請(qǐng)安裝TensorFlow1.0，Python3.5
項(xiàng)目地址：https://github.com/xiaobingchan/movie_recommender

前言
本項(xiàng)目使用文本卷積神經(jīng)網(wǎng)絡(luò)，并使用MovieLens數(shù)據(jù)集完成電影推薦的任務(wù)。
推薦系統(tǒng)在日常的網(wǎng)絡(luò)應(yīng)用中無(wú)處不在，比如網(wǎng)上購(gòu)物、網(wǎng)上買(mǎi)書(shū)、新聞app、社交網(wǎng)絡(luò)、音樂(lè)網(wǎng)站、電影網(wǎng)站等等等等，有人的地方就有推薦。根據(jù)個(gè)人的喜好，相同喜好人群的習(xí)慣等信息進(jìn)行個(gè)性化的內(nèi)容推薦。比如打開(kāi)新聞?lì)惖腶pp，因?yàn)橛辛藗€(gè)性化的內(nèi)容，每個(gè)人看到的新聞首頁(yè)都是不一樣的。
這當(dāng)然是很有用的，在信息爆炸的今天，獲取信息的途徑和方式多種多樣，人們花費(fèi)時(shí)間最多的不再是去哪獲取信息，而是要在眾多的信息中尋找自己感興趣的，這就是信息超載問(wèn)題。為了解決這個(gè)問(wèn)題，推薦系統(tǒng)應(yīng)運(yùn)而生。
協(xié)同過(guò)濾是推薦系統(tǒng)應(yīng)用較廣泛的技術(shù)，該方法搜集用戶的歷史記錄、個(gè)人喜好等信息，計(jì)算與其他用戶的相似度，利用相似用戶的評(píng)價(jià)來(lái)預(yù)測(cè)目標(biāo)用戶對(duì)特定項(xiàng)目的喜好程度。優(yōu)點(diǎn)是會(huì)給用戶推薦未瀏覽過(guò)的項(xiàng)目，缺點(diǎn)呢，對(duì)于新用戶來(lái)說(shuō)，沒(méi)有任何與商品的交互記錄和個(gè)人喜好等信息，存在冷啟動(dòng)問(wèn)題，導(dǎo)致模型無(wú)法找到相似的用戶或商品。
為了解決冷啟動(dòng)的問(wèn)題，通常的做法是對(duì)于剛注冊(cè)的用戶，要求用戶先選擇自己感興趣的話題、群組、商品、性格、喜歡的音樂(lè)類(lèi)型等信息，比如豆瓣FM：

先來(lái)看看數(shù)據(jù)
本項(xiàng)目使用的是MovieLens 1M 數(shù)據(jù)集，包含6000個(gè)用戶在近4000部電影上的1億條評(píng)論。
數(shù)據(jù)集分為三個(gè)文件：用戶數(shù)據(jù)users.dat，電影數(shù)據(jù)movies.dat和評(píng)分?jǐn)?shù)據(jù)ratings.dat。

用戶數(shù)據(jù)
分別有用戶ID、性別、年齡、職業(yè)ID和郵編等字段。

數(shù)據(jù)中的格式：UserID::Gender::Age::Occupation::Zip-code

Gender is denoted by a “M” for male and “F” for female
Age is chosen from the following ranges:

1: “Under 18”
18: “18-24”
25: “25-34”
35: “35-44”
45: “45-49”
50: “50-55”
56: “56+”
Occupation is chosen from the following choices:

0: “other” or not specified
1: “academic/educator”
2: “artist”
3: “clerical/admin”
4: “college/grad student”
5: “customer service”
6: “doctor/health care”
7: “executive/managerial”
8: “farmer”
9: “homemaker”
10: “K-12 student”
11: “l(fā)awyer”
12: “programmer”
13: “retired”
14: “sales/marketing”
15: “scientist”
16: “self-employed”
17: “technician/engineer”
18: “tradesman/craftsman”
19: “unemployed”
20: “writer”

其中UserID、Gender、Age和Occupation都是類(lèi)別字段，其中郵編字段是我們不使用的。
電影數(shù)據(jù)
分別有電影ID、電影名和電影風(fēng)格等字段。

數(shù)據(jù)中的格式：MovieID::Title::Genres

Titles are identical to titles provided by the IMDB (including
year of release)
Genres are pipe-separated and are selected from the following genres:

Action
Adventure
Animation
Children’s
Comedy
Crime
Documentary
Drama
Fantasy
Film-Noir
Horror
Musical
Mystery
Romance
Sci-Fi
Thriller
War
Western

MovieID是類(lèi)別字段，Title是文本，Genres也是類(lèi)別字段
評(píng)分?jǐn)?shù)據(jù)
分別有用戶ID、電影ID、評(píng)分和時(shí)間戳等字段。

數(shù)據(jù)中的格式：UserID::MovieID::Rating::Timestamp

UserIDs range between 1 and 6040
MovieIDs range between 1 and 3952
Ratings are made on a 5-star scale (whole-star ratings only)
Timestamp is represented in seconds since the epoch as returned by time(2)
Each user has at least 20 ratings

