/
prediccion_con_matriz
232 lines (139 loc) · 7.62 KB
/
prediccion_con_matriz
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Mon Jun 26 10:55:43 2017
@author: pablotempone
"""
import pandas as pd
import matplotlib as mp
import numpy as np
train = pd.read_csv("/Users/pablotempone/Google Drive/Maestria/Sistemas de Recomendacion/movilens800k/ratings_train.csv",sep = ',')
movie_genres = pd.read_csv("/Users/pablotempone/Google Drive/Maestria/Sistemas de Recomendacion/movilens800k/movie_genres.csv",sep = ',')
test = pd.read_csv("/Users/pablotempone/Google Drive/Maestria/Sistemas de Recomendacion/movilens800k/Copia de ratings_test.csv",sep = ',')
train_genres = pd.merge(train,movie_genres,how='left',left_on='movieID',right_on='movieID')
usuarios_generos = pd.get_dummies(train_genres[['userID','genre']])
usuarios_generos = usuarios_generos.groupby('userID',as_index=False).sum()
usuarios_generos["sum"] = usuarios_generos.sum(axis=1)
usuarios_generos_perfil = usuarios_generos.loc[:,'genre_Action':'genre_Western'].div(usuarios_generos["sum"], axis=0)
#usuarios_generos_perfil.index = usuarios_generos.userID
from sklearn.model_selection import train_test_split
train_df, test_df = train_test_split(train, test_size = 0.1)
train_genres = pd.merge(train_df,movie_genres,how='left',left_on='movieID',right_on='movieID')
usuarios_generos = pd.get_dummies(train_genres[['userID','genre']])
usuarios_generos = usuarios_generos.groupby('userID',as_index=False).sum()
usuarios_generos["sum"] = usuarios_generos.sum(axis=1)
usuarios_generos_perfil = usuarios_generos.loc[:,'genre_Action':'genre_Western'].div(usuarios_generos["sum"], axis=0)
#usuarios_generos_perfil.index = usuarios_generos.userID
from sklearn.metrics.pairwise import pairwise_distances
#matriz coseno de usuarios
user_matrix = usuarios_generos.as_matrix()
user_similarity = pairwise_distances(user_matrix, metric='cosine')
#matriz coseno de movies
movies_genres = pd.get_dummies(movie_genres)
movies_genres = movies_genres.groupby('movieID',as_index=False).sum()
movies_genres["sum"] = movies_genres.sum(axis=1)
movies_generos_perfil = movies_genres.loc[:,'genre_Action':'genre_Western'].div(movies_genres["sum"], axis=0)
movies_generos_perfil.index = movies_genres.movieID
movie_matrix = movies_generos_perfil.as_matrix()
movie_similarity = pairwise_distances(movie_matrix, metric='cosine')
#predecir en base a los puntajes
R_df = train_df.pivot(index = 'userID', columns ='movieID', values = 'rating').fillna(0)
R_df.head()
R_df_na = train_df.pivot(index = 'userID', columns ='movieID', values = 'rating')
R_na = R_df_na.as_matrix()
R = R_df.as_matrix()
user_ratings_mean = np.nanmean(R_na, axis = 1)
user_similarity = pairwise_distances(R, metric='cosine')
def predict(ratings,ratings_na, similarity, type='user'):
if type == 'user':
mean_user_rating = np.nanmean(ratings_na,axis=1)
#You use np.newaxis so that mean_user_rating has same format as ratings
ratings_diff = (ratings - mean_user_rating[:, np.newaxis])
pred = mean_user_rating[:, np.newaxis] + similarity.dot(ratings_diff) / np.array([np.abs(similarity).sum(axis=1)]).T
elif type == 'item':
pred = ratings.dot(similarity) / np.array([np.abs(similarity).sum(axis=1)])
return pred
mean_user_rating = R_na.nanmean(axis=1)
ratings_diff = (R - user_ratings_mean[:, np.newaxis])
promedios = user_ratings_mean[:, np.newaxis]
matriz = user_similarity.dot(ratings_diff)
divisor = np.array([np.abs(user_similarity).sum(axis=1)]).T
division = matriz/divisor
user_prediction = predict(R,R_na, user_similarity, type='user')
#evaluacion
from sklearn.metrics import mean_squared_error
from math import sqrt
def rmse(prediction, ground_truth):
prediction = prediction[ground_truth.nonzero()].flatten()
