NYU x DSC Datathon Olist Recommender System

1 | Initial Setup

• numpy and pandas for basic data-manipulation
• sklearn and associated modules to run our machine learning algorithms
• scipy associated modules (norm) to compute the cosime similarity later on
• seaborn to output graphs
• Miscellaneous packages for flow of the notebook

! pip install utils
! pip install scikit-learn

import pandas as pd
import numpy as np
from numpy.linalg import norm
import scipy as sc

from sklearn.neighbors import KNeighborsClassifier
from sklearn.impute import KNNImputer
from sklearn import preprocessing

from scipy.linalg import norm

import seaborn as sns

from typing import List
from utils import *
import re
from collections import Counter
import sys

• This notebook is connected to the associated Github repository for easy access and version contol
• Optionally, it is also tested and connected to Google Drive if possible

if 'google.colab' in sys.modules:
    %cd /content
    !rm -rf Datathon_F22
    !git clone https://github.com/sunnydigital/Datathon_F22
    %cd Datathon_F22

if 'google.colab' in sys.modules:
    from google.colab import drive
    drive.mount('/content/drive')

• After initally reading in the product_category_name_translation.csv we need to transform the data into a useable format
• An initial inspection of the number of products uder each category is displayed
• The english translations for product names are given (from the associated column in the product_category_name_translation.csv)
• For each category, the individual words making up the category is split using re.split

trans_df = pd.read_csv('datasets/product_category_name_translation.csv')

og_cat_name = trans_df['product_category_name'].to_numpy()
og_cat_name = np.append(og_cat_name, ['portateis_cozinha_e_preparadores_de_alimentos', 'pc_gamer', 'pcs', 'other'])

og_name_to_idx = {}
for i, name in enumerate(og_cat_name):
  og_name_to_idx[name] = i
display(og_name_to_idx)

product_categories = trans_df['product_category_name_english'].to_numpy()
product_categories = np.append(product_categories, ['portateis_cozinha_e_preparadores_de_alimentos', 'pc_gamer', 'pcs', 'other'])
display(product_categories)


product_categories_list = []
for category in product_categories:
   product_categories_list.append(re.split('_| ', category))

product_categories_list

{'beleza_saude': 0,
 'informatica_acessorios': 1,
 'automotivo': 2,
 'cama_mesa_banho': 3,
 'moveis_decoracao': 4,
 'esporte_lazer': 5,
 'perfumaria': 6,
 'utilidades_domesticas': 7,
 'telefonia': 8,
 'relogios_presentes': 9,
 'alimentos_bebidas': 10,
 'bebes': 11,
 'papelaria': 12,
 'tablets_impressao_imagem': 13,
 'brinquedos': 14,
 'telefonia_fixa': 15,
 'ferramentas_jardim': 16,
 'fashion_bolsas_e_acessorios': 17,
 'eletroportateis': 18,
 'consoles_games': 19,
 'audio': 20,
 'fashion_calcados': 21,
 'cool_stuff': 22,
 'malas_acessorios': 23,
 'climatizacao': 24,
 'construcao_ferramentas_construcao': 25,
 'moveis_cozinha_area_de_servico_jantar_e_jardim': 26,
 'construcao_ferramentas_jardim': 27,
 'fashion_roupa_masculina': 28,
 'pet_shop': 29,
 'moveis_escritorio': 30,
 'market_place': 31,
 'eletronicos': 32,
 'eletrodomesticos': 33,
 'artigos_de_festas': 34,
 'casa_conforto': 35,
 'construcao_ferramentas_ferramentas': 36,
 'agro_industria_e_comercio': 37,
 'moveis_colchao_e_estofado': 38,
 'livros_tecnicos': 39,
 'casa_construcao': 40,
 'instrumentos_musicais': 41,
 'moveis_sala': 42,
 'construcao_ferramentas_iluminacao': 43,
 'industria_comercio_e_negocios': 44,
 'alimentos': 45,
 'artes': 46,
 'moveis_quarto': 47,
 'livros_interesse_geral': 48,
 'construcao_ferramentas_seguranca': 49,
 'fashion_underwear_e_moda_praia': 50,
 'fashion_esporte': 51,
 'sinalizacao_e_seguranca': 52,
 'pcs': 73,
 'artigos_de_natal': 54,
 'fashion_roupa_feminina': 55,
 'eletrodomesticos_2': 56,
 'livros_importados': 57,
 'bebidas': 58,
 'cine_foto': 59,
 'la_cuisine': 60,
 'musica': 61,
 'casa_conforto_2': 62,
 'portateis_casa_forno_e_cafe': 63,
 'cds_dvds_musicais': 64,
 'dvds_blu_ray': 65,
 'flores': 66,
 'artes_e_artesanato': 67,
 'fraldas_higiene': 68,
 'fashion_roupa_infanto_juvenil': 69,
 'seguros_e_servicos': 70,
 'portateis_cozinha_e_preparadores_de_alimentos': 71,
 'pc_gamer': 72,
 'other': 74}



array(['health_beauty', 'computers_accessories', 'auto', 'bed_bath_table',
       'furniture_decor', 'sports_leisure', 'perfumery', 'housewares',
       'telephony', 'watches_gifts', 'food_drink', 'baby', 'stationery',
       'tablets_printing_image', 'toys', 'fixed_telephony',
       'garden_tools', 'fashion_bags_accessories', 'small_appliances',
       'consoles_games', 'audio', 'fashion_shoes', 'cool_stuff',
       'luggage_accessories', 'air_conditioning',
       'construction_tools_construction',
       'kitchen_dining_laundry_garden_furniture',
       'costruction_tools_garden', 'fashion_male_clothing', 'pet_shop',
       'office_furniture', 'market_place', 'electronics',
       'home_appliances', 'party_supplies', 'home_confort',
       'costruction_tools_tools', 'agro_industry_and_commerce',
       'furniture_mattress_and_upholstery', 'books_technical',
       'home_construction', 'musical_instruments',
       'furniture_living_room', 'construction_tools_lights',
       'industry_commerce_and_business', 'food', 'art',
       'furniture_bedroom', 'books_general_interest',
       'construction_tools_safety', 'fashion_underwear_beach',
       'fashion_sport', 'signaling_and_security', 'computers',
       'christmas_supplies', 'fashio_female_clothing',
       'home_appliances_2', 'books_imported', 'drinks', 'cine_photo',
       'la_cuisine', 'music', 'home_comfort_2',
       'small_appliances_home_oven_and_coffee', 'cds_dvds_musicals',
       'dvds_blu_ray', 'flowers', 'arts_and_craftmanship',
       'diapers_and_hygiene', 'fashion_childrens_clothes',
       'security_and_services',
       'portateis_cozinha_e_preparadores_de_alimentos', 'pc_gamer', 'pcs',
       'other'], dtype=object)





[['health', 'beauty'],
 ['computers', 'accessories'],
 ['auto'],
 ['bed', 'bath', 'table'],
 ['furniture', 'decor'],
 ['sports', 'leisure'],
 ['perfumery'],
 ['housewares'],
 ['telephony'],
 ['watches', 'gifts'],
 ['food', 'drink'],
 ['baby'],
 ['stationery'],
 ['tablets', 'printing', 'image'],
 ['toys'],
 ['fixed', 'telephony'],
 ['garden', 'tools'],
 ['fashion', 'bags', 'accessories'],
 ['small', 'appliances'],
 ['consoles', 'games'],
 ['audio'],
 ['fashion', 'shoes'],
 ['cool', 'stuff'],
 ['luggage', 'accessories'],
 ['air', 'conditioning'],
 ['construction', 'tools', 'construction'],
 ['kitchen', 'dining', 'laundry', 'garden', 'furniture'],
 ['costruction', 'tools', 'garden'],
 ['fashion', 'male', 'clothing'],
 ['pet', 'shop'],
 ['office', 'furniture'],
 ['market', 'place'],
 ['electronics'],
 ['home', 'appliances'],
 ['party', 'supplies'],
 ['home', 'confort'],
 ['costruction', 'tools', 'tools'],
 ['agro', 'industry', 'and', 'commerce'],
 ['furniture', 'mattress', 'and', 'upholstery'],
 ['books', 'technical'],
 ['home', 'construction'],
 ['musical', 'instruments'],
 ['furniture', 'living', 'room'],
 ['construction', 'tools', 'lights'],
 ['industry', 'commerce', 'and', 'business'],
 ['food'],
 ['art'],
 ['furniture', 'bedroom'],
 ['books', 'general', 'interest'],
 ['construction', 'tools', 'safety'],
 ['fashion', 'underwear', 'beach'],
 ['fashion', 'sport'],
 ['signaling', 'and', 'security'],
 ['computers'],
 ['christmas', 'supplies'],
 ['fashio', 'female', 'clothing'],
 ['home', 'appliances', '2'],
 ['books', 'imported'],
 ['drinks'],
 ['cine', 'photo'],
 ['la', 'cuisine'],
 ['music'],
 ['home', 'comfort', '2'],
 ['small', 'appliances', 'home', 'oven', 'and', 'coffee'],
 ['cds', 'dvds', 'musicals'],
 ['dvds', 'blu', 'ray'],
 ['flowers'],
 ['arts', 'and', 'craftmanship'],
 ['diapers', 'and', 'hygiene'],
 ['fashion', 'childrens', 'clothes'],
 ['security', 'and', 'services'],
 ['portateis', 'cozinha', 'e', 'preparadores', 'de', 'alimentos'],
 ['pc', 'gamer'],
 ['pcs'],
 ['other']]

2 | GloVe Embeddings

• An Indexer Object is created to map words to a dictionary
• read_wrod_embeddings is used to interface between the stored GloVe-50d embeddings in the file glove.6B.50d-relativized.txt and the dictionary protocal defined previously in the Indexer Object
• Finally an indexer object, word_indexer and the embedded dictionary embeddings object is returned

class Indexer(object):
    """
    Bijection between objects and integers starting at 0. Useful for mapping
    labels, features, etc. into coordinates of a vector space.

    Attributes:
        @objs_to_ints
        @ints_to_objs
    """
    def __init__(self):
        self.objs_to_ints = {}
        self.ints_to_objs = {}

    def __repr__(self):
        return str([str(self.get_object(i)) for i in range(0, len(self))])

    def __str__(self):
        return self.__repr__()

    def __len__(self):
        return len(self.objs_to_ints)

    def get_object(self, index):
        """
            @param index: integer index to look up
            @return: Returns the object corresponding to the particular index or None if not found
        """
        if (index not in self.ints_to_objs):
            return None
        else:
            return self.ints_to_objs[index]

    def contains(self, object):
        """
            @param object: object to look up
            @return: Returns True if it is in the Indexer, False otherwise
        """
        return self.index_of(object) != -1

    def index_of(self, object):
        """
            @param object: object to look up
            @return: Returns -1 if the object isn't present, index otherwise
        """
        if (object not in self.objs_to_ints):
            return -1
        else:
            return self.objs_to_ints[object]

    def add_and_get_index(self, object, add=True):
        """
        Adds the object to the index if it isn't present, always returns a nonnegative index
            @param object: object to look up or add
            @param add: True by default, False if we shouldn't add the object. If False, equivalent to index_of.
            @return: The index of the object
        """
        if not add:
            return self.index_of(object)
        if (object not in self.objs_to_ints):
            new_idx = len(self.objs_to_ints)
            self.objs_to_ints[object] = new_idx
            self.ints_to_objs[new_idx] = object
        return self.objs_to_ints[object]

def read_word_embeddings(embeddings_file: str):
    """
    Loads the given embeddings (ASCII-formatted) into a WordEmbeddings object. Augments this with an UNK embedding
    that is the 0 vector. Reads in all embeddings with no filtering -- you should only use this for relativized
    word embedding files.
        @param embeddings_file: path to the file containing embeddings
        @return: WordEmbeddings object reflecting the words and their embeddings
    """
    f = open(embeddings_file)
    word_indexer = Indexer()
    vectors = []
    # Make position 0 a PAD token, which can be useful if you
    word_indexer.add_and_get_index("PAD")
    # Make position 1 the UNK token
    word_indexer.add_and_get_index("UNK")
    for line in f:
        if line.strip() != "":
            space_idx = line.find(' ')
            word = line[:space_idx]
            numbers = line[space_idx+1:]
            float_numbers = [float(number_str) for number_str in numbers.split()]
            vector = np.array(float_numbers)
            word_indexer.add_and_get_index(word)
            # Append the PAD and UNK vectors to start. Have to do this weirdly because we need to read the first line
            # of the file to see what the embedding dim is
            if len(vectors) == 0:
                vectors.append(np.zeros(vector.shape[0]))
                vectors.append(np.zeros(vector.shape[0]))
            vectors.append(vector)
    f.close()
    print("Read in " + repr(len(word_indexer)) + " vectors of size " + repr(vectors[0].shape[0]))
    # Turn vectors into a 2-D numpy array
    return word_indexer, np.array(vectors)

word_indexer, embeddings = read_word_embeddings('glove/glove.6B.50d-relativized.txt')