評(píng)分字段Rating就是我們要學(xué)習(xí)的targets，時(shí)間戳字段我們不使用。
說(shuō)說(shuō)數(shù)據(jù)預(yù)處理
UserID、Occupation和MovieID不用變。
Gender字段：需要將‘F’和‘M’轉(zhuǎn)換成0和1。
Age字段：要轉(zhuǎn)成7個(gè)連續(xù)數(shù)字0~6。
Genres字段：是分類(lèi)字段，要轉(zhuǎn)成數(shù)字。首先將Genres中的類(lèi)別轉(zhuǎn)成字符串到數(shù)字的字典，然后再將每個(gè)電影的Genres字段轉(zhuǎn)成數(shù)字列表，因?yàn)橛行╇娪笆嵌鄠€(gè)Genres的組合。
Title字段：處理方式跟Genres字段一樣，首先創(chuàng)建文本到數(shù)字的字典，然后將Title中的描述轉(zhuǎn)成數(shù)字的列表。另外Title中的年份也需要去掉。
Genres和Title字段需要將長(zhǎng)度統(tǒng)一，這樣在神經(jīng)網(wǎng)絡(luò)中方便處理。空白部分用‘< PAD >’對(duì)應(yīng)的數(shù)字填充。
數(shù)據(jù)預(yù)處理的代碼可以在項(xiàng)目中找到：load_data函數(shù)

模型設(shè)計(jì)

通過(guò)研究數(shù)據(jù)集中的字段類(lèi)型，我們發(fā)現(xiàn)有一些是類(lèi)別字段，通常的處理是將這些字段轉(zhuǎn)成one hot編碼，但是像UserID、MovieID這樣的字段就會(huì)變成非常的稀疏，輸入的維度急劇膨脹，這是我們不愿意見(jiàn)到的，畢竟我這小筆記本不像大廠動(dòng)輒能處理數(shù)以億計(jì)維度的輸入：）
所以在預(yù)處理數(shù)據(jù)時(shí)將這些字段轉(zhuǎn)成了數(shù)字，我們用這個(gè)數(shù)字當(dāng)做嵌入矩陣的索引，在網(wǎng)絡(luò)的第一層使用了嵌入層，維度是（N，32）和（N，16）。
電影類(lèi)型的處理要多一步，有時(shí)一個(gè)電影有多個(gè)電影類(lèi)型，這樣從嵌入矩陣索引出來(lái)是一個(gè)（n，32）的矩陣，因?yàn)橛卸鄠€(gè)類(lèi)型嘛，我們要將這個(gè)矩陣求和，變成（1，32）的向量。
電影名的處理比較特殊，沒(méi)有使用循環(huán)神經(jīng)網(wǎng)絡(luò)，而是用了文本卷積網(wǎng)絡(luò)，下文會(huì)進(jìn)行說(shuō)明。
從嵌入層索引出特征以后，將各特征傳入全連接層，將輸出再次傳入全連接層，最終分別得到（1，200）的用戶特征和電影特征兩個(gè)特征向量。
我們的目的就是要訓(xùn)練出用戶特征和電影特征，在實(shí)現(xiàn)推薦功能時(shí)使用。得到這兩個(gè)特征以后，就可以選擇任意的方式來(lái)擬合評(píng)分了。我使用了兩種方式，一個(gè)是上圖中畫(huà)出的將兩個(gè)特征做向量乘法，將結(jié)果與真實(shí)評(píng)分做回歸，采用MSE優(yōu)化損失。因?yàn)楸举|(zhì)上這是一個(gè)回歸問(wèn)題，另一種方式是，將兩個(gè)特征作為輸入，再次傳入全連接層，輸出一個(gè)值，將輸出值回歸到真實(shí)評(píng)分，采用MSE優(yōu)化損失。
實(shí)際上第二個(gè)方式的MSE loss在0.8附近，第一個(gè)方式在1附近，5次迭代的結(jié)果。

文本卷積網(wǎng)絡(luò)
網(wǎng)絡(luò)看起來(lái)像下面這樣

圖片來(lái)自Kim Yoon的論文：Convolutional Neural Networks for Sentence Classification

將卷積神經(jīng)網(wǎng)絡(luò)用于文本的文章建議你閱讀Understanding Convolutional Neural Networks for NLP
網(wǎng)絡(luò)的第一層是詞嵌入層，由每一個(gè)單詞的嵌入向量組成的嵌入矩陣。下一層使用多個(gè)不同尺寸（窗口大小）的卷積核在嵌入矩陣上做卷積，窗口大小指的是每次卷積覆蓋幾個(gè)單詞。這里跟對(duì)圖像做卷積不太一樣，圖像的卷積通常用2x2、3x3、5x5之類(lèi)的尺寸，而文本卷積要覆蓋整個(gè)單詞的嵌入向量，所以尺寸是（單詞數(shù)，向量維度），比如每次滑動(dòng)3個(gè)，4個(gè)或者5個(gè)單詞。第三層網(wǎng)絡(luò)是max pooling得到一個(gè)長(zhǎng)向量，最后使用dropout做正則化，最終得到了電影Title的特征。

