NLP（二十七）：DeepCTR框架的使用

一 、算法介紹

左邊deep network，右邊FM，所以叫deepFM

包含兩個部分：

• Part1： FM(Factorization machines)，因子分解機部分

在傳統的一階線性回歸之上，加了一個二次項，可以表達兩兩特征的相互關系。

這里的公式可以簡化，減少計算量，下圖來至於網絡。

• Part2： Deep部分

deep部分是多層dnn網絡。

二、算法實現

實現部分，用Keras實現一個DeepFM 和·清塵·《FM、FMM、DeepFM整理（pytorch）》

講的比較清楚，這里引用keras實現來說明。

整體的網絡結構：

特征編碼#

特征可以分為3類：

• 連續型field，比如數字類型特征
• 單值離散型特征，比如gender，可選為male、female
• 多值離散型，比如tag，可以有多個

連續型field，可以拼接到一起，dense數據。

單值，多值field進行Onehot后，可見單值離散field對應的獨熱向量只有一位取1，而多值離散field對應的獨熱向量有多於一位取1，表示該field可以同時取多個特征值。

label	shop_score	gender=m	gender=f	interest=f	interest=c
0	0.2	1	0	1	1
1	0.8	0	1	0	1

FM 部分#

看公式：

先算 FM一次項：

• 連續型field 可以用Dense(1)層實現
• 單值離散型field 用Embedding(n,1), n是分類中值的個數
• 多值離散型field可以同時取多個特征值，為了batch training，必須對樣本進行補零padding。同樣可以用Embedding實現，因為有多個Embedding，可以取下平均值。

然后計算FM二次項，這里理解比較費勁一點。

·清塵·《FM、FMM、DeepFM整理（pytorch）》 深入淺出的講明白了這個過程，大家可以參見。

我們來看具體實現方面，這里的DeepFM模型CTR預估理論與實戰
講解更容易理解。

假設只有前面的C1和C2兩個Category的特征，詞典大小還是3和2。假設輸入還是C1=2，C2=2(下標從1開始)，則Embedding之后為V2=[e21,e22,e23,e24]和V5=[e51,e52,e53,e54]。

因為xi和xj同時不為零才需要計算，所以上面的公式里需要計算的只有i=2和j=5的情況。因此：

擴展到多個，比如C1,C2,C3，需要算內積

怎么用用矩陣乘法一次計算出來呢？我們可以看看這個

對應的代碼就是

Copy

       square_of_sum = tf.square(reduce_sum( concated_embeds_value, axis=1, keep_dims=True)) sum_of_square = reduce_sum( concated_embeds_value * concated_embeds_value, axis=1, keep_dims=True) cross_term = square_of_sum - sum_of_square cross_term = 0.5 * reduce_sum(cross_term, axis=2, keep_dims=False) 

其中concated_embeds_value是拼接起來的embeds_value。

Deep部分#

DNN比較簡單，FM的輸入和DNN的輸入都是同一個group_embedding_dict。

三、使用movielens 來測試

下載ml-100k 數據集

Copy

wget http://files.grouplens.org/datasets/movielens/ml-100k.zip
unzip ml-100k.zip

安裝相關軟件包，sklearn，deepctr

導入包：

Copy

import pandas import pandas as pd import sklearn from sklearn.metrics import log_loss, roc_auc_score from sklearn.model_selection import train_test_split from sklearn.preprocessing import LabelEncoder from tensorflow.python.keras.preprocessing.sequence import pad_sequences import tensorflow as tf from tqdm import tqdm from deepctr.models import DeepFM from deepctr.feature_column import SparseFeat, VarLenSparseFeat, get_feature_names import numpy as np 

讀取評分數據：

Copy

u_data = pd.read_csv("ml-100k/u.data", sep='\t', header=None) u_data.columns = ['user_id', 'movie_id', 'rating', 'timestamp'] 