Read in 14923 vectors of size 50

• get_embedding is a functon defined to obtain the embedding for any product category, meant to deal with multiple words describing any category
• Two ways of dealing with multi-word properties of product categories are included:
  • mean where the mean of embedding values for each word in the the GloVe-50d dictionary is found, for each word across all 50 dimensions
  • sum where the sum of embedding values for each word in the GloVe-50d dictionary is found, for each word, summed across all 50 dimensions
• The embedding for each product category is returned based on the previously determined methodology for word handling

def get_embedding(word_indexer, embeddings, product_categories, reduce='mean'):
    """
        @param word_indexer returns the index of the word in the embedding
        @param embeddings the embeddings matrix of all words in GloVe-50d
        @param product_categories the categories of all products represented in a [N x 1] array
    """
    category_embeddings = {}
    for category, category_list in zip(product_categories, product_categories_list):
        category_embed = []
        for word in category_list:
            index_word = word_indexer.index_of(word)
            embed = embeddings[index_word]
            category_embed.append(embed)
        if reduce == 'mean': 
            embed = np.mean(category_embed, axis=0)
        if reduce == 'sum':
            embed = np.sum(category_embed, axis=0)
        category_embeddings[category] = embed
    
    return category_embeddings

embeddings = get_embedding(word_indexer, embeddings, product_categories)
embeddings