核心代碼講解
完整代碼請(qǐng)見(jiàn)項(xiàng)目

#嵌入矩陣的維度
embed_dim = 32
#用戶ID個(gè)數(shù)
uid_max = max(features.take(0,1)) + 1 # 6040
#性別個(gè)數(shù)
gender_max = max(features.take(2,1)) + 1 # 1 + 1 = 2
#年齡類(lèi)別個(gè)數(shù)
age_max = max(features.take(3,1)) + 1 # 6 + 1 = 7
#職業(yè)個(gè)數(shù)
job_max = max(features.take(4,1)) + 1# 20 + 1 = 21

#電影ID個(gè)數(shù)
movie_id_max = max(features.take(1,1)) + 1 # 3952
#電影類(lèi)型個(gè)數(shù)
movie_categories_max = max(genres2int.values()) + 1 # 18 + 1 = 19
#電影名單詞個(gè)數(shù)
movie_title_max = len(title_set) # 5216

#對(duì)電影類(lèi)型嵌入向量做加和操作的標(biāo)志，考慮過(guò)使用mean做平均，但是沒(méi)實(shí)現(xiàn)mean
combiner = "sum"

#電影名長(zhǎng)度
sentences_size = title_count # = 15
#文本卷積滑動(dòng)窗口，分別滑動(dòng)2, 3, 4, 5個(gè)單詞
window_sizes = {2, 3, 4, 5}
#文本卷積核數(shù)量
filter_num = 8

#電影ID轉(zhuǎn)下標(biāo)的字典，數(shù)據(jù)集中電影ID跟下標(biāo)不一致，比如第5行的數(shù)據(jù)電影ID不一定是5
movieid2idx = {val[0]:i for i, val in enumerate(movies.values)}

超參
# Number of Epochs
num_epochs = 5
# Batch Size
batch_size = 256

dropout_keep = 0.5
# Learning Rate
learning_rate = 0.0001
# Show stats for every n number of batches
show_every_n_batches = 20

save_dir = './save'

輸入
定義輸入的占位符

def get_inputs():
? ? uid = tf.placeholder(tf.int32, [None, 1], name="uid")
? ? user_gender = tf.placeholder(tf.int32, [None, 1], name="user_gender")
? ? user_age = tf.placeholder(tf.int32, [None, 1], name="user_age")
? ? user_job = tf.placeholder(tf.int32, [None, 1], name="user_job")

? ? movie_id = tf.placeholder(tf.int32, [None, 1], name="movie_id")
? ? movie_categories = tf.placeholder(tf.int32, [None, 18], name="movie_categories")
? ? movie_titles = tf.placeholder(tf.int32, [None, 15], name="movie_titles")
? ? targets = tf.placeholder(tf.int32, [None, 1], name="targets")
? ? LearningRate = tf.placeholder(tf.float32, name = "LearningRate")
? ? dropout_keep_prob = tf.placeholder(tf.float32, name = "dropout_keep_prob")
? ? return uid, user_gender, user_age, user_job, movie_id, movie_categories, movie_titles, targets, LearningRate, dropout_keep_prob

構(gòu)建神經(jīng)網(wǎng)絡(luò)
定義User的嵌入矩陣
def get_user_embedding(uid, user_gender, user_age, user_job):
? ? with tf.name_scope("user_embedding"):
? ? ? ? uid_embed_matrix = tf.Variable(tf.random_uniform([uid_max, embed_dim], -1, 1), name = "uid_embed_matrix")
? ? ? ? uid_embed_layer = tf.nn.embedding_lookup(uid_embed_matrix, uid, name = "uid_embed_layer")

? ? ? ? gender_embed_matrix = tf.Variable(tf.random_uniform([gender_max, embed_dim // 2], -1, 1), name= "gender_embed_matrix")
? ? ? ? gender_embed_layer = tf.nn.embedding_lookup(gender_embed_matrix, user_gender, name = "gender_embed_layer")

? ? ? ? age_embed_matrix = tf.Variable(tf.random_uniform([age_max, embed_dim // 2], -1, 1), name="age_embed_matrix")
? ? ? ? age_embed_layer = tf.nn.embedding_lookup(age_embed_matrix, user_age, name="age_embed_layer")

? ? ? ? job_embed_matrix = tf.Variable(tf.random_uniform([job_max, embed_dim // 2], -1, 1), name = "job_embed_matrix")
? ? ? ? job_embed_layer = tf.nn.embedding_lookup(job_embed_matrix, user_job, name = "job_embed_layer")
? ? return uid_embed_layer, gender_embed_layer, age_embed_layer, job_embed_layer

