数据清洗

用python 处理数据movielens-100， 并用matlab文件保存。

# coding: utf-8
# ml-k100.py
import pandas as pd
import numpy as np
import time
import scipy.io as sio 


def to_date(x):
    if len(x)==0:
        return 1995
    tt = time.strptime(x.decode(), '%d-%b-%Y')
    return tt.tm_year + round(tt.tm_yday/ 366, 3)
    
def parse_line(line):
    item, title, date, _, _, *genres =  line.split(b'|') 
    #if len(date)==0: 
    #    print(line)  
    return [int(item), to_date(date)] + [int(v) for v in genres]

def transform(age):
    V = pd.qcut(age, 10)
    V1 = V.value_counts()
    v2 = V1.sort_index().cumsum() / len(age)
    return pd.Categorical.from_codes(V.cat.codes, v2.tolist()).astype(float) 


D = open('ml-100k/u.item', 'rb').read().splitlines()
D1 = pd.DataFrame(data=list(map(parse_line, D)), columns=['item', 'date'] + ['f{}'.format(i) for i in range(19)])
D1.date = transform(D1.date)

D2 = pd.read_csv('ml-100k/u.user', sep='|', header=None, names=['user','age','gender','occupation','zipcode'])
D2.age = transform(D2.age)
D3 = pd.get_dummies(D2.drop('zipcode', axis=1),  columns=['gender','occupation'])

D4 = pd.read_csv('ml-100k/u.data',sep='\t', header=None, names=['user','item','rate','timestamp'])
D4.drop('timestamp', axis=1,inplace=True)


sio.savemat('ml-100k.mat', {'user': D3.values, 'item': D1.values, 'rate': D4.values})