{'perfumery': array([-0.45596  , -0.13718  , -0.23754  , -0.63402  , -0.10487  ,
        -0.42744  ,  0.83699  ,  0.23567  , -0.078916 ,  0.5      ,
         0.08264  ,  0.2316   ,  0.77999  ,  0.47049  , -0.32597  ,
        -0.51283  ,  0.30945  ,  0.47881  ,  0.55381  , -0.24994  ,
        -0.47739  , -0.073831 ,  0.072328 , -0.10644  ,  0.25275  ,
         0.97813  , -0.38809  ,  0.29865  ,  0.29393  ,  0.35823  ,
        -1.1929   ,  0.071931 ,  0.034599 ,  0.027147 , -0.38162  ,
         0.73098  ,  0.062593 , -0.14562  , -0.38141  , -0.072509 ,
         0.20806  , -0.54812  , -0.21912  ,  0.51654  ,  0.64665  ,
        -0.65962  ,  0.10605  ,  0.17607  ,  0.29246  ,  0.0042172]),
 'artes': array([-0.45596  , -0.13718  , -0.23754  , -0.63402  , -0.10487  ,
        -0.42744  ,  0.83699  ,  0.23567  , -0.078916 ,  0.5      ,
         0.08264  ,  0.2316   ,  0.77999  ,  0.47049  , -0.32597  ,
        -0.51283  ,  0.30945  ,  0.47881  ,  0.55381  , -0.24994  ,
        -0.47739  , -0.073831 ,  0.072328 , -0.10644  ,  0.25275  ,
         0.97813  , -0.38809  ,  0.29865  ,  0.29393  ,  0.35823  ,
        -1.1929   ,  0.071931 ,  0.034599 ,  0.027147 , -0.38162  ,
         0.73098  ,  0.062593 , -0.14562  , -0.38141  , -0.072509 ,
         0.20806  , -0.54812  , -0.21912  ,  0.51654  ,  0.64665  ,
        -0.65962  ,  0.10605  ,  0.17607  ,  0.29246  ,  0.0042172]),
 'sport_leisure': array([-6.5777500e-01,  2.3660000e-01, -5.2690000e-01, -1.0300000e-03,
        -3.7766500e-01, -2.4336400e-01,  1.1789000e-01, -2.6644000e-01,
         1.3412000e-02,  2.1341800e-01,  5.8577000e-01,  6.4220000e-02,
        -8.8555000e-02,  1.7286500e-01,  2.5672500e-01, -3.4311500e-01,
         2.7832000e-01,  1.2754050e+00,  1.1235000e-02, -8.0732000e-01,
        -6.4122500e-01,  1.1676450e-01, -2.9102100e-01,  4.8428000e-01,
         1.5141000e-01, -1.3723500e-01, -6.3680000e-01,  3.7784500e-01,
         1.5137115e-01,  3.4792000e-01,  5.6390000e-01,  6.3726550e-01,
        -2.4579550e-01, -9.4106500e-02, -2.6462000e-01,  4.5140000e-01,
        -4.5003500e-02,  1.4046500e-01, -5.2468500e-01, -2.7534950e-01,
         1.5362100e-01, -5.9103000e-01,  6.8940000e-02,  4.5214500e-01,
         2.8340550e-01, -3.7257450e-01, -1.5898500e-01,  2.5181000e-01,
         5.0931500e-01, -1.2091640e-01]),
 'babies': array([ 0.69898 , -0.056312,  0.34178 , -1.4674  ,  0.54603 ,  1.7752  ,
        -0.57618 , -0.60623 ,  0.83499 ,  0.42334 ,  0.50787 , -0.57778 ,
         1.0339  ,  0.27791 ,  1.4515  ,  0.58806 , -0.96275 ,  0.33018 ,
         0.45507 , -0.24377 , -0.15741 ,  0.95016 ,  0.71083 ,  0.23539 ,
         0.66137 , -0.61362 , -0.50664 , -0.24646 ,  0.32122 , -0.50601 ,
         1.5551  ,  1.1833  , -0.19435 ,  0.40841 ,  0.50837 ,  0.11851 ,
         0.82978 , -0.22482 ,  0.84309 , -0.27989 , -1.2027  , -0.50551 ,
         0.49102 ,  0.5024  ,  1.5571  , -0.025314, -0.032675, -0.68783 ,
        -0.072284, -0.22732 ]),
 'domestic_utilities': array([-0.2169885 , -0.259665  , -0.089036  , -0.232085  ,  0.02123   ,
         0.02266   ,  0.303745  , -0.30865   ,  0.304807  ,  0.24873115,
         0.27609   ,  0.182715  ,  0.44881   ,  0.297305  ,  0.20334   ,
        -0.2240425 ,  0.0246    ,  0.3544    ,  0.21105   , -0.38053   ,
         0.289055  , -0.3395355 , -0.114346  , -0.12142   , -0.0075    ,
        -0.118635  ,  0.00909   ,  0.02115   ,  0.324625  ,  0.756815  ,
         1.0631    ,  0.6046155 ,  0.4585545 , -0.0684765 , -0.31949   ,
         0.4114665 , -0.3848335 , -0.1004675 , -0.565625  , -0.2515545 ,
        -0.339595  , -0.245639  ,  0.28604   ,  0.173715  ,  0.44828   ,
        -0.3250159 , -0.2789    ,  0.435085  ,  0.45348   ,  0.0225586 ]),
 'musical instruments': array([-0.06468   ,  0.322055  , -0.963365  , -0.114635  , -0.177675  ,
         0.54965   , -0.1838395 , -0.4549    , -0.64644   ,  1.3871    ,
         0.57586   ,  0.94313   ,  0.151145  ,  1.07977   , -0.49424   ,
        -0.380291  , -0.17956   ,  0.1215325 , -0.297625  , -0.85511   ,
         0.865405  ,  0.562785  , -0.186045  , -0.499275  ,  0.152454  ,
        -0.0582182 , -1.35372   , -0.50429   , -0.154655  , -0.653355  ,
         2.9801    , -0.507645  ,  0.638315  , -0.33097815,  0.5617    ,
         0.35252   ,  0.186936  ,  0.0288775 , -0.228075  ,  0.122783  ,
         0.05871   ,  0.03846   ,  0.147365  , -0.122755  , -0.14569   ,
         0.339465  ,  0.54428   , -0.093315  , -1.1327    ,  0.159625  ]),
 'cool_stuff': array([-4.932950e-01,  2.035540e-01, -8.313395e-02, -4.249600e-01,
         9.794000e-02, -4.397210e-01, -3.272700e-01, -2.336160e-01,
        -3.763900e-01,  6.511450e-01, -5.402300e-01,  3.683300e-01,
        -1.061500e-02,  6.245650e-01,  5.724015e-01,  5.704455e-01,
         3.654400e-01,  4.361400e-01, -1.085555e-01, -1.033350e+00,
        -5.114250e-02,  5.775950e-01,  8.279250e-01,  3.179500e-01,
         4.330600e-01, -8.627450e-01, -8.960600e-01,  1.115915e+00,
         1.720000e+00, -2.354450e-01,  2.476200e+00,  2.326425e-01,
        -7.899500e-03,  1.246350e-01, -1.432040e-01,  5.095900e-01,
        -3.377350e-01,  3.771100e-01,  3.613050e-01, -7.046750e-01,
         1.961800e-01,  3.419200e-01, -9.439450e-02,  3.296600e-01,
         4.638600e-01,  4.373350e-01,  4.684150e-01, -4.642925e-01,
         2.028000e-03,  6.441700e-01]),
 'furniture_decoration': array([-0.45596  , -0.13718  , -0.23754  , -0.63402  , -0.10487  ,
        -0.42744  ,  0.83699  ,  0.23567  , -0.078916 ,  0.5      ,
         0.08264  ,  0.2316   ,  0.77999  ,  0.47049  , -0.32597  ,
        -0.51283  ,  0.30945  ,  0.47881  ,  0.55381  , -0.24994  ,
        -0.47739  , -0.073831 ,  0.072328 , -0.10644  ,  0.25275  ,
         0.97813  , -0.38809  ,  0.29865  ,  0.29393  ,  0.35823  ,
        -1.1929   ,  0.071931 ,  0.034599 ,  0.027147 , -0.38162  ,
         0.73098  ,  0.062593 , -0.14562  , -0.38141  , -0.072509 ,
         0.20806  , -0.54812  , -0.21912  ,  0.51654  ,  0.64665  ,
        -0.65962  ,  0.10605  ,  0.17607  ,  0.29246  ,  0.0042172]),
 'home appliances': array([-0.11147   ,  0.160835  ,  0.01424   , -0.297908  ,  0.231035  ,
        -0.46969   , -0.473055  ,  0.333195  , -0.0387957 ,  0.258926  ,
        -0.216425  , -0.15653   ,  0.26442   ,  0.047085  ,  0.04186   ,
        -0.221237  ,  0.024125  ,  0.23801655,  0.0201    ,  0.02417   ,
        -0.166215  ,  0.2318245 , -0.276051  ,  0.153675  ,  0.016345  ,
        -0.324785  , -0.048915  ,  0.14006   ,  0.401985  , -0.12704   ,
         1.02255   ,  0.4103305 ,  0.1114745 ,  0.024657  ,  0.09831   ,
         0.6256    ,  0.1077315 ,  0.135195  ,  0.085595  , -0.2381195 ,
         0.013745  , -0.16949   , -0.228925  ,  0.31469   ,  0.319578  ,
        -0.04022   , -0.16518   , -0.316485  ,  0.24828   , -0.0955514 ]),
 'toys': array([ 0.11738 , -0.78864 ,  0.20307 ,  0.40609 ,  0.53572 ,  0.66853 ,
        -1.0303  , -2.0666  ,  0.67746 ,  0.41301 , -0.036541,  0.57572 ,
         0.2052  , -0.13067 ,  1.1972  ,  0.69488 , -0.7459  ,  1.0515  ,
         0.22374 , -1.0436  ,  1.1563  ,  0.12272 , -0.091888, -0.070431,
        -0.35183 , -0.37046 , -1.4255  , -0.078561,  0.716   , -1.3362  ,
         1.3186  ,  0.995   ,  0.52411 ,  0.38252 ,  0.73561 ,  1.1922  ,
         0.30399 , -0.19526 , -0.036646, -0.48951 , -0.36473 ,  0.2109  ,
         0.16709 ,  0.18673 ,  1.3641  ,  1.1111  , -0.17835 , -0.31437 ,
         0.12497 , -0.14976 ]),
 'bed table bath': array([ 0.121282  ,  1.07551   , -0.50220667, -0.33427333,  0.37935   ,
        -0.153277  , -0.141247  ,  0.02063833, -0.05749   , -0.60773667,
        -0.4423    ,  0.2403    , -0.190591  ,  0.27271   ,  0.21277667,
         0.86095667,  0.09901333,  0.06610467, -0.13987933, -0.73049667,
         0.68601   ,  0.55938333,  0.48579333,  0.25348   , -0.37527667,
        -0.423359  , -0.130161  ,  0.77310667,  0.42638333,  0.09753433,
         2.26126667,  0.27597467, -0.23237503,  0.67459667,  0.37090333,
         0.82384333,  0.57042   ,  0.48660667,  0.932469  , -0.33592333,
         0.46510333, -0.20701167, -0.30365333,  0.71547   ,  0.31835667,
         0.42810333,  0.18425667, -1.03925233,  0.38622333, -0.54126   ]),
 'construction_tools_security': array([ 0.50816667, -0.15552927,  0.61994   ,  0.21229467, -0.02190367,
        -0.32106667, -0.30894   , -0.80907333,  0.16325333, -0.592376  ,
        -0.16282   , -0.10930333, -0.347921  , -0.05749967, -0.32407333,
         0.31827333,  0.13833333,  0.25797333,  0.58916667, -0.494492  ,
         0.94635   , -0.05304667, -0.88858   , -0.78261   , -0.048928  ,
        -1.00035333,  0.02441667, -0.399411  ,  0.35588333,  0.09051667,
         3.3164    , -0.50588333, -0.39591667, -0.65001233,  0.01619   ,
         0.8768    ,  0.254316  ,  0.20291667, -0.09791267,  0.423946  ,
        -0.13481333, -0.37350333,  0.4858    ,  0.20931667,  0.53948267,
        -0.29450233, -0.04733667,  0.60774667,  0.18903   , -0.15894333]),
 'informatica_accessories': array([-0.45596  , -0.13718  , -0.23754  , -0.63402  , -0.10487  ,
        -0.42744  ,  0.83699  ,  0.23567  , -0.078916 ,  0.5      ,
         0.08264  ,  0.2316   ,  0.77999  ,  0.47049  , -0.32597  ,
        -0.51283  ,  0.30945  ,  0.47881  ,  0.55381  , -0.24994  ,
        -0.47739  , -0.073831 ,  0.072328 , -0.10644  ,  0.25275  ,
         0.97813  , -0.38809  ,  0.29865  ,  0.29393  ,  0.35823  ,
        -1.1929   ,  0.071931 ,  0.034599 ,  0.027147 , -0.38162  ,
         0.73098  ,  0.062593 , -0.14562  , -0.38141  , -0.072509 ,
         0.20806  , -0.54812  , -0.21912  ,  0.51654  ,  0.64665  ,
        -0.65962  ,  0.10605  ,  0.17607  ,  0.29246  ,  0.0042172]),
 'beauty_health': array([ 0.149462 ,  0.837515 , -0.658989 ,  0.11762  ,  0.156495 ,
         0.311425 , -0.2985085, -0.72018  ,  1.114425 ,  0.224965 ,
         0.18148  , -0.1437515,  0.36231  , -0.316605 ,  0.29906  ,
         0.187535 , -0.318833 ,  0.19092  ,  0.534318 , -0.15734  ,
        -0.095259 ,  1.056855 , -0.295065 , -0.079285 ,  0.14422  ,
        -1.22806  , -1.06188  , -0.15872  , -0.05257  ,  0.38244  ,
         2.6115   ,  0.681295 , -0.01553  , -0.803845 , -0.079915 ,
         0.240563 , -0.286601 ,  0.450905 ,  0.700525 , -0.472489 ,
        -0.326875 ,  0.212085 ,  0.4923   ,  0.25222  ,  0.270245 ,
         0.37235  , -0.448995 , -0.188735 ,  0.56276  ,  0.34234  ]),
 'bags_accessories': array([-0.09753  , -0.21828  ,  0.39673  , -0.541435 ,  0.28565  ,
        -0.250009 ,  0.16389  , -0.432615 ,  0.587842 , -0.11338  ,
         0.13238  ,  0.034195 ,  0.52179  ,  0.371215 ,  0.406465 ,
         0.015975 , -0.30545  ,  0.43989  ,  0.361695 , -1.02972  ,
         0.11384  , -0.02404  ,  0.438889 , -0.109875 , -0.292865 ,
         0.148435 , -0.41799  ,  0.785775 ,  0.695215 , -0.208865 ,
         0.1957   ,  0.6654155, -0.2863305,  0.4026785,  0.102555 ,
         1.22624  ,  0.2838865,  0.30733  ,  0.079865 ,  0.0776355,
         0.35376  ,  0.099455 ,  0.17175  ,  0.439675 ,  0.51477  ,
        -0.338477 , -0.051525 , -0.07325  ,  0.385935 , -0.5949414]),
 'garden_tools': array([-0.13122  , -0.45077  ,  0.20832  , -0.15525  , -0.0177905,
        -0.273765 ,  0.207975 , -0.589265 ,  0.100587 ,  0.265581 ,
         0.0123335,  0.303705 ,  0.4274135,  0.2284005, -0.441685 ,
        -0.167755 ,  0.30533  ,  0.4014   ,  0.67399  , -0.67052  ,
         0.135475 , -0.0920255, -0.201911 , -0.296385 ,  0.375905 ,
         0.38115  , -0.546515 , -0.447175 ,  0.41337  , -0.095075 ,
         0.8753   , -0.3589695, -0.2505255, -0.3293365, -0.10512  ,
         1.16314  , -0.0027195,  0.341125 , -0.140954 , -0.0208605,
         0.459025 , -0.554825 , -0.05076  ,  0.85532  ,  1.007625 ,
        -0.3457835,  0.583475 ,  0.42408  ,  0.202085 , -0.1239964]),
 'furniture_office': array([-0.100575  , -0.0411355 ,  0.043105  , -0.209865  ,  0.18825   ,
        -0.59084   ,  0.02806   , -0.02936   , -0.0272195 , -0.39515   ,
        -0.306055  , -0.054165  ,  0.205645  ,  0.2346751 ,  0.092645  ,
        -0.11909   , -0.277255  ,  0.274721  ,  0.58935   , -0.1465255 ,
         0.294605  ,  0.0838895 , -0.231836  , -0.210795  , -0.24414   ,
        -0.562935  , -0.1818305 ,  0.36765   , -0.19358   ,  0.27433   ,
         0.8427    , -0.0921145 , -0.2660805 , -0.5448265 , -0.340315  ,
         0.51402   ,  0.2321565 ,  0.036925  ,  0.131865  , -0.1284995 ,
        -0.039955  ,  0.025175  , -0.247405  ,  0.25581025, -0.044475  ,
        -0.40652   , -0.442715  ,  0.0650475 ,  0.508945  ,  0.2105486 ]),
 'automotive': array([-0.45596  , -0.13718  , -0.23754  , -0.63402  , -0.10487  ,
        -0.42744  ,  0.83699  ,  0.23567  , -0.078916 ,  0.5      ,
         0.08264  ,  0.2316   ,  0.77999  ,  0.47049  , -0.32597  ,
        -0.51283  ,  0.30945  ,  0.47881  ,  0.55381  , -0.24994  ,
        -0.47739  , -0.073831 ,  0.072328 , -0.10644  ,  0.25275  ,
         0.97813  , -0.38809  ,  0.29865  ,  0.29393  ,  0.35823  ,
        -1.1929   ,  0.071931 ,  0.034599 ,  0.027147 , -0.38162  ,
         0.73098  ,  0.062593 , -0.14562  , -0.38141  , -0.072509 ,
         0.20806  , -0.54812  , -0.21912  ,  0.51654  ,  0.64665  ,
        -0.65962  ,  0.10605  ,  0.17607  ,  0.29246  ,  0.0042172]),
 'electronic': array([-6.6594e-01, -6.5510e-01,  9.2148e-01,  6.8301e-01, -4.7822e-01,
        -2.8030e-01, -5.3848e-01, -1.0404e+00,  8.6666e-01,  9.4106e-01,
         6.8266e-01,  1.6785e-01,  2.7838e-01,  1.1826e+00, -4.1586e-01,
        -5.3318e-02, -1.2734e+00, -2.2728e-01,  8.7516e-02, -9.1505e-01,
         1.7087e+00, -1.1551e-01, -7.6449e-01,  2.7068e-02, -2.9328e-02,
        -7.2308e-01, -7.4972e-01, -7.2909e-01,  4.0463e-01, -5.0634e-01,
