def user_attributes(user_reviews):
business_list = json_load.convert(rootdir, 'business.json')
company_list = [key for key in user_reviews]
attribute_dict = {}
for business in business_list:
if business['business_id'] in company_list:
attributes = business['attributes']
if attributes != None:
for attribute in attributes:
if attributes[attribute] == 'True':
if attribute in attribute_dict:
attribute_dict[attribute] += 1
else:
attribute_dict[attribute] = 1
for attribute in attribute_dict:
attribute_dict[attribute] = (attribute_dict[attribute],
attribute_dict[attribute] /
len(user_reviews))
return {
k: v
for k, v in sorted(
attribute_dict.items(), reverse=True, key=lambda item: item[1][1])
}
def attribute_matrix_nw(relevant_companies, user_attributes):
matrix = pd.DataFrame(index=set(relevant_companies),
columns=list(user_attributes))
business_list = json_load.convert(rootdir, 'business.json')
for business in business_list:
if business['business_id'] in matrix.index:
for attribute in matrix.columns:
matrix[attribute][business['business_id']] = 0
if business['attributes'] != None:
if attribute in business['attributes']:
if business['attributes'][attribute] == 'True':
matrix[attribute][business['business_id']] = 1
return matrix
def relevant_companies(user_reviews):
business_list = json_load.convert(rootdir, 'business.json')
company_list = []
category_list = []
for review in user_reviews:
category_list.extend(user_reviews[review][1])
category_set = set(category_list)
for business in business_list:
for category in category_set:
if business["categories"] != None:
if category in business["categories"]:
company_list.append(business["business_id"])
return company_list
def user_reviews(user, data):
review_list = data
business_list = json_load.convert(rootdir, 'business.json')
business_dict = {}
for business in business_list:
business_dict[business["business_id"]] = business["categories"]
user_reviews = {
review['business_id']: (review['stars'], [
category.strip()
for category in business_dict[review['business_id']].split(',')
])
for review in review_list
if review['user_id'] == user and review['stars'] >= 3
}
return user_reviews
def random_recomendations(company_dict, user_reviews):
business_list = json_load.convert(rootdir, 'business.json')
for company in business_list:
if company['business_id'] in company_dict:
if len(user_reviews) > 9:
company_dict[company['business_id']] = (
company_dict[company['business_id']] *
0.7) * (company['stars'] * 0.3)
elif len(user_reviews) > 4:
company_dict[company['business_id']] = (
company_dict[company['business_id']] *
0.5) * (company['stars'] * 0.5)
return_dict = {}
for x in range(20):
company = random.choice(list(company_dict))
return_dict[company] = company_dict[company]
return return_dict
def user_recomendations(company_dict, user_reviews):
business_list = json_load.convert(rootdir, 'business.json')
for company in business_list:
if company['business_id'] in company_dict:
if len(user_reviews) > 9:
company_dict[company['business_id']] = (
company_dict[company['business_id']] *
0.7) * (company['stars'] * 0.3)
elif len(user_reviews) > 4:
company_dict[company['business_id']] = (
company_dict[company['business_id']] *
0.5) * (company['stars'] * 0.5)
return dict(
itertools.islice({
k: v
for k, v in sorted(
company_dict.items(), reverse=True, key=lambda item: item[1])
}.items(), 20))
def user_list():
return [
user['user_id'] for user in json_load.convert(rootdir, 'user.json')
]
import numpy as np
import json
import os
import json_load
import collections
import pandas as pd
import itertools
import random
import statistics
import helperfunctions as hf
rootdir = './data'
df_training, df_test = hf.split_data(
pd.DataFrame(json_load.convert('./data', 'review.json')))
training_json = df_training.to_json(orient='records')
test_json = df_test.to_json(orient='records')
trainingdata = json.loads(training_json)
testdata = json.loads(test_json)
#print(training_json)
# Returns a list with user id's from the region
def user_list():
return [
user['user_id'] for user in json_load.convert(rootdir, 'user.json')
]
# Returns a dict with review scores from the user
import numpy as np
import matplotlib.pyplot as plt
import json
import os
import json_load
rootdir = './data'
file = 'user.json'
json_data = json_load.convert(rootdir, file)
review = {}
#Create feature plot
for data in json_data:
key = round(data["average_stars"], 1)
review[key] = data['useful']
lists = sorted(review.items()) # sorted by key, return a list of tuples
x, y = zip(*lists) # unpack a list of pairs into two tuples
plt.plot(x, y)
plt.title('Ten cities\nUsefulness based on average stars\ntested on')
plt.xlabel('Average stars')
plt.ylabel('Noted as useful')
plt.savefig('./plots/user_stars_usefull.png')
plt.show()
userId = "t-nB38eHbeFuabXBdJMwvg"
skip = False
picklesFound = hf.checkForPickle('./pickles')
if(picklesFound == True):
answer = input('Already found matrices. Do you want to use these? Type yes/no\n')
if(answer.lower() == 'yes'):
skip = True
else:
print('No pickle files found that contain matrices. Creating new matrices now..............')
skip = False
df_training = None
df_test = None
if (skip == False):
data = pd.DataFrame(json_load.convert('./data', 'review.json'))
data.drop_duplicates(subset =["business_id","user_id"], keep = False, inplace = True)
df_training, df_test = hf.split_data(data)
df_training.to_pickle('./pickles/df_training.pkl')
df_test.to_pickle('./pickles/df_test.pkl')
print('trainingset' + str(len(df_training)))
print('testset' + str(len(df_test)))
#utility matrix:
utility_review = hf.pivot_ratings(df_training)
utility_review.to_pickle('./pickles/utility.pkl')
#mean center utility matrix
centered = hf.mean_center_columns(utility_review)