基于协同过滤算法的电影推荐实例

import pandas as pd 
import numpy as np  
import tensorflow as tf 


#Step1:数据特征处理
ratings_df = pd.read_csv("./data/ratings.csv")
movies_df = pd.read_csv("./data/movies.csv")


movies_df['movieRow'] = movies_df.index 
#筛选特征
movies_df = movies_df[['movieRow','movieId','title']]
movies_df.to_csv('moviesProcessed.csv',index=False,header=True,encoding='utf-8')
#用电影的行号代替电影的ID，节省内存
ratings_df = pd.merge(ratings_df,movies_df,on='movieId')
ratings_df = ratings_df[['userId','movieRow','rating']]
ratings_df.to_csv('ratingsProcessed.csv',index=False,header=True,encoding='utf-8')


#建立一个电影评分表rating：用户对每个电影的评分高低
#获取UserId的最大编号
userNo = ratings_df['userId'].max() + 1 
#获取电影ID的最大编号
movieNo = ratings_df['movieRow'].max() + 1 
#rating的维度：[movieNo,userNo]
rating = np.zeros((movieNo,userNo))
flag = 0 
ratings_df_length = np.shape(ratings_df)[0]

for index,row in ratings_df.iterrows():
    #iterrows()返回值为元组:(index,row)
    rating[int(row['movieRow']),int(row['userId'])]=row['rating']
    flag += 1 
    #打印进度
    print('processed %d,%d left' % (flag,ratings_df_length-flag))

#建立一个评分记录表record：评过分电影的对应位置为1，没有评过分的对应位置为0 
record = rating > 0 
record = np.array(record,dtype=int)

#Step2:构建模型

#对评分数据进行标准
def normalizeRatings(rating,record):
    m,n = rating.shape 
    #每个电影的评分均值
    rating_mean = np.zeros((m,1))
    #电影的真实评分
    rating_norm = np.zeros((m,n))
    for i in range(m):
        #统计一行中所有评分的，因为没评分的不参与平均值的计算
        idx = record[i,:] !=0 
        rating_mean[i] = np.mean(rating[i,idx])
        rating_norm[i,idx] -= rating_mean[i]
    return rating_norm,rating_mean

rating_norm,rating_mean = normalizeRatings(rating,record)
#处理Nan
rating_norm = np.nan_to_num(rating_norm)
rating_mean = np.nan_to_num(rating_mean)

#电影类型数量
num_features = 10
#电影内容矩阵：X_parameters
X_parameters = tf.Variable(tf.random_normal([movieNo,num_features],stddev=0.35))
#用户喜好矩阵：Theta_parameters
Theta_parameters = tf.Variable(tf.random_normal([userNo,num_features],stddev=0.35))


#损失函数
#transpose_b:对Theta_parameters进行转置 
xietong_formula = (1/2) * tf.reduce_sum(((tf.matmul(X_parameters,Theta_parameters,transpose_b = True) - rating_norm)*record)**2)
loss = xietong_formula + (1/2) * (tf.reduce_sum(X_parameters**2) + tf.reduce_sum(Theta_parameters**2))

#优化器和优化目标
optimizer = tf.train.AdamOptimizer(1e-4)
optim     = optimizer.minimize(loss)



#Step3：训练模型

#用tf.summary可视化训练过程
tf.summary.scalar('loss',loss)
#汇总summary信息
summaryMerged = tf.summary.merge_all()
#把信息保存到文件
filename = './movie_tensorboard'
writer = tf.summary.FileWriter(filename)


sess = tf.Session()
init = tf.global_variables_initializer()
sess.run(init)

for i in range(100):
    _,movie_summary = sess.run([optim,summaryMerged])
    writer.add_summary(movie_summary,i)


#Step4:模型评估
Current_X_parameters,Current_Theta_parameters = sess.run([X_parameters,Theta_parameters])
predicts = np.dot(Current_X_parameters,Current_Theta_parameters.T) + rating_mean
errors   = np.sqrt(np.sum((predicts - rating)**2))
print(errors)

参考教程：https://www.imooc.com/video/17130