         2.9845e+00, -5.8706e-01,  3.0288e-01, -5.0952e-01, -3.2629e-01,
         4.3184e-01,  2.9218e-01, -2.9649e-01, -1.7555e-03,  1.2518e-01,
         1.0972e+00,  1.6128e-01,  2.2262e-01,  6.7454e-01,  1.5689e-01,
         2.6961e-01,  7.8616e-01,  3.6450e-01, -4.2447e-01, -5.8735e-02]),
 'fashion_shoes': array([-0.529271  ,  0.187405  , -0.706595  , -0.150854  ,  0.257785  ,
        -0.062465  , -0.64265   , -1.31045   ,  0.1302085 ,  0.121555  ,
         0.15193675, -0.2591685 , -0.87015   ,  0.13515   ,  0.622815  ,
         0.30194   , -0.52321   ,  0.184355  ,  0.289615  , -1.27407   ,
         0.267221  ,  0.79496   , -0.58884   ,  0.123875  , -0.49169   ,
        -0.965705  , -1.28342   ,  0.82245   ,  0.54177   , -0.765955  ,
         1.8377    ,  0.77242   ,  0.12483   ,  0.23106   , -0.020515  ,
         0.487787  , -0.031645  ,  0.97589   , -0.577765  , -0.24196   ,
         0.33837   ,  0.358545  ,  0.41685   ,  0.10506   ,  0.753915  ,
        -0.498315  ,  0.04316   , -0.85085   ,  0.0436    ,  0.16844   ]),
 'telephony': array([-0.45596  , -0.13718  , -0.23754  , -0.63402  , -0.10487  ,
        -0.42744  ,  0.83699  ,  0.23567  , -0.078916 ,  0.5      ,
         0.08264  ,  0.2316   ,  0.77999  ,  0.47049  , -0.32597  ,
        -0.51283  ,  0.30945  ,  0.47881  ,  0.55381  , -0.24994  ,
        -0.47739  , -0.073831 ,  0.072328 , -0.10644  ,  0.25275  ,
         0.97813  , -0.38809  ,  0.29865  ,  0.29393  ,  0.35823  ,
        -1.1929   ,  0.071931 ,  0.034599 ,  0.027147 , -0.38162  ,
         0.73098  ,  0.062593 , -0.14562  , -0.38141  , -0.072509 ,
         0.20806  , -0.54812  , -0.21912  ,  0.51654  ,  0.64665  ,
        -0.65962  ,  0.10605  ,  0.17607  ,  0.29246  ,  0.0042172]),
 'stationary store': array([ 0.431825 , -0.159595 ,  0.7213   ,  0.0359543,  0.5905305,
        -0.276996 , -0.849492 , -0.3911   ,  0.629875 , -0.1776875,
         0.5078325, -0.59475  , -0.20207  ,  0.137035 ,  0.00635  ,
         0.40943  , -0.61595  ,  0.677255 , -0.25467  , -1.192875 ,
         0.8558665, -0.411145 , -0.565025 ,  0.826405 , -0.09058  ,
        -0.349225 , -0.38166  ,  0.59205  ,  1.1161   , -0.07926  ,
         1.7939   , -0.210345 ,  0.169255 , -0.096005 ,  0.29153  ,
         0.1508485,  0.194135 ,  0.55134  ,  0.201635 ,  0.543336 ,
         0.490335 , -0.2112085, -0.356055 ,  0.361025 , -0.00265  ,
        -0.22146  ,  0.32249  , -0.780135 ,  0.39682  , -0.009155 ]),
 'fashion_bolsas_e_acessorios': array([-0.3612355 , -0.0531225 , -0.496455  , -0.4822595 , -0.00221   ,
        -0.3775225 ,  0.4505125 , -0.1233975 , -0.08142025,  0.5212425 ,
         0.06148088,  0.0285875 ,  0.3185425 ,  0.4761475 , -0.11933   ,
        -0.41105   ,  0.146635  ,  0.276585  ,  0.4649475 , -0.38809   ,
        -0.2147625 ,  0.12867925, -0.1729265 ,  0.103575  ,  0.078905  ,
         0.530345  , -0.7632925 ,  0.3161625 ,  0.29196   ,  0.237145  ,
        -0.360725  ,  0.17938325,  0.11933925, -0.11785975, -0.3642725 ,
         0.566631  , -0.04145275,  0.16126   , -0.4949825 , -0.13645425,
         0.2658775 , -0.3131925 , -0.1265925 ,  0.33689   ,  0.6194475 ,
        -0.5635775 ,  0.15796   , -0.0090975 ,  0.30098   ,  0.2050404 ]),
 'pcs': array([-0.45596  , -0.13718  , -0.23754  , -0.63402  , -0.10487  ,
        -0.42744  ,  0.83699  ,  0.23567  , -0.078916 ,  0.5      ,
         0.08264  ,  0.2316   ,  0.77999  ,  0.47049  , -0.32597  ,
        -0.51283  ,  0.30945  ,  0.47881  ,  0.55381  , -0.24994  ,
        -0.47739  , -0.073831 ,  0.072328 , -0.10644  ,  0.25275  ,
         0.97813  , -0.38809  ,  0.29865  ,  0.29393  ,  0.35823  ,
        -1.1929   ,  0.071931 ,  0.034599 ,  0.027147 , -0.38162  ,
         0.73098  ,  0.062593 , -0.14562  , -0.38141  , -0.072509 ,
         0.20806  , -0.54812  , -0.21912  ,  0.51654  ,  0.64665  ,
        -0.65962  ,  0.10605  ,  0.17607  ,  0.29246  ,  0.0042172]),
 'house_construction': array([ 0.70555   ,  0.293185  ,  0.179376  , -0.08316   ,  0.05319   ,
         0.189115  , -0.89585   , -0.2394795 , -0.44347   , -0.729345  ,
        -0.9164635 , -0.29398   , -0.1271705 , -0.20140725, -0.0930295 ,
         0.31127   ,  0.028011  ,  0.237485  ,  0.254165  , -0.539355  ,
         1.42785   , -0.197095  , -1.077485  , -0.758905  , -0.204447  ,
        -1.257615  , -0.097675  ,  0.124841  , -0.017775  ,  0.153878  ,
         3.13085   , -0.98483   , -0.22145   , -0.2723735 ,  0.171395  ,
         0.0940355 ,  0.41796   ,  0.220105  ,  0.33042   ,  0.07027   ,
        -0.579885  , -0.345695  , -0.00575   , -0.055565  ,  0.007649  ,
         0.01957   , -0.75081   , -0.545575  ,  0.154991  , -0.043825  ]),
 'watches_gifts': array([ 0.13938  ,  0.490805 , -0.10947  ,  0.016295 ,  0.826165 ,
        -0.07303  , -0.56479  , -0.260535 ,  0.559345 ,  0.360701 ,
        -0.179925 ,  0.609995 ,  0.801685 , -0.12088  ,  1.1018   ,
         0.113595 , -0.871335 ,  0.008045 ,  0.19502  , -0.88376  ,
         0.995565 ,  0.2234605, -0.1203995, -0.45073  ,  0.03505  ,
        -0.87461  , -0.81466  , -0.01929  ,  0.48275  , -0.923825 ,
         1.1388   ,  1.145715 ,  0.375925 ,  0.1000245,  0.015275 ,
         0.761995 , -0.1934265,  0.38571  , -0.425525 , -0.185705 ,
         0.298155 ,  0.665945 ,  0.1537   ,  0.508755 ,  0.110155 ,
        -0.35469  , -0.365415 , -0.245515 , -0.395905 , -0.26568  ]),
 'construction_tools_construction': array([ 6.04326667e-01, -5.40133333e-02,  4.78586667e-01,  2.83800000e-01,
        -4.04690333e-01,  3.73033333e-02, -6.85740000e-01, -8.43093333e-01,
        -1.14010000e-01, -1.70072667e-01, -7.47566667e-03, -1.00790000e-01,
        -1.81647667e-01, -2.78849667e-01, -3.55140000e-01,  2.03773333e-01,
         1.87083333e-01,  5.66763333e-01,  3.55830000e-01, -6.04233333e-01,
         1.33764667e+00, -5.36506667e-01, -1.31845000e+00, -1.02777667e+00,
         1.58164000e-01, -5.60030000e-01, -8.85000000e-02, -4.28098667e-01,
         4.28930000e-01,  8.98000000e-03,  3.32730000e+00, -9.14663333e-01,
        -2.63333333e-04, -2.79684667e-01,  9.72200000e-02,  7.08906667e-01,
         3.31869333e-01,  6.13723333e-01,  2.10327333e-01,  5.75602667e-01,
        -2.77223333e-01, -8.02490000e-01,  3.61586667e-01,  2.26533333e-01,
         4.81825333e-01, -3.60482333e-01, -7.28600000e-02,  1.81796667e-01,
         2.17276667e-01, -2.75956667e-01]),
 'pet_shop': array([ 0.383305  , -0.22558   , -0.1947675 , -0.02225955,  0.552615  ,
         0.13605   , -1.42445   , -0.9175    ,  0.82867   ,  0.1530525 ,
        -0.018925  ,  0.265945  ,  0.60104   ,  0.45376   ,  0.504435  ,
         0.33318   , -0.437065  ,  0.640445  ,  0.09386   , -0.64959   ,
         0.901288  ,  0.2274495 , -0.5736005 ,  0.627565  , -0.20507   ,
        -1.05225   , -0.846905  ,  0.1250828 ,  0.569395  , -0.734295  ,
         1.6631    ,  0.17512   , -0.3570085 ,  0.62313   ,  0.11078   ,
         0.520045  ,  0.22995   ,  0.216055  ,  0.75598   , -0.06704   ,
         0.154995  ,  0.1381225 , -0.166715  ,  0.671815  ,  1.207155  ,
         0.16389   , -0.39704   , -0.644335  ,  0.89743   ,  0.186325  ]),
 'small appliances': array([ 3.4297000e-01,  3.9515000e-02, -8.8830000e-02, -4.2744500e-01,
         3.0806500e-01,  6.3250000e-02, -6.7200000e-03, -1.9590000e-02,
        -5.8000000e-05, -2.2235000e-02, -4.0970000e-02, -8.0975000e-02,
         8.6283000e-01,  3.9184500e-01, -4.4992500e-01, -2.5332900e-01,
         4.1254000e-01,  5.3558000e-01,  4.9550000e-03, -6.7982000e-01,
        -1.7003000e-01, -3.7142050e-01, -4.6756000e-02,  1.7112500e-01,
        -8.8070000e-02, -7.2185000e-02, -4.1208000e-01,  4.2770500e-01,
         3.4677500e-01,  4.9295000e-02,  1.3307000e+00, -1.2948950e-01,
         2.0801950e-01,  4.3545000e-02, -9.2545000e-02,  6.1857500e-01,
        -1.7431850e-01,  8.0300000e-03, -1.1555500e-01, -6.6786000e-02,
         1.0083015e-01, -1.8001000e-01, -1.2839150e-01,  4.0629500e-01,
         5.5394500e-01, -3.2650805e-01, -1.0066500e-01, -2.3265000e-01,
         1.1707450e-01, -2.2713140e-01]),
 'agro_industry_and_trade': array([-3.6232000e-01, -1.6120825e-01, -3.0137925e-01,  2.7522175e-01,
         3.0500000e-04, -1.9281200e-02, -4.3572750e-01, -5.8117250e-01,
         1.4762350e-01,  1.8315050e-01,  2.2254500e-01,  3.1581500e-01,
        -2.4672750e-02,  2.6693750e-01,  1.0409750e-01, -8.1837500e-03,
         2.3921475e-01,  2.3425000e-02,  3.2642500e-01, -4.5122400e-01,
         6.0444750e-01, -1.7562525e-01, -1.8230300e-01, -3.1608150e-01,
        -1.7505000e-02, -7.9666750e-01, -2.9927500e-02, -6.0552500e-02,
         2.2260000e-01,  1.8546000e-01,  2.4350500e+00,  1.0978275e-01,
         2.2210275e-01, -6.2197750e-02, -4.0215250e-01, -5.1220000e-02,
        -5.3526675e-01,  1.9079650e-01, -7.6040000e-02,  2.1851750e-01,
        -4.5848575e-01, -3.6085000e-02,  3.8913000e-01, -5.1767000e-02,
         4.7119500e-01, -4.3097500e-02, -4.2938000e-01,  2.7493000e-01,
         1.1408500e-01, -5.4280700e-02]),
 'in': array([ 0.33042  ,  0.24995  , -0.60874  ,  0.10923  ,  0.036372 ,
         0.151    , -0.55083  , -0.074239 , -0.092307 , -0.32821  ,
         0.09598  , -0.82269  , -0.36717  , -0.67009  ,  0.42909  ,
         0.016496 , -0.23573  ,  0.12864  , -1.0953   ,  0.43334  ,
         0.57067  , -0.1036   ,  0.20422  ,  0.078308 , -0.42795  ,
        -1.7984   , -0.27865  ,  0.11954  , -0.12689  ,  0.031744 ,
         3.8631   , -0.17786  , -0.082434 , -0.62698  ,  0.26497  ,
        -0.057185 , -0.073521 ,  0.46103  ,  0.30862  ,  0.12498  ,
        -0.48609  , -0.0080272,  0.031184 , -0.36576  , -0.42699  ,
         0.42164  , -0.11666  , -0.50703  , -0.027273 , -0.53285  ]),
 'living room furniture': array([ 1.67513333e-01,  4.02106667e-01, -7.22100000e-02, -7.26840000e-01,
         7.17410000e-01, -2.41123333e-01, -2.67453333e-01, -2.62083333e-01,
        -1.82842000e-01,  8.89833333e-02, -1.13746667e-01, -2.46113333e-01,
         6.51142667e-01,  3.34633333e-01,  4.22633333e-01,  1.46620000e-01,
         1.24803333e-01,  3.68426667e-01,  2.28036667e-01, -2.87353333e-01,
         3.40700000e-02,  7.51323000e-01, -1.59634000e-01,  8.91543333e-02,
         2.33490000e-01, -2.35480000e-01, -3.49280000e-01,  2.41256667e-01,
         4.24886667e-01,  3.46933333e-02,  1.59026667e+00,  2.18807000e-01,
        -4.28693333e-02, -3.34637667e-01, -3.76666667e-04,  6.62646667e-01,
         2.05654333e-01,  3.62533333e-02,  3.48250000e-01, -2.08463000e-01,
        -1.81054000e-01, -8.17216667e-02,  1.29733333e-01,  6.26040000e-01,
         1.02564667e-01, -1.90523333e-01, -1.45210000e-01, -6.81410000e-01,
         9.24056667e-02, -1.19250933e-01]),
 'signaling_and_security': array([ 1.11156667e-01,  9.64066667e-04,  9.95133333e-02, -1.89446333e-01,
         1.71883333e-01, -2.29106667e-01,  2.11993333e-01, -2.31163333e-01,
         3.70213333e-02, -3.91980000e-01, -3.18066667e-02,  5.87833333e-02,
        -1.59613333e-01,  1.03470000e-01, -1.95366667e-01,  2.85850000e-02,
         1.07656333e-01,  1.13783333e-01,  3.79966667e-01, -2.66565333e-01,
         1.08866667e-01,  3.58706333e-01,  3.67493333e-02, -2.50102000e-01,
        -1.60673333e-01, -8.17556667e-01,  2.05733333e-02,  3.56850000e-02,
         1.54873000e-01,  1.41806667e-01,  2.04663333e+00,  2.11610333e-01,
        -2.49500333e-01, -5.60964333e-01, -7.81233333e-02,  6.12480000e-01,
         1.91433333e-03, -1.82360000e-01, -3.06660000e-01,  2.23910333e-01,
        -6.81943333e-02,  6.15733333e-02,  2.96260000e-01, -1.25100000e-02,
         3.70623333e-01, -2.83243333e-01, -2.56300000e-01,  2.52823333e-01,
         7.58333333e-02, -1.04514267e-01]),
 'air conditioning': array([ 0.08439  ,  0.046005 ,  0.26098  , -0.27408  , -0.793375 ,
        -0.803605 , -0.1245665,  0.010855 ,  0.4847655, -0.353585 ,
         0.91126  ,  0.1183835, -0.063325 ,  0.872935 , -0.636735 ,
         0.54369  , -0.73477  ,  0.976155 , -0.937524 , -1.147215 ,
         0.059735 ,  0.38916  , -0.273395 , -0.6447   , -0.405955 ,
        -1.05922  , -0.005155 ,  0.604635 ,  1.055195 ,  0.8969   ,
         2.568    ,  0.68065  , -0.350425 , -0.405875 ,  0.188045 ,
         0.6840155,  0.685145 ,  0.197325 ,  0.460865 ,  0.616075 ,
         0.3456   ,  0.371473 ,  0.03534  ,  0.30622  , -0.018955 ,
         0.251155 ,  0.197905 ,  0.02026  , -0.889225 ,  0.85391  ]),
 'consoles_games': array([-0.670455  ,  0.318735  , -0.084189  ,  0.50554   , -0.189155  ,
        -0.334235  , -0.06092   , -0.120755  , -0.273838  ,  0.441805  ,
         0.054707  ,  0.018565  , -0.313205  ,  0.294875  ,  0.529165  ,
        -0.654975  ,  0.08529   ,  0.45546   , -0.714095  , -0.091431  ,
        -0.54849   , -0.1767905 ,  0.284039  ,  0.01622   ,  0.148317  ,
         0.132665  , -0.079865  , -0.1243    ,  0.354785  , -0.435235  ,
         1.04355   ,  0.8325155 ,  0.0698745 , -0.1962215 , -0.136555  ,
         0.71047   ,  0.4870315 ,  0.3173    , -0.665315  , -0.1829195 ,
        -0.135135  , -0.66811   , -0.17432   ,  0.580815  ,  0.115815  ,
         0.16498   ,  0.0049695 ,  0.1625    ,  0.14231545, -0.0138379 ]),
 'general_interest_books': array([-0.13378667,  0.40633   , -0.180478  , -0.32258   ,  0.57698667,
        -0.15125333, -0.81235333, -0.78724667, -0.17347333, -0.18801   ,
        -0.10828667,  0.84404333, -0.32388033, -0.50571   ,  0.66861067,
        -0.10638333, -0.25498667, -0.40634   ,  0.00961433, -0.27365133,
         0.75540633, -0.20779667,  0.14485397, -0.28896333, -0.15458667,
        -1.32330667, -0.607391  , -0.59106667, -0.26451667,  0.41909667,
         2.96093333, -0.07119   ,  0.06087667, -0.59353333, -0.0925    ,
        -0.23531667, -0.59946833,  0.08378333,  0.00778733, -0.10507333,