將User的嵌入矩陣一起全連接生成User的特征
def get_user_feature_layer(uid_embed_layer, gender_embed_layer, age_embed_layer, job_embed_layer):
? ? with tf.name_scope("user_fc"):
? ? ? ? #第一層全連接
? ? ? ? uid_fc_layer = tf.layers.dense(uid_embed_layer, embed_dim, name = "uid_fc_layer", activation=tf.nn.relu)
? ? ? ? gender_fc_layer = tf.layers.dense(gender_embed_layer, embed_dim, name = "gender_fc_layer", activation=tf.nn.relu)
? ? ? ? age_fc_layer = tf.layers.dense(age_embed_layer, embed_dim, name ="age_fc_layer", activation=tf.nn.relu)
? ? ? ? job_fc_layer = tf.layers.dense(job_embed_layer, embed_dim, name = "job_fc_layer", activation=tf.nn.relu)

? ? ? ? #第二層全連接
? ? ? ? user_combine_layer = tf.concat([uid_fc_layer, gender_fc_layer, age_fc_layer, job_fc_layer], 2) ?#(?, 1, 128)
? ? ? ? user_combine_layer = tf.contrib.layers.fully_connected(user_combine_layer, 200, tf.tanh) ?#(?, 1, 200)

? ? ? ? user_combine_layer_flat = tf.reshape(user_combine_layer, [-1, 200])
? ? return user_combine_layer, user_combine_layer_flat

定義Movie ID的嵌入矩陣
def get_movie_id_embed_layer(movie_id):
? ? with tf.name_scope("movie_embedding"):
? ? ? ? movie_id_embed_matrix = tf.Variable(tf.random_uniform([movie_id_max, embed_dim], -1, 1), name = "movie_id_embed_matrix")
? ? ? ? movie_id_embed_layer = tf.nn.embedding_lookup(movie_id_embed_matrix, movie_id, name = "movie_id_embed_layer")
? ? return movie_id_embed_layer

對(duì)電影類(lèi)型的多個(gè)嵌入向量做加和
def get_movie_categories_layers(movie_categories):
? ? with tf.name_scope("movie_categories_layers"):
? ? ? ? movie_categories_embed_matrix = tf.Variable(tf.random_uniform([movie_categories_max, embed_dim], -1, 1), name = "movie_categories_embed_matrix")
? ? ? ? movie_categories_embed_layer = tf.nn.embedding_lookup(movie_categories_embed_matrix, movie_categories, name = "movie_categories_embed_layer")
? ? ? ? if combiner == "sum":
? ? ? ? ? ? movie_categories_embed_layer = tf.reduce_sum(movie_categories_embed_layer, axis=1, keep_dims=True)
? ? # ? ? elif combiner == "mean":

? ? return movie_categories_embed_layer

Movie Title的文本卷積網(wǎng)絡(luò)實(shí)現(xiàn)
def get_movie_cnn_layer(movie_titles):
? ? #從嵌入矩陣中得到電影名對(duì)應(yīng)的各個(gè)單詞的嵌入向量
? ? with tf.name_scope("movie_embedding"):
? ? ? ? movie_title_embed_matrix = tf.Variable(tf.random_uniform([movie_title_max, embed_dim], -1, 1), name = "movie_title_embed_matrix")
? ? ? ? movie_title_embed_layer = tf.nn.embedding_lookup(movie_title_embed_matrix, movie_titles, name = "movie_title_embed_layer")
? ? ? ? movie_title_embed_layer_expand = tf.expand_dims(movie_title_embed_layer, -1)

? ? #對(duì)文本嵌入層使用不同尺寸的卷積核做卷積和最大池化
? ? pool_layer_lst = []
? ? for window_size in window_sizes:
? ? ? ? with tf.name_scope("movie_txt_conv_maxpool_{}".format(window_size)):
? ? ? ? ? ? filter_weights = tf.Variable(tf.truncated_normal([window_size, embed_dim, 1, filter_num],stddev=0.1),name = "filter_weights")
? ? ? ? ? ? filter_bias = tf.Variable(tf.constant(0.1, shape=[filter_num]), name="filter_bias")

? ? ? ? ? ? conv_layer = tf.nn.conv2d(movie_title_embed_layer_expand, filter_weights, [1,1,1,1], padding="VALID", name="conv_layer")
? ? ? ? ? ? relu_layer = tf.nn.relu(tf.nn.bias_add(conv_layer,filter_bias), name ="relu_layer")

? ? ? ? ? ? maxpool_layer = tf.nn.max_pool(relu_layer, [1,sentences_size - window_size + 1 ,1,1], [1,1,1,1], padding="VALID", name="maxpool_layer")
? ? ? ? ? ? pool_layer_lst.append(maxpool_layer)

? ? #Dropout層
? ? with tf.name_scope("pool_dropout"):
? ? ? ? pool_layer = tf.concat(pool_layer_lst, 3, name ="pool_layer")
? ? ? ? max_num = len(window_sizes) * filter_num
? ? ? ? pool_layer_flat = tf.reshape(pool_layer , [-1, 1, max_num], name = "pool_layer_flat")? ? ? ? dropout_layer = tf.nn.dropout(pool_layer_flat, dropout_keep_prob, name = "dropout_layer")
? ? return pool_layer_flat, dropout_layer