有評分的設置為1，隨機采用未評分的

Copy

def neg_sample(u_data, neg_rate=1): # 全局隨機采樣 item_ids = u_data['movie_id'].unique() print('start neg sample') neg_data = [] # 負采樣 for user_id, hist in tqdm(u_data.groupby('user_id')): # 當前用戶movie rated_movie_list = hist['movie_id'].tolist() candidate_set = list(set(item_ids) - set(rated_movie_list)) neg_list_id = np.random.choice(candidate_set, size=len(rated_movie_list) * neg_rate, replace=True) for id in neg_list_id: neg_data.append([user_id, id, -1, 0]) u_data_neg = pd.DataFrame(neg_data) u_data_neg.columns = ['user_id', 'movie_id', 'rating', 'timestamp'] u_data = pandas.concat([u_data, u_data_neg]) print('end neg sample') return u_data 

讀取item數據

Copy

u_item = pd.read_csv("ml-100k/u.item", sep='|', header=None, error_bad_lines=False) genres_columns = ['Action', 'Adventure', 'Animation', 'Children', 'Comedy', 'Crime', 'Documentary', 'Drama', 'Fantasy', 'Film_Noir', 'Horror', 'Musical', 'Mystery', 'Romance', 'Sci-Fi', 'Thriller', 'War', 'Western'] u_item.columns = ['movie_id', 'title', 'release_date', 'video_date', 'url', 'unknown'] + genres_columns 

處理genres並刪除單獨的genres列

Copy

     genres_list = []
    for index, row in u_item.iterrows(): genres = [] for item in genres_columns: if row[item]: genres.append(item) genres_list.append('|'.join(genres)) u_item['genres'] = genres_list for item in genres_columns: del u_item[item] 

讀取用戶信息：

Copy

  # user id | age | gender | occupation(職業) | zip code（郵編，地區) u_user = pd.read_csv("ml-100k/u.user", sep='|', header=None) u_user.columns = ['user_id', 'age', 'gender', 'occupation', 'zip'] 

join到一起：

Copy

 data = pandas.merge(u_data, u_item, on="movie_id", how='left') data = pandas.merge(data, u_user, on="user_id", how='left') data.to_csv('ml-100k/data.csv', index=False) 

處理特征：

Copy

sparse_features = ["movie_id", "user_id", "gender", "age", "occupation", "zip", ] data[sparse_features] = data[sparse_features].astype(str) target = ['rating'] # 評分 data['rating'] = [1 if int(x) >= 0 else 0 for x in data['rating']] 

先特征編碼：

Copy

for feat in sparse_features: lbe = LabelEncoder() data[feat] = lbe.fit_transform(data[feat]) 

處理genres特征，一個movie有多個genres，先拆分，然后編碼為數字，注意是從1開始；由於每個movie的genres長度不一樣，可以計算最大長度，位數不足的后面補零（pad_sequences，在post補0）

Copy

	 def split(x): key_ans = x.split('|') for key in key_ans: if key not in key2index: # Notice : input value 0 is a special "padding",so we do not use 0 to encode valid feature for sequence input key2index[key] = len(key2index) + 1 return list(map(lambda x: key2index[x], key_ans)) key2index = {} genres_list = list(map(split, data['genres'].values)) genres_length = np.array(list(map(len, genres_list))) max_len = max(genres_length) # Notice : padding=`post` genres_list = pad_sequences(genres_list, maxlen=max_len, padding='post', ) 

構建deepctr的特征列，主要分為兩類特征，一是定長的SparseFeat，稀疏的類別特征，二是可變長度的VarLenSparseFeat，像genres這樣的包含多個的。

Copy

	   fixlen_feature_columns = [SparseFeat(feat, data[feat].nunique(), embedding_dim=4) for feat in sparse_features] use_weighted_sequence = False if use_weighted_sequence: varlen_feature_columns = [VarLenSparseFeat(SparseFeat('genres', vocabulary_size=len( key2index) + 1, embedding_dim=4), maxlen=max_len, combiner='mean', weight_name='genres_weight')] # Notice : value 0 is for padding for sequence input feature else: varlen_feature_columns = [VarLenSparseFeat(SparseFeat('genres', vocabulary_size=len( key2index) + 1, embedding_dim=4), maxlen=max_len, combiner='mean', weight_name=None)] # Notice : value 0 is for padding for sequence input feature linear_feature_columns = fixlen_feature_columns + varlen_feature_columns dnn_feature_columns = fixlen_feature_columns + varlen_feature_columns feature_names = get_feature_names(linear_feature_columns + dnn_feature_columns) 

