第2关：动手搭建电影推荐系统 实验记录

# -*- coding: utf-8 -*-
import numpy as np
def build_rating(m,n,ratings_df):
    '''
    m(int):用户数
    n(int):电影数
    ratings_df(DataFrame):用户对电影评分记录表
    rating(ndarray):用户-电影评分矩阵
    '''
    #*********Begin*********#
    #创建电影评分矩阵
    #获取用户数与电影数
    userNo = m
    movieNo = n
    #创建电影评分表
    rating = np.zeros((userNo,movieNo))
    for index,row in ratings_df.iterrows():
        rating[int(row['userId']),int(row['movieRow'])]=row['rating']
    #*********End*********#
    return rating
#训练模型
def fit(lr,alpha,d,n_iter,data):
    '''
    userID(int):推荐用户ID
    lr(float):学习率
    alpha(float):权重衰减系数
    d(int):矩阵分解因子
    n_iter(int):训练轮数
    data(ndarray):用户-电影评分矩阵
    ''' 
    #*********Begin*********#
    #获取用户数与电影数
    m,n = data.shape 
    #初始化喜好矩阵与内容矩阵，值为0-1之间，符合正态分布  
    x = np.random.uniform(0,1,(m,d))
    w = np.random.uniform(0,1,(d,n))
    #创建评分记录表，无评分记为0，有评分记为1
    record = np.array(data>0,dtype=int)
    #梯度下降，更新参数           
    for i in range(n_iter):
        x_grads = np.dot(np.multiply(record,np.dot(x,w)-data),w.T)
        w_grads = np.dot(x.T,np.multiply(record,np.dot(x,w)-data))
        x = alpha*x - lr*x_grads
        w = alpha*w - lr*w_grads
    #预测
    predict = np.dot(x,w)
    #*********End*********#
    return predict
#进行推荐   
def recommend(userID,predict,data):
    '''
    userID(int):推荐用户ID
    predict(ndarray):预测结果
    data(ndarray):用户-电影评分矩阵
    '''
    #*********Begin*********#
    #将用户未看过的电影分值从低到高进行排列
    recommend = np.argsort(predict[userID-1])
    a = recommend[-1]
    b = recommend[-2]
    c = recommend[-3]
    d = recommend[-4]
    e = recommend[-5]
    #*********End*********#
    return recommend
    
    

---

# -*- coding: utf-8 -*-
import numpy as np
def build_rating(m,n,ratings_df):
    '''
    m(int):用户数
    n(int):电影数
    ratings_df(DataFrame):用户对电影评分记录表
    rating(ndarray):用户-电影评分矩阵
    '''
    #*********Begin*********#
    #创建电影评分矩阵
    rating = np.zeros((m,n))
    for index,row in ratings_df.iterrows():
        rating[int(row['userId']),int(row['movieRow'])]=row['rating']
    #*********End*********#
    return rating
#训练模型
def fit(lr,alpha,d,n_iter,data):
    '''
    userID(int):推荐用户ID
    lr(float):学习率
    alpha(float):权重衰减系数
    d(int):矩阵分解因子
    n_iter(int):训练轮数
    data(ndarray):用户-电影评分矩阵
    ''' 
    #*********Begin*********#
    #获取用户数与电影数
    m,n = data.shape 
    #初始化喜好矩阵与内容矩阵，值为0-1之间，符合正态分布  
    x = np.zeros((m,d))
    w = np.zeros((d,n))
    i = 0
    while  i < m:
        j = 0
        while j < d:
            a = np.random.normal(loc = 0.5, scale = 0.25)
            if a >= 0 and a <= 1:
                x[i][j] = a
                j += 1
        i += 1
    i = 0
    while i < d:
        j = 0
        while j < n:
            a = np.random.normal(loc = 0.5, scale = 0.25)
            if a >= 0 and a <= 1:
                w[i][j] = a
                j += 1
        i += 1
    #创建评分记录表，无评分记为0，有评分记为1
    record = np.zeros(data.shape)
    i = 0
    while i < m:
        j = 0
        while j < n:
            if data[i][j] == 0:
                record[i][j] = 0
            else :
                record[i][j] = 1  
            j += 1
        i += 1
    #梯度下降，更新参数           
    loss = np.mean(np.multiply((data-np.dot(x,w))**2,record))
    x_grads = np.dot(np.multiply(record,np.dot(x,w)-data),w.T)
    w_grads = np.dot(x.T,np.multiply(record,np.dot(x,w)-data))
    for i in range(n_iter):
        x_grads = np.dot(np.multiply(record,np.dot(x,w)-data),w.T)
        w_grads = np.dot(x.T,np.multiply(record,np.dot(x,w)-data))
        x = alpha*x - lr*x_grads
        w = alpha*w - lr*w_grads
    #预测
    predict = np.dot(x,w)
    #*********End*********#
    return predict
#进行推荐   
def recommend(userID,predict,data):
    '''
    userID(int):推荐用户ID
    predict(ndarray):预测结果
    data(ndarray):用户-电影评分矩阵
    '''
    #*********Begin*********#
    #将用户未看过的电影分值从低到高进行排列
    a,b = data.shape 
    record = np.zeros(data.shape)
    i = 0
    while i < a:
        j = 0
        while j < b:
            if data[i][j] == 0:
                record[i][j] = 0
            else :
                record[i][j] = 1  
            j += 1
        i += 1
    r = record[userID - 1]
    m = predict[userID - 1]
    n = range(b)
    dic = dict(zip(n,m))
    for i in range(b):
        if r[i] == 1:
            del dic[i]
    recommend = np.zeros(len(dic))
    for i in range(len(dic)):
        k = list(dic.keys())
        recommend[i] = k[0]
        for l in dic.keys():
            if dic[l] >= dic[recommend[i]]:
                recommend[i] = l
        del dic[recommend[i]]
    recommend = recommend[::-1]
    #*********End*********#
    return recommend