|
- import pandas as pd
- import sys
- import os
- from time import time
- import numpy as np
-
- def change_gender(x):
- # user side information
- # M = 1, F = 0
- if x == 'M':
- return 1
- else:
- return 0
-
- def normalize_QP_matrix(x):
- return x/x.sum(axis=0)
-
- def user_flat(user_info_file):
- if not os.path.exists('data/X.txt'):
- u_cols = ['user_id', 'age', 'sex', 'occupation', 'zip_code']
- users = pd.read_csv(user_info_file, sep='|', names=u_cols,
- encoding='latin-1')
- values = users['sex'].apply(lambda x:change_gender(x))
- users['new_sex'] = np.array(values)
- user_age = np.array(users.age.tolist())
- users['new_age'] = (user_age-np.mean(user_age))/np.std(user_age)
- user_side_info = np.array(users[['new_sex','new_age']]).T
- print('[Info]: finish generate user flat side info')
- np.savetxt('data/X.txt', user_side_info)
- else:
- print('[Info]:already have user flat information')
-
- def user_hierarchy(user_info_file,occupation_file):
- if not os.path.exists('data/user_hierarchy.txt'):
- u_cols = ['user_id', 'age', 'sex', 'occupation', 'zip_code']
- users = pd.read_csv(user_info_file, sep='|', names=u_cols,
- encoding='latin-1')
- # user hierarchical structure
- occupation = pd.read_csv(occupation_file,sep='\t',encoding = 'latin-1',header=None,names=['occu'])
- # occupation encoding
- occupation_dict = dict(zip(occupation.occu.tolist(),range(occupation.shape[0])))
- user_hier_id = users.user_id.tolist()
- user_hier_occu = users.occupation.tolist()
- user_hier_occu_id = list(map(occupation_dict.get,user_hier_occu))
- users_hierarchical_matrix = np.zeros([len(user_hier_id),len(occupation_dict)])
- for i in range(len(user_hier_id)):
- users_hierarchical_matrix[user_hier_id[i]-1,user_hier_occu_id[i]]=1
- print('[Info]: finish generate user hierarchy info matrix')
- np.savetxt('data/user_hierarchy.txt',normalize_QP_matrix(users_hierarchical_matrix))
- else:
- print('[Info]:already have user hierarchical information')
-
- def item_flat(item_info_file):
- if not os.path.exists('data/Y.txt'):
- # movies contain two level hierarchical structures and movie side informatin(title,release_date)
- m_cols = ['movie_id', 'title', 'release_date', 'video_release_date', 'imdb_url','unknown', 'Action', 'Adventure', 'Animation',"Children's",
- "Comedy", 'Crime','Documentary', 'Drama','Fantasy','Film-Noir','Horror','Musical','Mystery','Romance','Sci-Fi','Thriller','War','Western']
- movies = pd.read_csv(item_info_file, sep='|', names=m_cols,
- encoding='latin-1')
- movies_flat = movies[['unknown', 'Action', 'Adventure', 'Animation',"Children's","Comedy", 'Crime','Documentary', 'Drama','Fantasy','Film-Noir','Horror','Musical','Mystery','Romance','Sci-Fi','Thriller','War','Western']]
- movies_flat_matrix = np.array(movies_flat).T
- print('[Info]: finish generate item flat side info')
- np.savetxt('data/Y.txt',movies_flat_matrix)
- else:
- print('[Info]:already have item flat side information')
-
- def item_hierarchy(item_info_file):
- if not os.path.exists('data/item_hierarchy.txt'):
- # movies contain two level hierarchical structures and movie side informatin(title,release_date)
- m_cols = ['movie_id', 'title', 'release_date', 'video_release_date', 'imdb_url','unknown', 'Action', 'Adventure', 'Animation',"Children's",
- "Comedy", 'Crime','Documentary', 'Drama','Fantasy','Film-Noir','Horror','Musical','Mystery','Romance','Sci-Fi','Thriller','War','Western']
- movies = pd.read_csv(item_info_file, sep='|', names=m_cols,
- encoding='latin-1')
- movies_hierarchical = movies[['unknown', 'Action', 'Adventure', 'Animation',"Children's","Comedy", 'Crime','Documentary', 'Drama','Fantasy','Film-Noir','Horror','Musical','Mystery','Romance','Sci-Fi','Thriller','War','Western']]
- movies_hierarchical_matrix = np.array(movies_hierarchical)
- print('[Info]: finish generate item hierarchy side info')
- np.savetxt('data/item_hierarchy.txt',normalize_QP_matrix(movies_hierarchical_matrix))
- else:
- print('[Info]:already have item hierarchy information')
-
- if __name__ == "__main__":
- user_flat('data/ml-100k/u.user')
- user_hierarchy('data/ml-100k/u.user','data/ml-100k/u.occupation')
- item_flat('data/ml-100k/u.item')
- item_hierarchy('data/ml-100k/u.item')
|