封裝訓練數據，先shuffle（亂排）數據，然后生成dict input數據。

Copy

data = sklearn.utils.shuffle(data) train_model_input = {name: data[name] for name in sparse_features} # train_model_input["genres"] = genres_list 

構建DeepFM模型，由於目標值是0,1，因此采用binary，損失函數用binary_crossentropy

Copy

model = DeepFM(linear_feature_columns, dnn_feature_columns, task='binary') model.compile(optimizer=tf.keras.optimizers.Adam(), loss='binary_crossentropy', metrics=['AUC', 'Precision', 'Recall']) model.summary() 

訓練模型：

Copy

model.fit(train_model_input, data[target].values, batch_size=256, epochs=20, verbose=2, validation_split=0.2 ) 

開始訓練：

Copy

Epoch 1/20 625/625 - 3s - loss: 0.5081 - auc: 0.8279 - precision: 0.7419 - recall: 0.7695 - val_loss: 0.4745 - val_auc: 0.8513 - val_precision: 0.7563 - val_recall: 0.7936 Epoch 2/20 625/625 - 2s - loss: 0.4695 - auc: 0.8538 - precision: 0.7494 - recall: 0.8105 - val_loss: 0.4708 - val_auc: 0.8539 - val_precision: 0.7498 - val_recall: 0.8127 Epoch 3/20 625/625 - 2s - loss: 0.4652 - auc: 0.8564 - precision: 0.7513 - recall: 0.8139 - val_loss: 0.4704 - val_auc: 0.8545 - val_precision: 0.7561 - val_recall: 0.8017 Epoch 4/20 625/625 - 2s - loss: 0.4624 - auc: 0.8579 - precision: 0.7516 - recall: 0.8146 - val_loss: 0.4724 - val_auc: 0.8542 - val_precision: 0.7296 - val_recall: 0.8526 Epoch 5/20 625/625 - 2s - loss: 0.4607 - auc: 0.8590 - precision: 0.7521 - recall: 0.8173 - val_loss: 0.4699 - val_auc: 0.8550 - val_precision: 0.7511 - val_recall: 0.8141 Epoch 6/20 625/625 - 2s - loss: 0.4588 - auc: 0.8602 - precision: 0.7545 - recall: 0.8165 - val_loss: 0.4717 - val_auc: 0.8542 - val_precision: 0.7421 - val_recall: 0.8265 Epoch 7/20 625/625 - 2s - loss: 0.4574 - auc: 0.8610 - precision: 0.7535 - recall: 0.8192 - val_loss: 0.4722 - val_auc: 0.8547 - val_precision: 0.7549 - val_recall: 0.8023 Epoch 8/20 625/625 - 2s - loss: 0.4561 - auc: 0.8619 - precision: 0.7543 - recall: 0.8201 - val_loss: 0.4717 - val_auc: 0.8548 - val_precision: 0.7480 - val_recall: 0.8185 Epoch 9/20 625/625 - 2s - loss: 0.4531 - auc: 0.8643 - precision: 0.7573 - recall: 0.8210 - val_loss: 0.4696 - val_auc: 0.8583 - val_precision: 0.7598 - val_recall: 0.8103 Epoch 10/20 625/625 - 2s - loss: 0.4355 - auc: 0.8768 - precision: 0.7787 - recall: 0.8166 - val_loss: 0.4435 - val_auc: 0.8769 - val_precision: 0.7756 - val_recall: 0.8293 Epoch 11/20 625/625 - 2s - loss: 0.4093 - auc: 0.8923 - precision: 0.7915 - recall: 0.8373 - val_loss: 0.4301 - val_auc: 0.8840 - val_precision: 0.7806 - val_recall: 0.8390 Epoch 12/20 625/625 - 2s - loss: 0.3970 - auc: 0.8988 - precision: 0.7953 - recall: 0.8497 - val_loss: 0.4286 - val_auc: 0.8867 - val_precision: 0.7903 - val_recall: 0.8299 Epoch 13/20 625/625 - 2s - loss: 0.3896 - auc: 0.9029 - precision: 0.8001 - recall: 0.8542 - val_loss: 0.4253 - val_auc: 0.8888 - val_precision: 0.7913 - val_recall: 0.8322 Epoch 14/20 625/625 - 2s - loss: 0.3825 - auc: 0.9067 - precision: 0.8038 - recall: 0.8584 - val_loss: 0.4205 - val_auc: 0.8917 - val_precision: 0.7885 - val_recall: 0.8506 Epoch 15/20 625/625 - 2s - loss: 0.3755 - auc: 0.9102 - precision: 0.8074 - recall: 0.8624 - val_loss: 0.4204 - val_auc: 0.8940 - val_precision: 0.7868 - val_recall: 0.8607 Epoch 16/20 625/625 - 2s - loss: 0.3687 - auc: 0.9136 - precision: 0.8117 - recall: 0.8653 - val_loss: 0.4176 - val_auc: 0.8956 - val_precision: 0.8097 - val_recall: 0.8236 Epoch 17/20 625/625 - 2s - loss: 0.3617 - auc: 0.9170 - precision: 0.8155 - recall: 0.8682 - val_loss: 0.4166 - val_auc: 0.8966 - val_precision: 0.8056 - val_recall: 0.8354 Epoch 18/20 625/625 - 2s - loss: 0.3553 - auc: 0.9201 - precision: 0.8188 - recall: 0.8716 - val_loss: 0.4168 - val_auc: 0.8977 - val_precision: 0.7996 - val_recall: 0.8492 Epoch 19/20 625/625 - 2s - loss: 0.3497 - auc: 0.9227 - precision: 0.8214 - recall: 0.8741 - val_loss: 0.4187 - val_auc: 0.8973 - val_precision: 0.8079 - val_recall: 0.8358 Epoch 20/20 625/625 - 2s - loss: 0.3451 - auc: 0.9248 - precision: 0.8244 - recall: 0.8753 - val_loss: 0.4210 - val_auc: 0.8982 - val_precision: 0.7945 - val_recall: 0.8617 