         0.04876   ,  0.14108   ,  0.00764667,  0.14884   , -0.19461133,
         0.05654733, -0.11801333,  0.47503367, -0.10037333,  0.214126  ]),
 'construction_tools_tools': array([ 0.39892333, -0.40918667,  0.56638333,  0.30366   , -0.16770067,
        -0.04139333, -0.55339   , -1.12864667,  0.08304   , -0.06945533,
        -0.03272433,  0.13751   , -0.05340533, -0.14626933, -0.45627   ,
         0.19054667,  0.24414667,  0.44537667,  0.575     , -0.84766667,
         1.04299333, -0.32336333, -0.8973    , -0.75705333,  0.328612  ,
        -0.38793   , -0.39672   , -0.81054933,  0.48087   , -0.2697    ,
         3.1354    , -0.85226667, -0.26795667, -0.48275233,  0.1343    ,
         1.15210333,  0.13191867,  0.72079667,  0.15491467,  0.30319533,
         0.21638333, -0.68201   ,  0.23959333,  0.71031667,  0.92521267,
        -0.19621467,  0.49402   ,  0.42694333,  0.16449333, -0.26408333]),
 'fashion_underwear_e_moda_praia': array([-0.4035864 , -0.074576  , -0.418086  , -0.3687174 ,  0.05163   ,
        -0.194718  ,  0.3565326 , -0.331838  ,  0.0718638 ,  0.35747   ,
         0.1141707 , -0.073368  ,  0.101738  ,  0.3884804 ,  0.048456  ,
        -0.15035   , -0.142352  ,  0.329326  ,  0.495374  , -0.508146  ,
        -0.11091   ,  0.2938714 , -0.0429112 ,  0.08110864, -0.133962  ,
         0.190236  , -0.884034  ,  0.48503   ,  0.476228  ,  0.004268  ,
        -0.240858  ,  0.2754786 , -0.0439746 , -0.0122178 , -0.208024  ,
         0.6074088 , -0.04877   ,  0.298082  , -0.523692  , -0.1423354 ,
         0.27784   , -0.189706  , -0.001552  ,  0.313756  ,  0.645318  ,
        -0.640938  ,  0.166912  , -0.243998  ,  0.271598  ,  0.07568232]),
 'fashion_clothes_men': array([-0.20766067,  0.18115   , -0.58055   , -0.36946267,  0.71611   ,
        -0.06227667, -0.480974  , -0.56409   , -0.18878433, -0.17501   ,
         0.1055745 , -0.55299   , -0.601128  , -0.04001   ,  0.75262667,
         0.03583333, -0.38519333,  0.06401   ,  0.00884333, -1.02360333,
         0.10532333,  1.05513667, -0.07654333,  0.40878733, -0.4867    ,
        -1.12438667, -1.28289333,  0.28985667,  0.52306667, -0.48607333,
         2.29503333,  0.7971    , -0.01128033, -0.1677    ,  0.21522667,
         0.67615133, -0.14004467,  0.78054667, -0.49585867, -0.15393933,
         0.16505333,  0.25155667,  0.61591   ,  0.17914333,  0.84348333,
        -0.51744   , -0.09429   , -0.87343   ,  0.09008   , -0.10341667]),
 'furniture_cozinha_area_de_servico_jantar_e_jardim': array([-0.12591   ,  0.05804125, -0.3714275 , -0.45058238, -0.17667313,
        -0.6250375 ,  0.4318925 ,  0.2268    , -0.097417  ,  0.32719125,
         0.15450375,  0.0841875 ,  0.55648   ,  0.065975  , -0.3158175 ,
        -0.53394125,  0.372565  ,  0.294245  ,  0.2417575 , -0.24151   ,
        -0.4950775 , -0.04474862,  0.01052183,  0.04897125,  0.0280925 ,
         0.58375375, -0.39934212,  0.327625  ,  0.249945  ,  0.282065  ,
        -0.1091875 , -0.13682175, -0.11248325, -0.11242225, -0.23426225,
         0.45630125,  0.04041513, -0.06066375, -0.14173625,  0.15078825,
         0.22956875, -0.37743613,  0.01378625,  0.17155463,  0.274235  ,
        -0.62712375, -0.06439375, -0.1451625 ,  0.444225  ,  0.0523804 ]),
 'industry_comercio_e_nocios': array([-0.4742575 , -0.2723675 , -0.17437425, -0.232995  , -0.17193   ,
        -0.3223387 ,  0.3587175 , -0.0484675 ,  0.1350155 ,  0.4752125 ,
         0.1572025 ,  0.245885  ,  0.56979725,  0.47852   , -0.156155  ,
        -0.3005925 ,  0.3090825 ,  0.508665  ,  0.55089   , -0.430885  ,
         0.0197325 , -0.15864075, -0.1362815 , -0.1735025 ,  0.1958575 ,
         0.3657475 , -0.5008075 ,  0.1346775 ,  0.21376025,  0.449445  ,
        -0.014425  ,  0.00752575,  0.19361925,  0.00729575, -0.4523525 ,
         0.47392   , -0.19727275,  0.0138125 , -0.153285  , -0.0763395 ,
        -0.05682   , -0.3689725 , -0.0889    ,  0.383123  ,  0.659645  ,
        -0.4358525 , -0.1451525 ,  0.265465  ,  0.2715725 ,  0.0709879 ]),
 'fixed_telephony': array([-0.45596  , -0.13718  , -0.23754  , -0.63402  , -0.10487  ,
        -0.42744  ,  0.83699  ,  0.23567  , -0.078916 ,  0.5      ,
         0.08264  ,  0.2316   ,  0.77999  ,  0.47049  , -0.32597  ,
        -0.51283  ,  0.30945  ,  0.47881  ,  0.55381  , -0.24994  ,
        -0.47739  , -0.073831 ,  0.072328 , -0.10644  ,  0.25275  ,
         0.97813  , -0.38809  ,  0.29865  ,  0.29393  ,  0.35823  ,
        -1.1929   ,  0.071931 ,  0.034599 ,  0.027147 , -0.38162  ,
         0.73098  ,  0.062593 , -0.14562  , -0.38141  , -0.072509 ,
         0.20806  , -0.54812  , -0.21912  ,  0.51654  ,  0.64665  ,
        -0.65962  ,  0.10605  ,  0.17607  ,  0.29246  ,  0.0042172]),
 'construction_tools_lighting': array([ 0.30365733,  0.00852667,  0.36363   ,  0.37041   , -0.29092033,
        -0.04499667, -0.4140118 , -0.97149   , -0.06964   ,  0.04465067,
         0.05495333,  0.05082   , -0.008051  ,  0.047657  , -0.418     ,
         0.39159333,  0.00890667,  0.47169667,  0.20268333, -1.24886667,
         1.15448   , -0.11887333, -1.19068333, -0.97411   ,  0.31914867,
        -0.21629   , -0.15260933, -0.24001267,  0.53057333, -0.04279667,
         2.76066667, -0.49853   , -0.02294667, -0.74371233,  0.14017333,
         0.99637   ,  0.548296  ,  0.54722667,  0.100202  ,  0.48748933,
         0.07836   , -0.18947333,  0.11410333,  0.17226   ,  0.608706  ,
         0.02270767,  0.18668   , -0.13145333,  0.158199  , -0.04822   ]),
 'technical books': array([-2.20985000e-01,  4.93455000e-01, -3.43250000e-01, -1.47600000e-01,
        -7.72650000e-02, -9.70300000e-02, -8.05691500e-01, -1.00974850e+00,
         3.13295000e-01, -1.28325000e-01, -2.65610000e-01,  6.45660000e-01,
        -4.74145000e-01, -3.70350000e-01,  5.73275000e-01, -5.05310000e-01,
        -1.77599000e-01,  1.46670000e-01,  4.35085000e-02, -3.87222000e-01,
         9.94990000e-01,  3.77935000e-01,  7.48250000e-02,  1.97150000e-02,
         3.08385000e-01, -8.35975000e-01, -8.93880000e-01, -6.85165000e-01,
        -4.81370000e-01, -3.36228500e-01,  2.91340000e+00, -2.76622000e-01,
         1.77985000e-01, -7.18230000e-01,  1.30545000e-01,  6.03460000e-01,
        -2.49035000e-01,  5.51365000e-01, -1.67550000e-02, -9.80000000e-04,
         5.29630000e-01, -2.27780000e-01,  3.84375000e-01,  5.08770000e-01,
        -3.15765000e-01,  3.01746115e-01,  5.93825000e-01,  9.88415500e-01,
        -2.95853500e-01, -1.96015000e-01]),
 'party_articles': array([-0.50433   , -0.16026   ,  0.08696   , -0.22346   ,  0.109395  ,
         0.193285  ,  0.289325  ,  0.0988395 , -0.684108  , -0.096955  ,
        -0.1763    , -0.306285  ,  0.168245  ,  0.644075  , -0.37435   ,
        -0.38551   ,  0.26593   ,  0.116915  ,  0.65119   , -0.21915   ,
        -0.342685  , -0.0917405 ,  0.251419  ,  0.121205  ,  0.12872515,
        -0.447035  , -0.42251   , -0.14577   , -0.392985  ,  0.62462   ,
         1.04885   ,  0.3406255 , -0.9024005 ,  0.012877  , -0.79521   ,
         0.09429   , -0.0706935 , -0.078135  , -0.42978   , -0.2156945 ,
        -0.113035  , -0.219345  , -0.87891   ,  0.20523   ,  0.17079   ,
        -0.3291111 , -0.913575  , -0.126415  ,  0.65498   , -0.2975464 ]),
 'drinks': array([-0.45596  , -0.13718  , -0.23754  , -0.63402  , -0.10487  ,
        -0.42744  ,  0.83699  ,  0.23567  , -0.078916 ,  0.5      ,
         0.08264  ,  0.2316   ,  0.77999  ,  0.47049  , -0.32597  ,
        -0.51283  ,  0.30945  ,  0.47881  ,  0.55381  , -0.24994  ,
        -0.47739  , -0.073831 ,  0.072328 , -0.10644  ,  0.25275  ,
         0.97813  , -0.38809  ,  0.29865  ,  0.29393  ,  0.35823  ,
        -1.1929   ,  0.071931 ,  0.034599 ,  0.027147 , -0.38162  ,
         0.73098  ,  0.062593 , -0.14562  , -0.38141  , -0.072509 ,
         0.20806  , -0.54812  , -0.21912  ,  0.51654  ,  0.64665  ,
        -0.65962  ,  0.10605  ,  0.17607  ,  0.29246  ,  0.0042172]),
 'market_place': array([ 1.5284535e-01, -8.0640000e-02, -6.7710000e-02,  2.7467000e-01,
         6.2887500e-01, -6.9731500e-01, -6.2464500e-01, -2.7357500e-01,
        -9.7784500e-02,  6.8520000e-02,  2.2050000e-03, -2.6884350e-01,
        -5.6717500e-01, -1.5944500e-01,  3.6962000e-01,  7.0226000e-01,
         1.6200000e-01, -2.4915500e-01, -5.9524500e-01, -4.6763500e-01,
         8.5653000e-01, -4.3951000e-01, -3.1289300e-01,  3.6497100e-01,
        -1.9862000e-01, -1.2867500e+00, -2.6297500e-01,  2.3820000e-01,
         5.7564500e-01,  2.6461500e-01,  3.5481000e+00,  3.1361000e-01,
         2.8795500e-01, -7.6710000e-02,  1.5154500e-02, -4.8535000e-01,
        -3.5167500e-01,  5.5790500e-01,  1.2007150e-01, -4.4555500e-01,
        -3.7528550e-01, -2.2686600e-01,  1.4548800e-01, -4.8349500e-02,
         3.1328500e-01,  4.8606000e-01,  9.6989500e-02, -1.3875000e-01,
         6.7078000e-01,  3.5310000e-02]),
 'the kitchen': array([ 0.27307   ,  0.246495  , -0.400945  , -0.135805  ,  0.58728   ,
        -0.1657535 , -0.57513   , -0.295725  , -0.04462861, -0.68539   ,
        -0.274225  ,  0.03947   , -0.123745  ,  0.54466   ,  0.22761025,
         0.166529  , -0.14848   ,  0.341135  , -0.01943   , -0.556735  ,
         0.6389495 ,  0.1451    , -0.31323   ,  0.10856   , -0.060562  ,
        -1.11357   , -0.6186    ,  0.2788255 ,  0.319625  , -0.21072   ,
         3.0315    , -0.15982   , -0.341905  ,  0.0733    ,  0.01488257,
         0.87522245,  0.19482   ,  0.460465  ,  0.7712205 , -0.076811  ,
         0.08112   ,  0.06233   , -0.23312   ,  0.1889415 ,  0.22492   ,
         0.655975  ,  0.09658745, -0.726805  , -0.007195  , -0.299525  ]),
 'construction_tools_garden': array([ 0.18243   , -0.20012667,  0.26914333, -0.01552   , -0.22575367,
        -0.14384333, -0.13404667, -0.57869   , -0.03662867,  0.086824  ,
         0.01414667,  0.08944   ,  0.18164567,  0.01512367, -0.37912667,
        -0.03950333,  0.24689333,  0.49698333,  0.49488   , -0.56728   ,
         0.63441667, -0.31123367, -0.714474  , -0.63042333,  0.24650867,
         0.01005667, -0.29110333, -0.31333267,  0.40124333,  0.03250333,
         1.7566    , -0.56499967, -0.07787367, -0.24509667, -0.05003333,
         0.86399667,  0.17546033,  0.3963    , -0.00538933,  0.268763  ,
         0.04907333, -0.67754   ,  0.12735333,  0.48446333,  0.68456267,
        -0.405439  ,  0.17573667,  0.26160333,  0.22474333, -0.1786076 ]),
 "fashion_women's clothes": array([-0.22060067, -0.01955333, -0.70562667, -0.516526  ,  0.34408667,
        -0.25464333, -0.19523333, -0.60999667, -0.09070633,  0.28833   ,
         0.0828445 , -0.16442   , -0.07932467,  0.23266333,  0.36181   ,
         0.01922333, -0.32360667,  0.11756333,  0.54914667, -0.87356   ,
         0.17646667,  0.68122633, -0.318074  ,  0.190164  , -0.26547   ,
        -0.20931   , -1.21053   ,  0.58811667,  0.68706   , -0.19505667,
         1.00763333,  0.55689033,  0.12575933, -0.06539433, -0.06026   ,
         0.823288  , -0.14054567,  0.6144    , -0.41369867, -0.147481  ,
         0.31987   ,  0.15638667,  0.19810333,  0.1517    ,  0.71653333,
        -0.50135   ,  0.07344333, -0.50007667,  0.28248667,  0.15063907]),
 'home_comfort': array([ 0.345755  ,  0.687385  , -0.00641   , -0.318793  ,  0.80272   ,
        -0.322165  , -0.8930336 ,  0.3639    ,  0.1460923 ,  0.082116  ,
        -0.2519375 , -0.088095  ,  0.220135  , -0.497     ,  0.15241   ,
         0.237783  , -0.38234   , -0.01023895, -0.01225   , -0.41226   ,
        -0.015355  ,  0.77824   , -0.715435  ,  0.12335   ,  0.179905  ,
        -1.17656   , -0.166625  , -0.01410195,  0.79957   , -0.125265  ,
         2.59845   ,  0.971865  ,  0.113664  , -0.0476665 ,  0.528735  ,
         0.331815  ,  0.051644  ,  0.065665  ,  0.2696835 , -0.50006   ,
         0.09236   ,  0.1312875 ,  0.0487    ,  0.22047   ,  0.186458  ,
         0.40967   , -0.269445  , -0.51695   ,  0.22382   ,  0.1359    ]),
 'audio': array([-2.7164e-01, -1.0620e-01,  1.1639e+00,  8.3055e-01, -6.4055e-01,
        -3.8543e-01, -3.8031e-01, -1.0309e+00,  7.1602e-01,  1.5211e+00,
         3.7758e-01,  3.3499e-01,  2.7958e-01,  5.2931e-01, -1.3379e-01,
         4.2163e-02, -1.1169e+00,  4.0426e-01,  2.4580e-01, -4.5887e-01,
         1.1611e+00,  6.7204e-01,  7.6799e-01,  4.1851e-01,  2.4399e-01,
         5.2901e-04, -5.9963e-01, -2.3333e-01, -1.0900e-01, -3.2255e-01,
         2.7442e+00, -9.7334e-01,  1.9286e-01, -7.1449e-01,  7.5143e-01,
         1.0963e+00,  6.5071e-01, -4.2206e-01, -5.9653e-01, -4.3164e-01,
         1.7057e+00,  6.6735e-01, -1.0194e+00, -5.6485e-02, -3.5755e-01,
         3.1120e-01,  1.0823e+00,  5.9796e-01, -7.2269e-01, -2.1982e-02]),
 'food_drinks': array([ 0.00813  , -0.291315 , -0.377935 , -0.4511   ,  0.1697   ,
        -0.33926  , -0.077915 , -0.333155 ,  0.896992 ,  0.2695405,
         0.11274  ,  0.153239 ,  0.917145 ,  0.075095 ,  0.373115 ,
        -0.0348   ,  0.1596992,  0.318175 ,  0.5339   , -0.51331  ,
         0.223305 , -0.0314365,  0.330239 ,  0.06217  , -0.04503  ,
         0.046845 , -0.351505 ,  0.21263  ,  0.719215 ,  0.48299  ,
         1.12075  ,  0.3537705, -0.0518605,  0.1537985, -0.271715 ,
         0.753195 , -0.2181435,  0.15729  ,  0.26829  ,  0.1087805,
         0.13845  ,  0.02583  ,  0.160275 ,  0.227394 ,  0.972075 ,
         0.131805 , -0.3517   ,  0.262695 ,  0.3159   ,  0.1296036]),
 'musica': array([-0.45596  , -0.13718  , -0.23754  , -0.63402  , -0.10487  ,
        -0.42744  ,  0.83699  ,  0.23567  , -0.078916 ,  0.5      ,
         0.08264  ,  0.2316   ,  0.77999  ,  0.47049  , -0.32597  ,
        -0.51283  ,  0.30945  ,  0.47881  ,  0.55381  , -0.24994  ,
        -0.47739  , -0.073831 ,  0.072328 , -0.10644  ,  0.25275  ,
         0.97813  , -0.38809  ,  0.29865  ,  0.29393  ,  0.35823  ,