將Movie的各個(gè)層一起做全連接

def get_movie_feature_layer(movie_id_embed_layer, movie_categories_embed_layer, dropout_layer):with tf.name_scope("movie_fc"):#第一層全連接movie_id_fc_layer = tf.layers.dense(movie_id_embed_layer, embed_dim, name = "movie_id_fc_layer", activation=tf.nn.relu)movie_categories_fc_layer = tf.layers.dense(movie_categories_embed_layer, embed_dim, name = "movie_categories_fc_layer", activation=tf.nn.relu)#第二層全連接movie_combine_layer = tf.concat([movie_id_fc_layer, movie_categories_fc_layer, dropout_layer], 2) ?#(?, 1, 96)movie_combine_layer = tf.contrib.layers.fully_connected(movie_combine_layer, 200, tf.tanh) ?#(?, 1, 200)movie_combine_layer_flat = tf.reshape(movie_combine_layer, [-1, 200])return movie_combine_layer, movie_combine_layer_flat

構(gòu)建計(jì)算圖

tf.reset_default_graph() train_graph = tf.Graph() with train_graph.as_default():#獲取輸入占位符uid, user_gender, user_age, user_job, movie_id, movie_categories, movie_titles, targets, lr, dropout_keep_prob = get_inputs()#獲取User的4個(gè)嵌入向量uid_embed_layer, gender_embed_layer, age_embed_layer, job_embed_layer = get_user_embedding(uid, user_gender, user_age, user_job)#得到用戶特征user_combine_layer, user_combine_layer_flat = get_user_feature_layer(uid_embed_layer, gender_embed_layer, age_embed_layer, job_embed_layer)#獲取電影ID的嵌入向量movie_id_embed_layer = get_movie_id_embed_layer(movie_id)#獲取電影類(lèi)型的嵌入向量movie_categories_embed_layer = get_movie_categories_layers(movie_categories)#獲取電影名的特征向量pool_layer_flat, dropout_layer = get_movie_cnn_layer(movie_titles)#得到電影特征movie_combine_layer, movie_combine_layer_flat = get_movie_feature_layer(movie_id_embed_layer,?movie_categories_embed_layer,?dropout_layer)#計(jì)算出評(píng)分，要注意兩個(gè)不同的方案，inference的名字（name值）是不一樣的，后面做推薦時(shí)要根據(jù)name取得tensorwith tf.name_scope("inference"):#將用戶特征和電影特征作為輸入，經(jīng)過(guò)全連接，輸出一個(gè)值的方案 # ? ? ? ? inference_layer = tf.concat([user_combine_layer_flat, movie_combine_layer_flat], 1) ?#(?, 200) # ? ? ? ? inference = tf.layers.dense(inference_layer, 1, # ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? kernel_initializer=tf.truncated_normal_initializer(stddev=0.01),? # ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? kernel_regularizer=tf.nn.l2_loss, name="inference")#簡(jiǎn)單的將用戶特征和電影特征做矩陣乘法得到一個(gè)預(yù)測(cè)評(píng)分inference = tf.matmul(user_combine_layer_flat, tf.transpose(movie_combine_layer_flat))? ? with tf.name_scope("loss"):# MSE損失，將計(jì)算值回歸到評(píng)分cost = tf.losses.mean_squared_error(targets, inference )loss = tf.reduce_mean(cost)# 優(yōu)化損失? # ? ? train_op = tf.train.AdamOptimizer(lr).minimize(loss) ?#costglobal_step = tf.Variable(0, name="global_step", trainable=False)optimizer = tf.train.AdamOptimizer(lr)gradients = optimizer.compute_gradients(loss) ?#costtrain_op = optimizer.apply_gradients(gradients, global_step=global_step)

訓(xùn)練網(wǎng)絡(luò)