最后我們測試下數據：

Copy

 pred_ans = model.predict(train_model_input, batch_size=256) count = 0 for (i, j) in zip(pred_ans, data['rating'].values): print(i, j) count += 1 if count > 10: break 

輸出如下：

 

四、參考

• deepFM in pytorch
• 皮果提《Factorization Machines 學習筆記（二）模型方程》
• ·清塵·《FM、FMM、DeepFM整理（pytorch）》
• 用Keras實現一個DeepFM
• https://www.cnblogs.com/xiaoqi/p/deepfm.html

五、實戰

import pandas as pd
from sklearn.metrics import log_loss, roc_auc_score
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelEncoder, MinMaxScaler
from deepctr.models import *
from deepctr.feature_column import SparseFeat, DenseFeat, get_feature_names
from tensorflow.python.keras.callbacks import EarlyStopping
from tensorflow.python.keras.models import  save_model,load_model
from deepctr.layers import custom_objects
import json
class DeepFmModel(object):
    def __init__(self):
        self.data = pd.read_csv('../data/train.txt', sep = "\t")
        # 類別特征(16)
        fixlen_category_columns = ['m_sex', 'm_access_frequencies', 'm_twoA', 'm_twoB', 'm_twoC',
                                   'm_twoD', 'm_twoE', 'm_categoryA', 'm_categoryB', 'm_categoryC',
                                   'm_categoryD', 'm_categoryE', 'm_num_interest_topic',
                                   'num_topic_attention_intersection',
                                   'q_num_topic_words',
                                   'num_topic_interest_intersection'
                                   ]
        # 數值特征(7)
        fixlen_number_columns = ['m_salt_score', 'm_num_atten_topic', 'q_num_title_chars_words',
                                 'q_num_desc_chars_words', 'q_num_desc_words', 'q_num_title_words',
                                 'days_to_invite'
                                 ]
        target = ['label']
        #sparse_類別標簽
        self.sparse_features = fixlen_category_columns
        #dense_線性標簽
        self.dense_features = fixlen_number_columns