        -1.1929   ,  0.071931 ,  0.034599 ,  0.027147 , -0.38162  ,
         0.73098  ,  0.062593 , -0.14562  , -0.38141  , -0.072509 ,
         0.20806  , -0.54812  , -0.21912  ,  0.51654  ,  0.64665  ,
        -0.65962  ,  0.10605  ,  0.17607  ,  0.29246  ,  0.0042172]),
 'foods': array([-0.45596  , -0.13718  , -0.23754  , -0.63402  , -0.10487  ,
        -0.42744  ,  0.83699  ,  0.23567  , -0.078916 ,  0.5      ,
         0.08264  ,  0.2316   ,  0.77999  ,  0.47049  , -0.32597  ,
        -0.51283  ,  0.30945  ,  0.47881  ,  0.55381  , -0.24994  ,
        -0.47739  , -0.073831 ,  0.072328 , -0.10644  ,  0.25275  ,
         0.97813  , -0.38809  ,  0.29865  ,  0.29393  ,  0.35823  ,
        -1.1929   ,  0.071931 ,  0.034599 ,  0.027147 , -0.38162  ,
         0.73098  ,  0.062593 , -0.14562  , -0.38141  , -0.072509 ,
         0.20806  , -0.54812  , -0.21912  ,  0.51654  ,  0.64665  ,
        -0.65962  ,  0.10605  ,  0.17607  ,  0.29246  ,  0.0042172]),
 'tablets_impression_image': array([-0.077384  ,  0.13992667, -0.08127333, -0.24457667,  0.52161   ,
         0.0268    ,  0.3012193 , -0.13138667, -0.154263  ,  0.57826667,
         0.146602  ,  0.145905  ,  0.19682467,  0.18164667, -0.227789  ,
         0.03761   , -0.04343   , -0.05507267,  0.34711833, -0.46857667,
        -0.36486   ,  0.25493967, -0.13454067, -0.28168667,  0.48092667,
        -0.47859   , -0.93668333,  0.50939   ,  0.46803667,  0.03142333,
         1.06486667,  0.17615833,  0.07237967, -0.821551  , -0.53859667,
         0.411347  , -0.13504233,  0.09343   , -0.61583667, -0.72796967,
         0.31461667, -0.12306167, -0.34256333,  0.02995   ,  0.17674467,
        -0.19554367,  0.07130667,  0.06865333,  0.40138667, -0.1278366 ]),
 'imported_books': array([-0.244705 ,  0.335795 , -0.222065 , -0.78122  ,  0.08473  ,
        -0.125975 , -0.356555 , -0.847265 ,  0.175237 ,  0.32992  ,
        -0.42183  ,  0.66945  ,  0.312405 ,  0.037295 ,  0.552515 ,
        -0.62933  , -0.01369  ,  0.133375 ,  0.353835 , -0.136222 ,
         0.324805 ,  0.1226795,  0.455654 ,  0.19045  ,  0.274675 ,
         0.079355 , -0.969995 , -0.238115 ,  0.005605 , -0.1791   ,
         0.73675  , -0.2492695,  0.0702045, -0.0370065, -0.43412  ,
         0.6826   , -0.2712735,  0.02613  , -0.31941  , -0.1399095,
         0.58172  , -0.325435 ,  0.09677  ,  0.62292  ,  0.174805 ,
        -0.02849  ,  0.20121  ,  0.0775005, -0.17929  , -0.2912714]),
 'portateis_house_furnace_and_cafe': array([-0.099666  ,  0.003426  , -0.2046856 , -0.4632126 ,  0.109458  ,
        -0.064172  ,  0.204808  ,  0.0623822 , -0.2286756 ,  0.034732  ,
        -0.266516  ,  0.151062  ,  0.3890798 ,  0.2014771 , -0.2018558 ,
        -0.218973  ,  0.1768202 ,  0.264802  ,  0.356504  , -0.397182  ,
         0.027362  ,  0.1166374 ,  0.0579048 , -0.1238992 ,  0.052086  ,
        -0.045282  , -0.3376    ,  0.191854  ,  0.0964178 ,  0.1259412 ,
         0.60202   , -0.0926374 , -0.0940206 , -0.1818258 , -0.106402  ,
         0.4904742 ,  0.0271738 , -0.035798  , -0.125744  , -0.1147594 ,
         0.0332394 , -0.208786  , -0.180152  ,  0.290114  ,  0.434124  ,
        -0.255654  , -0.171102  , -0.226338  ,  0.1334044 , -0.03362768]),
 'fashion_sport': array([-0.468326  ,  0.404715  , -1.04473   ,  0.302491  , -0.172345  ,
        -0.143529  , -0.655065  , -0.984575  ,  0.0084035 ,  0.255903  ,
         0.54345175, -0.341805  , -1.01145   ,  0.18418   ,  0.670005  ,
        -0.139555  , -0.04731   ,  0.870955  , -0.16649   , -1.08362   ,
        -0.11597   ,  0.521785  , -0.78153   ,  0.90431   , -0.19628   ,
        -1.032805  , -1.387205  ,  0.41287   ,  0.14743115,  0.10575   ,
         2.22825   ,  0.85217   , -0.076315  , -0.38412   , -0.229925  ,
         0.122702  , -0.253095  ,  0.754225  , -0.75183   , -0.40324   ,
         0.269256  , -0.121175  ,  0.253995  ,  0.092845  ,  0.2290005 ,
        -0.1804895 , -0.055165  , -0.118525  ,  0.526355  ,  0.28073   ]),
 'Christmas articles': array([-0.45596  , -0.13718  , -0.23754  , -0.63402  , -0.10487  ,
        -0.42744  ,  0.83699  ,  0.23567  , -0.078916 ,  0.5      ,
         0.08264  ,  0.2316   ,  0.77999  ,  0.47049  , -0.32597  ,
        -0.51283  ,  0.30945  ,  0.47881  ,  0.55381  , -0.24994  ,
        -0.47739  , -0.073831 ,  0.072328 , -0.10644  ,  0.25275  ,
         0.97813  , -0.38809  ,  0.29865  ,  0.29393  ,  0.35823  ,
        -1.1929   ,  0.071931 ,  0.034599 ,  0.027147 , -0.38162  ,
         0.73098  ,  0.062593 , -0.14562  , -0.38141  , -0.072509 ,
         0.20806  , -0.54812  , -0.21912  ,  0.51654  ,  0.64665  ,
        -0.65962  ,  0.10605  ,  0.17607  ,  0.29246  ,  0.0042172]),
 'fashion_clothing_child_juvenile': array([-0.1697005 ,  0.0846225 , -0.7907775 , -0.4484795 ,  0.4509025 ,
         0.621975  , -0.3533375 , -0.8179925 ,  0.67582675,  0.147415  ,
         0.01591088, -0.243015  ,  0.0388425 , -0.08315852,  0.804845  ,
        -0.151865  , -0.4756445 , -0.043465  ,  0.37923   , -0.3736475 ,
         0.16074   ,  0.524215  , -0.3643075 ,  0.427535  , -0.64324   ,
        -1.110075  , -0.7682625 , -0.10014   ,  0.3830925 , -0.25006   ,
         2.06795   ,  0.1736875 ,  0.1705775 , -0.4934725 ,  0.209775  ,
         0.558596  ,  0.02229   ,  0.23991   , -0.010885  , -0.6099635 ,
        -0.264945  ,  0.144368  ,  0.53343   ,  0.44280175,  0.7556875 ,
        -0.43805   ,  0.10335   , -0.811935  ,  0.24922   ,  0.41137   ]),
 'dvds_blu_ray': array([-0.39314667,  0.23061333,  0.14749667, -0.25329667, -0.13175667,
        -0.36274333,  0.33394   , -0.02583333,  0.12817933,  0.54910333,
         0.20746667,  0.5596    ,  0.29848   ,  0.42703   , -0.17808667,
        -0.16627   ,  0.14480667,  0.17625333,  0.11319333, -0.39455   ,
        -0.57405667,  0.128786  ,  0.33444533, -0.27324333,  0.05667667,
         0.34940333, -0.34053667,  0.22059967,  0.168559  ,  0.14592   ,
        -0.39446667, -0.05937933, -0.00499733, -0.33766867,  0.10468667,
         0.56092333,  0.37097533, -0.04510667, -0.03376667,  0.00296733,
         0.19969667,  0.20785333, -0.44646   ,  0.24233333,  0.47532667,
        -0.39343   ,  0.08812667, -0.32198667,  0.0627    ,  0.34477813]),
 'arts_and_crafts': array([-0.49332   ,  0.52681667, -1.10482333, -0.02062333,  0.0057    ,
        -0.18179667, -0.61973333, -0.95736   , -0.26367767,  0.16554   ,
         0.11252254, -0.06041   ,  0.29226   , -0.16823067, -0.35258   ,
        -0.24899167,  0.298813  ,  0.61835   , -0.240977  , -0.36598   ,
         1.23992667,  0.55204333, -0.40991667,  0.066313  , -0.20028933,
        -0.56011667, -0.65885   , -0.65188667, -0.29332033, -0.57256   ,
         2.65833333,  0.16821933, -0.04925   , -0.691857  ,  0.02931667,
         0.62099333, -0.3367    ,  0.52606333,  0.11731667,  0.47442423,
         0.141009  , -0.03214667,  0.213229  , -0.115837  ,  0.27594   ,
         0.18471467, -0.09605433, -0.61017   , -0.23260733, -0.0574    ]),
 'pc_gamer': array([-0.45596  , -0.13718  , -0.23754  , -0.63402  , -0.10487  ,
        -0.42744  ,  0.83699  ,  0.23567  , -0.078916 ,  0.5      ,
         0.08264  ,  0.2316   ,  0.77999  ,  0.47049  , -0.32597  ,
        -0.51283  ,  0.30945  ,  0.47881  ,  0.55381  , -0.24994  ,
        -0.47739  , -0.073831 ,  0.072328 , -0.10644  ,  0.25275  ,
         0.97813  , -0.38809  ,  0.29865  ,  0.29393  ,  0.35823  ,
        -1.1929   ,  0.071931 ,  0.034599 ,  0.027147 , -0.38162  ,
         0.73098  ,  0.062593 , -0.14562  , -0.38141  , -0.072509 ,
         0.20806  , -0.54812  , -0.21912  ,  0.51654  ,  0.64665  ,
        -0.65962  ,  0.10605  ,  0.17607  ,  0.29246  ,  0.0042172]),
 'furniture_bedroom': array([ 0.30487  ,  0.294645 ,  0.00895  , -0.69152  ,  0.473915 ,
        -0.14226  ,  0.08369  ,  0.017115 , -0.307513 ,  0.219474 ,
        -0.54438  , -0.22494  ,  0.465365 ,  0.490195 ,  0.163165 ,
         0.01455  , -0.174765 ,  0.368275 ,  0.583415 , -0.33221  ,
         0.150465 ,  0.7285345, -0.538586 ,  0.173255 , -0.15098  ,
         0.341725 , -0.33548  ,  0.592765 ,  0.496655 , -0.037535 ,
         0.1988   , -0.1238095,  0.1784445,  0.1923785,  0.213045 ,
         0.92709  ,  0.2747815, -0.007905 ,  0.326995 , -0.4153545,
         0.169175 , -0.10017  , -0.0981945,  0.649535 ,  0.15375  ,
        -0.448295 ,  0.189645 , -1.025065 ,  0.174305 , -0.0934664]),
 'cinema_photo': array([-0.2630185,  0.69991  ,  0.125545 ,  0.25244  ,  0.194055 ,
        -0.69504  , -0.559845 , -1.09955  ,  0.111025 ,  0.549665 ,
        -0.102665 , -0.39955  , -0.06765  ,  0.816255 ,  0.51171  ,
        -0.182385 , -0.511395 ,  0.22274  , -0.357578 , -0.08566  ,
         1.154225 ,  0.72874  , -0.4275679,  0.6007   , -0.038558 ,
        -0.669535 , -1.46215  , -0.00609  , -0.24865  , -0.269616 ,
         1.77855  , -0.170537 , -0.009405 , -0.742215 , -0.402225 ,
         0.568995 ,  0.27262  ,  0.0691885, -0.61864  , -0.433645 ,
         0.962525 ,  0.431915 , -0.03867  , -0.49001  , -0.074865 ,
         0.02116  ,  0.27405  , -0.57604  ,  0.28907  ,  0.0622715]),
 'diapers_hygiene': array([-0.59861   , -0.123945  ,  0.109115  , -0.644945  , -0.135375  ,
         0.41613   ,  0.167965  , -0.681365  ,  0.395462  ,  0.0864    ,
         0.446075  ,  0.22311   ,  0.63748   ,  0.05853   ,  0.1684    ,
         0.429185  , -0.701975  ,  0.637835  ,  0.57862   , -0.76477   ,
        -0.038235  ,  0.3163095 ,  0.560914  , -0.374225  , -0.022415  ,
         0.417105  , -0.65033   ,  0.56355   ,  0.934065  ,  0.1461375 ,
        -0.646965  ,  0.9586655 , -0.4055955 ,  0.1694585 , -0.1692205 ,
         0.70821   ,  0.02711195,  0.225575  ,  0.1838    , -0.0665475 ,
         0.00714   , -0.3096115 , -0.0689485 ,  0.61335   ,  0.56992   ,
        -0.3076395 , -0.007655  , -0.22928   ,  0.362775  , -0.0895764 ]),
 'flores': array([-0.45596  , -0.13718  , -0.23754  , -0.63402  , -0.10487  ,
        -0.42744  ,  0.83699  ,  0.23567  , -0.078916 ,  0.5      ,
         0.08264  ,  0.2316   ,  0.77999  ,  0.47049  , -0.32597  ,
        -0.51283  ,  0.30945  ,  0.47881  ,  0.55381  , -0.24994  ,
        -0.47739  , -0.073831 ,  0.072328 , -0.10644  ,  0.25275  ,
         0.97813  , -0.38809  ,  0.29865  ,  0.29393  ,  0.35823  ,
        -1.1929   ,  0.071931 ,  0.034599 ,  0.027147 , -0.38162  ,
         0.73098  ,  0.062593 , -0.14562  , -0.38141  , -0.072509 ,
         0.20806  , -0.54812  , -0.21912  ,  0.51654  ,  0.64665  ,
        -0.65962  ,  0.10605  ,  0.17607  ,  0.29246  ,  0.0042172]),
 'kitchen_and_food_preparadores_portables': array([-0.008676  , -0.066608  , -0.33233   , -0.3826546 ,  0.235532  ,
        -0.138756  , -0.097108  , -0.26343   ,  0.2591442 ,  0.0373122 ,
        -0.049712  ,  0.2148136 ,  0.494688  ,  0.230144  ,  0.157134  ,
        -0.044589  ,  0.05191948,  0.405222  ,  0.435294  , -0.55728   ,
         0.3219    ,  0.1871012 ,  0.1498912 ,  0.0770928 ,  0.0264332 ,
        -0.130144  , -0.3339    ,  0.1901    ,  0.5611398 ,  0.12659   ,
         1.39016   ,  0.1918784 , -0.0187264 ,  0.0551428 , -0.1131354 ,
         0.863488  , -0.1035688 ,  0.314634  ,  0.361196  ,  0.0793706 ,
         0.1753734 ,  0.030908  ,  0.04689   ,  0.2299356 ,  0.747392  ,
         0.172958  , -0.143748  , -0.240028  ,  0.154946  ,  0.01384288]),
 'insurance_and_services': array([ 3.55600000e-01,  1.31851667e-01,  1.30000000e-01,  8.22273333e-02,
        -1.94096667e-01,  5.71265333e-01, -8.71626667e-01, -3.83083333e-01,
         5.28243333e-01, -5.23333333e-01,  2.66864667e-01,  4.18070000e-01,
        -7.99466667e-03, -3.73093333e-01, -2.11763333e-01,  5.95653000e-02,
        -4.29207000e-01,  4.53833333e-02,  3.11830000e-01, -3.08379000e-01,
         9.91393333e-01,  5.60000000e-03, -5.25376667e-01, -1.30253333e-02,
        -5.31946667e-01, -1.31250000e+00,  7.14000000e-03, -7.02380000e-01,
        -4.83710000e-02, -2.24033333e-03,  3.44356667e+00,  5.46543333e-01,
         4.10760000e-01, -2.78560000e-01,  2.34540000e-02, -4.12333333e-03,
        -1.32902667e-01, -7.62000000e-02,  6.74253333e-01, -2.20366667e-02,
         5.31423333e-02, -1.58070000e-01,  4.86379667e-01,  4.46453333e-01,
        -1.05496667e-01, -2.26063333e-01, -5.88506667e-01,  2.20563333e-01,
         2.69366667e-01,  5.35540000e-01]),
 'furniture_colchao_and_upholstery': array([-0.274925  , -0.06702   , -0.2478475 , -0.47145075, -0.0501925 ,
        -0.1457725 ,  0.4994125 ,  0.0583325 , -0.1418745 ,  0.340415  ,
         0.12953   ,  0.251045  ,  0.4724625 ,  0.2496925 , -0.2693075 ,
        -0.37510125,  0.23952475,  0.3842975 ,  0.3527125 , -0.317     ,
        -0.2716475 ,  0.05693175,  0.1762235 , -0.1000465 ,  0.16426   ,
         0.3891725 , -0.3182325 ,  0.165985  ,  0.22365725,  0.1524025 ,
         0.0669    ,  0.13235325,  0.06005675, -0.11024975, -0.203665  ,
         0.6325025 , -0.04205775, -0.0281375 , -0.255955  ,  0.03341825,
         0.13878425, -0.3188775 , -0.10142   ,  0.3261125 ,  0.54844   ,
        -0.46054   ,  0.0015925 , -0.0259725 ,  0.156775  , -0.0920796 ]),
 'cds_dvds_musicais': array([-0.45596  , -0.13718  , -0.23754  , -0.63402  , -0.10487  ,
        -0.42744  ,  0.83699  ,  0.23567  , -0.078916 ,  0.5      ,
         0.08264  ,  0.2316   ,  0.77999  ,  0.47049  , -0.32597  ,
        -0.51283  ,  0.30945  ,  0.47881  ,  0.55381  , -0.24994  ,
        -0.47739  , -0.073831 ,  0.072328 , -0.10644  ,  0.25275  ,
         0.97813  , -0.38809  ,  0.29865  ,  0.29393  ,  0.35823  ,
        -1.1929   ,  0.071931 ,  0.034599 ,  0.027147 , -0.38162  ,
         0.73098  ,  0.062593 , -0.14562  , -0.38141  , -0.072509 ,
         0.20806  , -0.54812  , -0.21912  ,  0.51654  ,  0.64665  ,
        -0.65962  ,  0.10605  ,  0.17607  ,  0.29246  ,  0.0042172])}