%matplotlib inline %config InlineBackend.figure_format = 'retina' import matplotlib.pyplot as plt import time import datetimelosses = {'train':[], 'test':[]}with tf.Session(graph=train_graph) as sess:#搜集數(shù)據(jù)給tensorBoard用# Keep track of gradient values and sparsitygrad_summaries = []for g, v in gradients:if g is not None:grad_hist_summary = tf.summary.histogram("{}/grad/hist".format(v.name.replace(':', '_')), g)sparsity_summary = tf.summary.scalar("{}/grad/sparsity".format(v.name.replace(':', '_')), tf.nn.zero_fraction(g))grad_summaries.append(grad_hist_summary)grad_summaries.append(sparsity_summary)grad_summaries_merged = tf.summary.merge(grad_summaries)# Output directory for models and summariestimestamp = str(int(time.time()))out_dir = os.path.abspath(os.path.join(os.path.curdir, "runs", timestamp))print("Writing to {}\n".format(out_dir))# Summaries for loss and accuracyloss_summary = tf.summary.scalar("loss", loss)# Train Summariestrain_summary_op = tf.summary.merge([loss_summary, grad_summaries_merged])train_summary_dir = os.path.join(out_dir, "summaries", "train")train_summary_writer = tf.summary.FileWriter(train_summary_dir, sess.graph)# Inference summariesinference_summary_op = tf.summary.merge([loss_summary])inference_summary_dir = os.path.join(out_dir, "summaries", "inference")inference_summary_writer = tf.summary.FileWriter(inference_summary_dir, sess.graph)sess.run(tf.global_variables_initializer())saver = tf.train.Saver()for epoch_i in range(num_epochs):#將數(shù)據(jù)集分成訓(xùn)練集和測(cè)試集，隨機(jī)種子不固定train_X,test_X, train_y, test_y = train_test_split(features, ?targets_values, ?test_size = 0.2, ?random_state = 0) ?train_batches = get_batches(train_X, train_y, batch_size)test_batches = get_batches(test_X, test_y, batch_size)#訓(xùn)練的迭代，保存訓(xùn)練損失for batch_i in range(len(train_X) // batch_size):x, y = next(train_batches)categories = np.zeros([batch_size, 18])for i in range(batch_size):categories[i] = x.take(6,1)[i]titles = np.zeros([batch_size, sentences_size])for i in range(batch_size):titles[i] = x.take(5,1)[i]feed = {uid: np.reshape(x.take(0,1), [batch_size, 1]),user_gender: np.reshape(x.take(2,1), [batch_size, 1]),user_age: np.reshape(x.take(3,1), [batch_size, 1]),user_job: np.reshape(x.take(4,1), [batch_size, 1]),movie_id: np.reshape(x.take(1,1), [batch_size, 1]),movie_categories: categories, ?#x.take(6,1)movie_titles: titles, ?#x.take(5,1)targets: np.reshape(y, [batch_size, 1]),dropout_keep_prob: dropout_keep, #dropout_keeplr: learning_rate}step, train_loss, summaries, _ = sess.run([global_step, loss, train_summary_op, train_op], feed) ?#costlosses['train'].append(train_loss)train_summary_writer.add_summary(summaries, step) ?## Show every <show_every_n_batches> batchesif (epoch_i * (len(train_X) // batch_size) + batch_i) % show_every_n_batches == 0:time_str = datetime.datetime.now().isoformat()print('{}: Epoch {:>3} Batch {:>4}/{} ? train_loss = {:.3f}'.format(time_str,epoch_i,batch_i,(len(train_X) // batch_size),train_loss))#使用測(cè)試數(shù)據(jù)的迭代for batch_i ?in range(len(test_X) // batch_size):x, y = next(test_batches)categories = np.zeros([batch_size, 18])for i in range(batch_size):categories[i] = x.take(6,1)[i]titles = np.zeros([batch_size, sentences_size])for i in range(batch_size):titles[i] = x.take(5,1)[i]feed = {uid: np.reshape(x.take(0,1), [batch_size, 1]),user_gender: np.reshape(x.take(2,1), [batch_size, 1]),user_age: np.reshape(x.take(3,1), [batch_size, 1]),user_job: np.reshape(x.take(4,1), [batch_size, 1]),movie_id: np.reshape(x.take(1,1), [batch_size, 1]),movie_categories: categories, ?#x.take(6,1)movie_titles: titles, ?#x.take(5,1)targets: np.reshape(y, [batch_size, 1]),dropout_keep_prob: 1,lr: learning_rate}step, test_loss, summaries = sess.run([global_step, loss, inference_summary_op], feed) ?#cost#保存測(cè)試損失losses['test'].append(test_loss)inference_summary_writer.add_summary(summaries, step) ?#time_str = datetime.datetime.now().isoformat()if (epoch_i * (len(test_X) // batch_size) + batch_i) % show_every_n_batches == 0:print('{}: Epoch {:>3} Batch {:>4}/{} ? test_loss = {:.3f}'.format(time_str,epoch_i,batch_i,(len(test_X) // batch_size),test_loss))# Save Modelsaver.save(sess, save_dir) ?#, global_step=epoch_iprint('Model Trained and Saved')

在 TensorBoard 中查看可視化結(jié)果
獲取 Tensors

def get_tensors(loaded_graph):uid = loaded_graph.get_tensor_by_name("uid:0")user_gender = loaded_graph.get_tensor_by_name("user_gender:0")user_age = loaded_graph.get_tensor_by_name("user_age:0")user_job = loaded_graph.get_tensor_by_name("user_job:0")movie_id = loaded_graph.get_tensor_by_name("movie_id:0")movie_categories = loaded_graph.get_tensor_by_name("movie_categories:0")movie_titles = loaded_graph.get_tensor_by_name("movie_titles:0")targets = loaded_graph.get_tensor_by_name("targets:0")dropout_keep_prob = loaded_graph.get_tensor_by_name("dropout_keep_prob:0")lr = loaded_graph.get_tensor_by_name("LearningRate:0")#兩種不同計(jì)算預(yù)測(cè)評(píng)分的方案使用不同的name獲取tensor inference # ? ? inference = loaded_graph.get_tensor_by_name("inference/inference/BiasAdd:0")inference = loaded_graph.get_tensor_by_name("inference/MatMul:0")#movie_combine_layer_flat = loaded_graph.get_tensor_by_name("movie_fc/Reshape:0")user_combine_layer_flat = loaded_graph.get_tensor_by_name("user_fc/Reshape:0")return uid, user_gender, user_age, user_job, movie_id, movie_categories, movie_titles, targets, lr, dropout_keep_prob, inference, movie_combine_layer_flat, user_combine_layer_flat

