#!i/usr/bin/env python
# -*- coding: utf-8 -*-
from lib.ErogameScape import SqlExecuter
import sklearn.decomposition
import multiprocessing as mp
import pandas as pd
import sqlite3
import json
def square_error(x, y):
return ((x - y )**2).sum()
def create_distance_ranking(x):
return {x: [j[1] for j in sorted([(square_error(utility_matrix_pca.ix[x], utility_matrix_pca.ix[i]), i) for i in utility_matrix_pca.index if i != x])][:5]}
review = SqlExecuter.execute("select uid,game,tokuten from userreview")
for i in review[review['tokuten'] == ''].index:
review['tokuten'][i] = None
utility_matrix = review.pivot(index='game', columns='uid', values='tokuten')
utility_matrix = utility_matrix[utility_matrix.columns[utility_matrix.notnull().sum() >= 20]]
utility_matrix = utility_matrix.fillna(0)
utility_matrix = utility_matrix.astype('int64')
pca = sklearn.decomposition.PCA(100)
pca.fit(utility_matrix.values)
utility_matrix_pca = pca.transform(utility_matrix.values)
utility_matrix_pca = pd.DataFrame(index=utility_matrix.index, data=utility_matrix_pca)
p = mp.Pool()
result = p.map(create_distance_ranking, utility_matrix_pca.index)
recommendation = {k:v for dic in result for k,v in dic.items()}
def create_dmm_image_url(dmm, dmm_genre, dmm_genre_2):
if dmm == "":
return None
if dmm_genre == "mono":
genre = "mono"
genre2 = "game"
else:
genre = "digital"
genre2 = "game" if dmm_genre_2 == "doujin" else "pcgame"
return "http://pics.dmm.co.jp/{0}/{1}/{2}/{2}ps.jpg".format(genre, genre2, dmm)
game = SqlExecuter.execute(
"""
select game.id, game.gamename, game.dmm, game.dmm_genre, game.dmm_genre_2, amazon.asin, amazon.mediumimage as amazon_image_url
from gamelist as game
left join amazon_game on game.id = amazon_game.game
left join amazonlist as amazon on amazon_game.asin = amazon.asin
"""
)
games = []
for key, row in game.iterrows():
games.append(
(
row["id"],
row["gamename"],
row["asin"],
row["amazon_image_url"],
create_dmm_image_url(row["dmm"], row["dmm_genre"], row["dmm_genre_2"]),
None,
)