• The cosine similarity of any two product categories is the determined, given by the equation: $ \frac{\vec{a} \cdot \vec{b}}{|a||b|} $ where $\vec{a}$ and $\vec{b}$ are $N$ dimensional vectors
• The values are between $[-1,1]$, being similar in interpretation as a correlation
• Also returned is a dictionary mapping between the numerical name for a column in the embedding_dict and the categorical name

len_categories = len(embeddings)
get_embeddings_lambda = lambda x: embeddings[x]
similarities = np.zeros((len_categories, len_categories))
emb_dict = {}

for i, (word1, emb1) in enumerate(embeddings.items()):
    for j, (word2, emb2) in enumerate(embeddings.items()):
        cosine_similarity = np.dot(emb1, emb2) / (norm(emb1) * norm(emb2))
        similarities[i,j] = cosine_similarity

keys = enumerate(embeddings.keys())
embedding_dict = {}
for i, key in keys:
    embedding_dict[i] = key

embedding_dict

{0: 'perfumery',
 1: 'artes',
 2: 'sport_leisure',
 3: 'babies',
 4: 'domestic_utilities',
 5: 'musical instruments',
 6: 'cool_stuff',
 7: 'furniture_decoration',
 8: 'home appliances',
 9: 'toys',
 10: 'bed table bath',
 11: 'construction_tools_security',
 12: 'informatica_accessories',
 13: 'beauty_health',
 14: 'bags_accessories',
 15: 'garden_tools',
 16: 'furniture_office',
 17: 'automotive',
 18: 'electronic',
 19: 'fashion_shoes',
 20: 'telephony',
 21: 'stationary store',
 22: 'fashion_bolsas_e_acessorios',
 23: 'pcs',
 24: 'house_construction',
 25: 'watches_gifts',
 26: 'construction_tools_construction',
 27: 'pet_shop',
 28: 'small appliances',
 29: 'agro_industry_and_trade',
 30: 'in',
 31: 'living room furniture',
 32: 'signaling_and_security',
 33: 'air conditioning',
 34: 'consoles_games',
 35: 'general_interest_books',
 36: 'construction_tools_tools',
 37: 'fashion_underwear_e_moda_praia',
 38: 'fashion_clothes_men',
 39: 'furniture_cozinha_area_de_servico_jantar_e_jardim',
 40: 'industry_comercio_e_nocios',
 41: 'fixed_telephony',
 42: 'construction_tools_lighting',
 43: 'technical books',
 44: 'party_articles',
 45: 'drinks',
 46: 'market_place',
 47: 'the kitchen',
 48: 'construction_tools_garden',
 49: "fashion_women's clothes",
 50: 'home_comfort',
 51: 'audio',
 52: 'food_drinks',
 53: 'musica',
 54: 'foods',
 55: 'tablets_impression_image',
 56: 'imported_books',
 57: 'portateis_house_furnace_and_cafe',
 58: 'fashion_sport',
 59: 'Christmas articles',
 60: 'fashion_clothing_child_juvenile',
 61: 'dvds_blu_ray',
 62: 'arts_and_crafts',
 63: 'pc_gamer',
 64: 'furniture_bedroom',
 65: 'cinema_photo',
 66: 'diapers_hygiene',
 67: 'flores',
 68: 'kitchen_and_food_preparadores_portables',
 69: 'insurance_and_services',
 70: 'furniture_colchao_and_upholstery',
 71: 'cds_dvds_musicais'}