指定用戶和電影進(jìn)行評(píng)分
這部分就是對(duì)網(wǎng)絡(luò)做正向傳播，計(jì)算得到預(yù)測(cè)的評(píng)分

def rating_movie(user_id_val, movie_id_val):loaded_graph = tf.Graph() ?#with tf.Session(graph=loaded_graph) as sess: ?## Load saved modelloader = tf.train.import_meta_graph(load_dir + '.meta')loader.restore(sess, load_dir)# Get Tensors from loaded modeluid, user_gender, user_age, user_job, movie_id, movie_categories, movie_titles, targets, lr, dropout_keep_prob, inference,_, __ = get_tensors(loaded_graph) ?#loaded_graphcategories = np.zeros([1, 18])categories[0] = movies.values[movieid2idx[movie_id_val]][2]titles = np.zeros([1, sentences_size])titles[0] = movies.values[movieid2idx[movie_id_val]][1]feed = {uid: np.reshape(users.values[user_id_val-1][0], [1, 1]),user_gender: np.reshape(users.values[user_id_val-1][1], [1, 1]),user_age: np.reshape(users.values[user_id_val-1][2], [1, 1]),user_job: np.reshape(users.values[user_id_val-1][3], [1, 1]),movie_id: np.reshape(movies.values[movieid2idx[movie_id_val]][0], [1, 1]),movie_categories: categories, ?#x.take(6,1)movie_titles: titles, ?#x.take(5,1)dropout_keep_prob: 1}# Get Predictioninference_val = sess.run([inference], feed) ?return (inference_val)

生成Movie特征矩陣
將訓(xùn)練好的電影特征組合成電影特征矩陣并保存到本地

loaded_graph = tf.Graph() ?# movie_matrics = [] with tf.Session(graph=loaded_graph) as sess: ?## Load saved modelloader = tf.train.import_meta_graph(load_dir + '.meta')loader.restore(sess, load_dir)# Get Tensors from loaded modeluid, user_gender, user_age, user_job, movie_id, movie_categories, movie_titles, targets, lr, dropout_keep_prob, _, movie_combine_layer_flat, __ = get_tensors(loaded_graph) ?#loaded_graphfor item in movies.values:categories = np.zeros([1, 18])categories[0] = item.take(2)titles = np.zeros([1, sentences_size])titles[0] = item.take(1)feed = {movie_id: np.reshape(item.take(0), [1, 1]),movie_categories: categories, ?#x.take(6,1)movie_titles: titles, ?#x.take(5,1)dropout_keep_prob: 1}movie_combine_layer_flat_val = sess.run([movie_combine_layer_flat], feed) ?movie_matrics.append(movie_combine_layer_flat_val)pickle.dump((np.array(movie_matrics).reshape(-1, 200)), open('movie_matrics.p', 'wb')) movie_matrics = pickle.load(open('movie_matrics.p', mode='rb'))

生成User特征矩陣
將訓(xùn)練好的用戶特征組合成用戶特征矩陣并保存到本地

loaded_graph = tf.Graph() ?# users_matrics = [] with tf.Session(graph=loaded_graph) as sess: ?## Load saved modelloader = tf.train.import_meta_graph(load_dir + '.meta')loader.restore(sess, load_dir)# Get Tensors from loaded modeluid, user_gender, user_age, user_job, movie_id, movie_categories, movie_titles, targets, lr, dropout_keep_prob, _, __,user_combine_layer_flat = get_tensors(loaded_graph) ?#loaded_graphfor item in users.values:feed = {uid: np.reshape(item.take(0), [1, 1]),user_gender: np.reshape(item.take(1), [1, 1]),user_age: np.reshape(item.take(2), [1, 1]),user_job: np.reshape(item.take(3), [1, 1]),dropout_keep_prob: 1}user_combine_layer_flat_val = sess.run([user_combine_layer_flat], feed) ?users_matrics.append(user_combine_layer_flat_val)pickle.dump((np.array(users_matrics).reshape(-1, 200)), open('users_matrics.p', 'wb')) users_matrics = pickle.load(open('users_matrics.p', mode='rb'))