        #對數據預處理，nan填充0、-1
        self.data[self.sparse_features] = self.data[self.sparse_features].fillna('-1', )
        self.data[self.dense_features] = self.data[self.dense_features].fillna(0, )
        self.target = ['label']
        #對類別標簽硬編碼，由數字表示類別
        for feat in self.sparse_features:
            lbe = LabelEncoder()
            self.data[feat] = lbe.fit_transform(self.data[feat])
        #對線性標簽歸一化到0-1
        mms = MinMaxScaler(feature_range=(0, 1))
        self.data[self.dense_features] = mms.fit_transform(self.data[self.dense_features])

        #對兩種特征詞嵌入
        fixlen_feature_columns = \
            [SparseFeat(feat, vocabulary_size=self.data[feat].max() + 1, embedding_dim=4)
             for i, feat in enumerate(self.sparse_features)] \
            + [DenseFeat(feat, 1, ) for feat in self.dense_features]
        self.dnn_feature_columns = fixlen_feature_columns
        self.linear_feature_columns = fixlen_feature_columns

        #得到詞嵌入的列名稱
        feature_names = get_feature_names(self.linear_feature_columns + self.dnn_feature_columns)

        self.train, self.test = train_test_split(self.data, test_size=0.1, random_state=2020)
        self.train_model_input = {name: self.train[name] for name in feature_names}
        self.test_model_input = {name: self.test[name] for name in feature_names}

    def train_model(self):
        # 4.Define Model,train,predict and evaluate
        model = DeepFM(self.linear_feature_columns, self.dnn_feature_columns, task='binary')
        model.compile("adam", "binary_crossentropy",
                      metrics=['binary_crossentropy', "accuracy"], )
        #按照某個指標提前停止
        es = EarlyStopping(monitor='val_binary_crossentropy', patience=3)
        hi = model.fit(self.train_model_input, self.train[self.target].values,
                            batch_size=256, epochs=100, verbose=2, validation_split=0.2,callbacks=[es] )

        save_model(model, '../checkpoint/DeepFM.cpt')# save_model, same as before
        with open("../logs/deep_fm_train.txt", "w", encoding="utf8") as f:
            for k, v in hi.history.items():
                f.write(k + "\t" + json.dumps(v) + "\n")

    def predictData(self):
        model = load_model('../checkpoint/DeepFM.cpt',custom_objects)# load_model,just add a parameter
        pred_ans = model.predict(self.test_model_input, batch_size=256)

        with open("../logs/deep_fm_test.txt", "w", encoding="utf8") as f:
            f.write("test LogLoss" + "\t" + str(round(log_loss(self.test[self.target].values, pred_ans), 4)) + "\n")
            f.write("test AUC" + "\t" + str(round(roc_auc_score(self.test[self.target].values, pred_ans), 4)) + "\n")
            import numpy as np
            label = np.squeeze(self.test[self.target].values).astype(int)
            pre = np.around(np.squeeze(pred_ans)).astype(int)
            from sklearn.metrics import accuracy_score
            f.write("accuracy_score" + "\t" + str(accuracy_score(label, pre)) + "\n")
            from sklearn.metrics import recall_score
            f.write("recall_score" + "\t" + str(recall_score(label, pre)) + "\n")
            from sklearn.metrics import precision_score
            f.write("precision_score" + "\t" + str(precision_score(label, pre)) + "\n")
            from sklearn.metrics import f1_score
            f.write("f1_score" + "\t" + str(f1_score(label, pre)) + "\n")
            from sklearn.metrics import confusion_matrix
            confusion = confusion_matrix(label, pre)
            import seaborn as sns
            import matplotlib.pyplot as plt
            sns.heatmap(confusion, annot=True, xticklabels=['Negative', 'Positive'], yticklabels=['Negative', 'Positive'], cmap=plt.cm.Blues)
            plt.ylabel("Label")
            plt.xlabel("Predicted")
            plt.savefig("../logs/deep_fm.png")
            plt.show()


if __name__ == '__main__':
    DeepFmModel = DeepFmModel()
    DeepFmModel.train_model()
    DeepFmModel.predictData()