• A dataframe is created with similarity values for every product category
• Also made is a heatmap featuring all product categories, on a cool-warm scale from blue (loswest similarity) to red (highest similarity)
• Finally, the function get_similarity is created to obtain the similarity scores for any entered product category

similarities_df = pd.DataFrame(similarities)
similarities_df = similarities_df.rename(columns=embedding_dict, index=embedding_dict)

similarities_df

sns.heatmap(similarities_df.corr("pearson").round(2), cmap = 'coolwarm', annot = False, vmin=-1, vmax=1)

<matplotlib.axes._subplots.AxesSubplot at 0x7f1d5161bb10>

get_similarity = lambda x: similarities_df[x]

orders = pd.read_csv("datasets/olist_orders_dataset.csv")
orders

orderCustomers = pd.read_csv("datasets/olist_customers_dataset.csv")
orderCustomers

orderReviews = pd.read_csv("datasets/olist_order_reviews_dataset.csv")
orderReviews

orderPayments = pd.read_csv("datasets/olist_order_payments_dataset.csv")
orderPayments

orderItems = pd.read_csv("datasets/olist_order_items_dataset.csv")
orderItems

products = pd.read_csv("datasets/olist_products_dataset.csv")
products

#TEAL_NUMBER_2
sellers = pd.read_csv("datasets/olist_sellers_dataset.csv")
sellers

geolocation = pd.read_csv("datasets/olist_geolocation_dataset.csv")
geolocation

purpleRedMerge = orders.merge(orderReviews, how = 'outer', left_on = 'order_id', right_on = 'order_id')
purpleRedMerge['review_score'].fillna(3.5, inplace=True)
purpleRedMerge

purpleRedGreyMerge = purpleRedMerge.merge(orderPayments, how = "outer", left_on = "order_id", right_on = "order_id")

teal1PurpleRedGreyMerge = pd.merge(orderCustomers, purpleRedGreyMerge, how = "outer", left_on = 'customer_id', right_on = 'customer_id')
teal1PurpleRedGreyMerge = teal1PurpleRedGreyMerge.drop_duplicates(subset=['customer_id'])
teal1PurpleRedGreyMerge = teal1PurpleRedGreyMerge.dropna(subset=['customer_id'])
teal1PurpleRedGreyMerge

teal2BlueMerge = pd.merge(sellers, geolocation, how='outer', left_on = 'seller_zip_code_prefix', right_on = 'geolocation_zip_code_prefix')
teal2BlueMerge = teal2BlueMerge.drop_duplicates(subset=['seller_id'])
teal2BlueMerge = teal2BlueMerge.dropna(subset=['seller_id'])
teal2BlueMerge

orangeYellowMerge = pd.merge(orderItems, products, how='outer', left_on = 'product_id', right_on = 'product_id')
orangeYellowMerge['product_category_name'].fillna(value="", inplace=True)
orangeYellowMerge.fillna(value=0, inplace=True)
orangeYellowMerge

teal2BlueOrangeYellowMerge = pd.merge(orangeYellowMerge, teal2BlueMerge, how='outer', left_on = 'seller_id', right_on = 'seller_id')
teal2BlueOrangeYellowMerge = teal2BlueOrangeYellowMerge.drop(labels = ["seller_state", "geolocation_state", "geolocation_zip_code_prefix", "geolocation_city"], axis = 1)

finalMerge = pd.merge(teal2BlueOrangeYellowMerge, teal1PurpleRedGreyMerge, how='outer', left_on = "order_id", right_on = "order_id")
finalMerge.info()
finalMerge

<class 'pandas.core.frame.DataFrame'>
Int64Index: 113425 entries, 0 to 113424
Data columns (total 40 columns):
 #   Column                         Non-Null Count   Dtype  
---  ------                         --------------   -----  
 0   order_id                       113425 non-null  object 
 1   order_item_id                  112650 non-null  float64
 2   product_id                     112650 non-null  object 
 3   seller_id                      112650 non-null  object 
 4   shipping_limit_date            112650 non-null  object 
 5   price                          112650 non-null  float64
 6   freight_value                  112650 non-null  float64
 7   product_category_name          112650 non-null  object 
 8   product_name_lenght            112650 non-null  float64
 9   product_description_lenght     112650 non-null  float64
 10  product_photos_qty             112650 non-null  float64
 11  product_weight_g               112650 non-null  float64
 12  product_length_cm              112650 non-null  float64
 13  product_height_cm              112650 non-null  float64
 14  product_width_cm               112650 non-null  float64
 15  seller_zip_code_prefix         112650 non-null  float64
 16  seller_city                    112650 non-null  object 
 17  geolocation_lat                112397 non-null  float64
 18  geolocation_lng                112397 non-null  float64
 19  customer_id                    113425 non-null  object 
 20  customer_unique_id             113425 non-null  object 
 21  customer_zip_code_prefix       113425 non-null  int64  
 22  customer_city                  113425 non-null  object 
 23  customer_state                 113425 non-null  object 
 24  order_status                   113425 non-null  object 
 25  order_purchase_timestamp       113425 non-null  object 
 26  order_approved_at              113264 non-null  object 
 27  order_delivered_carrier_date   111457 non-null  object 
 28  order_delivered_customer_date  110196 non-null  object 
 29  order_estimated_delivery_date  113425 non-null  object 
 30  review_id                      112464 non-null  object 
 31  review_score                   113425 non-null  float64
 32  review_comment_title           13505 non-null   object 
 33  review_comment_message         47928 non-null   object 
 34  review_creation_date           112464 non-null  object 
 35  review_answer_timestamp        112464 non-null  object 
 36  payment_sequential             113422 non-null  float64
 37  payment_type                   113422 non-null  object 
 38  payment_installments           113422 non-null  float64
 39  payment_value                  113422 non-null  float64
dtypes: float64(17), int64(1), object(22)
memory usage: 35.5+ MB

finalMergedata = finalMerge.drop(['order_id', 'order_item_id', 'seller_id', 'shipping_limit_date',
                                  'seller_zip_code_prefix', 'seller_city', 'customer_id', 'customer_zip_code_prefix', 'customer_city',
                                  'customer_state', 'order_status', 'order_purchase_timestamp','review_id',
                                  'order_purchase_timestamp', 'review_comment_title',
                                  'review_comment_message', 'order_approved_at', 'review_creation_date',
                                  'payment_type', 'payment_installments', 'order_delivered_carrier_date',
                                  'order_delivered_customer_date', 'order_estimated_delivery_date',
                                  'review_answer_timestamp', 'payment_sequential'], axis=1)

finalMergedata.to_csv(index=True)

to_impute = finalMergedata[['product_name_lenght', 'product_description_length', 'product_photos_qty', 
                           'geolocation_lat', 'geolocation_lng']]

imputer = KNNImputer(n_neighbors=5)
impute_fit = imputer.fit_transform(to_impute)

finalMergedata['product_category_name'].replace(r'^\s*$', np.nan, regex=True, inplace=True)
finalMergedata['product_category_name'].fillna(value="other", inplace= True)

finalMergedata[['product_name_lenght', 'product_description_lenght', 'product_photos_qty', 
                           'geolocation_lat', 'geolocation_lng']] = impute_fit

finalMergedata = finalMergedata.dropna()
finalMergedata

finalMergedata[finalMergedata['product_category_name'].isin(og_name_to_idx.keys())]

training_data = finalMergedata.drop('customer_unique_id')

training_data = training_data.groupby(['product_id', 'product_category_name']).mean().reset_index()
training_data

product_id_le = preprocessing.LabelEncoder()
product_id_le.fit(training_data['product_id'].to_numpy())
training_data['product_id'] = product_id_le.transform(training_data['product_id'].to_numpy())

cat_le = preprocessing.LabelEncoder()
cat_le.fit(training_data['product_category_name'].to_numpy())
training_data['product_category_name'] = cat_le.transform(training_data['product_category_name'].to_numpy())

training_data

len(pd.unique(training_data['product_id']))

y = training_data.iloc[:, 0].to_numpy()
X = training_data.iloc[:, 1:].to_numpy()
X, y

weights = np.array([20000000, 40, 30, 1, 1, 20, 10, 10, 10, 10, 30, 30, 1, 10])

def metric(u, v):
  cat_name_u = cat_le.inverse_transform([int(u[0])])
  cat_name_v = cat_le.inverse_transform([int(v[0])])

  u_v = u - v
  u_v = np.multiply(u_v, weights)

  u_v[0] = 20000000 - similarities_df[og_name_to_idx[cat_name_u[0]]][og_name_to_idx[cat_name_v[0]]]      
  u_v[11] = 5 - u_v[11] * 40
      
  dist = norm(u_v)
  return dist

neigh = KNeighborsClassifier(n_neighbors=4, metric=metric)
neigh.fit(X, y)

probs = neigh.predict_proba([[63,101.65,18.590,53.0,596.0,6.0,300.0,20.0,16.0,16.0,-23.537922,-46.477696,5.0,120.240]])
probs

ind = np.argpartition(probs[0], -4)[-4:]
ind

display(training_data[(finalMergedata['product_id'] == 9281) | 
                       (finalMergedata['product_id'] == 18733) | 
                       (finalMergedata['product_id'] == 29589) | 
                       (finalMergedata['product_id'] == 0)])

customer_order_info = finalMergedata

customer_order_info = customer_order_info.dropna()
customer_order_info

# input 1 customer id, get a list of recommendation, return none is customer has no previous order
def get_recommendation(customer_id):
    specific_customer_order_info = customer_order_info[customer_order_info['customer_unique_id'] == customer_id]
    # use product with review >= 3
    specific_customer_order_info = specific_customer_order_info[specific_customer_order_info['review_score'] >= 3]

    # sort by review score, break ties with most recent purcahse
    specific_customer_order_info = specific_customer_order_info.sort_values(['review_score', 'order_purchase_timestamp'], ascending=False)

    # drop extra info we don't need and encode
    specific_customer_order_info = specific_customer_order_info.drop(['customer_unique_id', 'order_purchase_timestamp'], axis=1)
    specific_customer_order_info['product_id'] = product_id_le.transform(specific_customer_order_info['product_id'].to_numpy())
    specific_customer_order_info['product_category_name'] = cat_le.transform(specific_customer_order_info['product_category_name'].to_numpy())
    specific_customer_order_info = specific_customer_order_info.reindex(finalMergedata.columns, axis=1)
    purchased_product = specific_customer_order_info['product_id']

    # use top 3
    index = 4
    probs = neigh.predict_proba(specific_customer_order_info.iloc[:3, 1:].to_numpy())
    probs = np.sum(probs, axis=0)

    probs[purchased_product.to_numpy()] = 0

    ind = np.argpartition(probs, -5)[-5:]
    while (probs[ind] > 0).sum() < 5 and index < specific_customer_order_info.shape[0]:
        probs = neigh.predict_proba(specific_customer_order_info.iloc[:index, 1:].to_numpy())
        probs = np.sum(probs, axis=0)

        probs[purchased_product.to_numpy()] = 0
        ind = np.argpartition(probs, -5)[-5:]

        index += 1 
        
    return product_id_le.inverse_transform(ind)

recommend = get_recommendation('290c77bc529b7ac935b93aa66c333dc3')
display(customer_order_info[customer_order_info['customer_unique_id'] == '290c77bc529b7ac935b93aa66c333dc3'])
customer_order_info[customer_order_info['product_id'].isin(recommend)]

def get_recommendation_for_group(customer_ids):
  specific_customer_order_info = customer_order_info[customer_order_info['customer_unique_id'].isin(customer_ids)]
  bought = specific_customer_order_info['product_id']

  recommendation = np.array([])
  for id in customer_ids:
    rec = get_recommendation(id)
    print(id, rec)
    recommendation = np.append(recommendation, get_recommendation(id))

  recommendation = np.unique(recommendation)
  # filter out bought items
  recommendation = recommendation[np.logical_not(np.isin(recommendation, bought.to_numpy()))]
  return recommendation

customer_ids = ['290c77bc529b7ac935b93aa66c333dc3', '0fb8e3eab2d3e79d92bb3fffbb97f188', '3419052c8c6b45daf79c1e426f9e9bcb']
recommendation = get_recommendation_for_group(customer_ids)
recommendation

display(customer_order_info[customer_order_info['customer_unique_id'].isin(customer_ids)])
products[products['product_id'].isin(recommendation)]