開(kāi)始推薦電影
使用生產(chǎn)的用戶特征矩陣和電影特征矩陣做電影推薦

推薦同類(lèi)型的電影
思路是計(jì)算當(dāng)前看的電影特征向量與整個(gè)電影特征矩陣的余弦相似度，取相似度最大的top_k個(gè)，這里加了些隨機(jī)選擇在里面，保證每次的推薦稍稍有些不同。

def recommend_same_type_movie(movie_id_val, top_k = 20):loaded_graph = tf.Graph() ?#with tf.Session(graph=loaded_graph) as sess: ?## Load saved modelloader = tf.train.import_meta_graph(load_dir + '.meta')loader.restore(sess, load_dir)norm_movie_matrics = tf.sqrt(tf.reduce_sum(tf.square(movie_matrics), 1, keep_dims=True))normalized_movie_matrics = movie_matrics / norm_movie_matrics#推薦同類(lèi)型的電影probs_embeddings = (movie_matrics[movieid2idx[movie_id_val]]).reshape([1, 200])probs_similarity = tf.matmul(probs_embeddings, tf.transpose(normalized_movie_matrics))sim = (probs_similarity.eval())# ? ? results = (-sim[0]).argsort()[0:top_k]# ? ? print(results)print("您看的電影是：{}".format(movies_orig[movieid2idx[movie_id_val]]))print("以下是給您的推薦：")p = np.squeeze(sim)p[np.argsort(p)[:-top_k]] = 0p = p / np.sum(p)results = set()while len(results) != 5:c = np.random.choice(3883, 1, p=p)[0]results.add(c)for val in (results):print(val)print(movies_orig[val])return results

推薦您喜歡的電影
思路是使用用戶特征向量與電影特征矩陣計(jì)算所有電影的評(píng)分，取評(píng)分最高的top_k個(gè)，同樣加了些隨機(jī)選擇部分。

def recommend_your_favorite_movie(user_id_val, top_k = 10):loaded_graph = tf.Graph() ?#with tf.Session(graph=loaded_graph) as sess: ?## Load saved modelloader = tf.train.import_meta_graph(load_dir + '.meta')loader.restore(sess, load_dir)#推薦您喜歡的電影probs_embeddings = (users_matrics[user_id_val-1]).reshape([1, 200])probs_similarity = tf.matmul(probs_embeddings, tf.transpose(movie_matrics))sim = (probs_similarity.eval())# ? ? print(sim.shape)# ? ? results = (-sim[0]).argsort()[0:top_k]# ? ? print(results)# ? ? sim_norm = probs_norm_similarity.eval()# ? ? print((-sim_norm[0]).argsort()[0:top_k])print("以下是給您的推薦：")p = np.squeeze(sim)p[np.argsort(p)[:-top_k]] = 0p = p / np.sum(p)results = set()while len(results) != 5:c = np.random.choice(3883, 1, p=p)[0]results.add(c)for val in (results):print(val)print(movies_orig[val])return results

看過(guò)這個(gè)電影的人還看了（喜歡）哪些電影
首先選出喜歡某個(gè)電影的top_k個(gè)人，得到這幾個(gè)人的用戶特征向量。
然后計(jì)算這幾個(gè)人對(duì)所有電影的評(píng)分
選擇每個(gè)人評(píng)分最高的電影作為推薦
同樣加入了隨機(jī)選擇

import randomdef recommend_other_favorite_movie(movie_id_val, top_k = 20):loaded_graph = tf.Graph() ?#with tf.Session(graph=loaded_graph) as sess: ?## Load saved modelloader = tf.train.import_meta_graph(load_dir + '.meta')loader.restore(sess, load_dir)probs_movie_embeddings = (movie_matrics[movieid2idx[movie_id_val]]).reshape([1, 200])probs_user_favorite_similarity = tf.matmul(probs_movie_embeddings, tf.transpose(users_matrics))favorite_user_id = np.argsort(probs_user_favorite_similarity.eval())[0][-top_k:]# ? ? print(normalized_users_matrics.eval().shape)# ? ? print(probs_user_favorite_similarity.eval()[0][favorite_user_id])# ? ? print(favorite_user_id.shape)print("您看的電影是：{}".format(movies_orig[movieid2idx[movie_id_val]]))print("喜歡看這個(gè)電影的人是：{}".format(users_orig[favorite_user_id-1]))probs_users_embeddings = (users_matrics[favorite_user_id-1]).reshape([-1, 200])probs_similarity = tf.matmul(probs_users_embeddings, tf.transpose(movie_matrics))sim = (probs_similarity.eval())# ? ? results = (-sim[0]).argsort()[0:top_k]# ? ? print(results)# ? ? print(sim.shape)# ? ? print(np.argmax(sim, 1))p = np.argmax(sim, 1)print("喜歡看這個(gè)電影的人還喜歡看：")results = set()while len(results) != 5:c = p[random.randrange(top_k)]results.add(c)for val in (results):print(val)print(movies_orig[val])return results

結(jié)論
以上就是實(shí)現(xiàn)的常用的推薦功能，將網(wǎng)絡(luò)模型作為回歸問(wèn)題進(jìn)行訓(xùn)練，得到訓(xùn)練好的用戶特征矩陣和電影特征矩陣進(jìn)行推薦。

?

?

總結(jié)

以上是生活随笔為你收集整理的TensorFlow实战——深度学习训练个性化推荐系统的全部?jī)?nèi)容，希望文章能夠幫你解決所遇到的問(wèn)題。

如果覺(jué)得生活随笔網(wǎng)站內(nèi)容還不錯(cuò)，歡迎將生活随笔推薦給好友。