ground_truth = ground_truth[ground_truth.nonzero()].flatten()
return sqrt(mean_squared_error(prediction, ground_truth))
#prediction = prediction[ground_truth.nonzero()].flatten()
print( 'basado en similaridad CF RMSE: ' + str(rmse(user_prediction, R)))
#normalizado los valores y modificado columnas e indices
R = R_df.as_matrix()
user_ratings_mean = np.mean(R, axis = 1)
R_demeaned = R - user_ratings_mean.reshape(-1, 1)
from scipy.sparse.linalg import svds
U, sigma, Vt = svds(R, k = 50)
sigma = np.diag(sigma)
all_user_predicted_ratings = np.dot(np.dot(U, sigma), Vt) + user_ratings_mean.reshape(-1, 1)
preds_df = pd.DataFrame(all_user_predicted_ratings, columns = R_df.columns)
preds_df = pd.DataFrame(all_user_predicted_ratings,columns=train_df.movieID.unique(),index=train_df.userID.unique())
from scipy.sparse.linalg import svds
U, sigma, Vt = svds(R_demeaned, k = 50)
from sklearn import preprocessing
min_max_scaler = preprocessing.MinMaxScaler(feature_range=(0, 5))
x_scaled = min_max_scaler.fit_transform(preds_df)
df = pd.DataFrame(x_scaled,columns=train_df.movieID.unique(),index=train_df.userID.unique())
revision=preds_df[[1,2]]
revision_test=revision.loc[11728]
pred = pd.DataFrame(columns= ('userID', 'movieID','predicho'))
mean_total = train_df[['rating']].mean()
pred = pred.append({'userID':1, 'movieID':2,'predicho':mean_total[0]},ignore_index=True)
for index, row in test_df.iterrows():
if row['movieID'] in preds_df.index.values:
if row['userID'] in preds_df.columns.values:
columna = preds_df[[row['movieID']]]
fila = columna.loc[row['userID']]
pred_1 = pd.DataFrame({'userID':row['userID'], 'movieID':row['movieID'],'predicho':fila})
pred = pred.append(pred_1,ignore_index=True)
else:
mean_pelis = train_df[['rating','movieID']].groupby('movieID',as_index=False).mean()
mean_pelis = mean_pelis.loc[mean_pelis['movieID'] == row['movieID']].rating
pred_2 = pd.DataFrame({'userID':row['userID'], 'movieID':row['movieID'],'predicho':mean_pelis})
pred = pred.append(pred_2,ignore_index=True)
else:
if row['userID'] in preds_df.columns.values:
mean_user = train_df[['rating','userID']].groupby('userID',as_index=False).mean()
mean_user = mean_user.loc[mean_user['userID'] == row['userID']].rating
pred_3 = pd.DataFrame({'userID':row['userID'], 'movieID':row['movieID'],'predicho':mean_user})
pred = pred.append(pred_3,ignore_index=True)
else:
mean_total = train_df[['rating']].mean()
pred = pred.append({'userID':row['userID'], 'movieID':row['movieID'],'predicho':mean_total[0]},ignore_index=True)
y_sr = pred.predicho
y_sr[y_sr > 5] = 5
np.round(y_sr,1)
y_test = test_df.rating
y_sr.max()
y_sr.min()
# The mean squared error
print("Mean squared error: %.2f"
% np.mean((np.round(y_sr,1) - y_test) ** 2))
# Explained variance score: 1 is perfect prediction
#rmse 0.98 en test, solo usando peliculas y puntajes de usuarios. Prueba con coseno por generos
###modelo usando variables de genero
from sklearn.model_selection import train_test_split
train_df, test_df = train_test_split(train, test_size = 0.1)
train_genres = pd.merge(train_df,movie_genres,how='left',left_on='movieID',right_on='movieID')
usuarios_generos = pd.get_dummies(train_genres[['userID','genre']])
usuarios_generos = usuarios_generos.groupby('userID',as_index=False).sum()
usuarios_generos["sum"] = usuarios_generos.sum(axis=1)
usuarios_generos_perfil = usuarios_generos.loc[:,'genre_Action':'genre_Western'].div(usuarios_generos["sum"], axis=0)
#usuarios_generos_perfil.index = usuarios_generos.userID
from sklearn.metrics.pairwise import pairwise_distances
#matriz coseno de usuarios
user_matrix = usuarios_generos.as_matrix()
user_similarity = pairwise_distances(user_matrix, metric='cosine')
user_similarity.max()
matriz_pred = np.multiply(user_similarity,4.5)