Python與機(jī)器學(xué)習(xí)（第2版）（微課版） 課件 8-04-拓展實(shí)訓(xùn)：影評數(shù)據(jù)分析拓展實(shí)訓(xùn)說明

項(xiàng)目背景|項(xiàng)目概述|學(xué)習(xí)目標(biāo)|任務(wù)實(shí)施步驟拓展實(shí)訓(xùn)影評數(shù)據(jù)分析應(yīng)用【數(shù)據(jù)挖掘應(yīng)用】課程負(fù)責(zé)人：陳清華01101111011010111101010000101101010100111101

熟練使用循環(huán)神經(jīng)網(wǎng)絡(luò)算法進(jìn)行情感分析；

熟練使用Keras庫設(shè)計(jì)實(shí)現(xiàn)循環(huán)神經(jīng)網(wǎng)絡(luò)；技能學(xué)習(xí)目標(biāo)

會使用混淆矩陣對模型進(jìn)行評估。項(xiàng)目總體要求原始數(shù)據(jù)結(jié)果呈現(xiàn)IMDB是影評數(shù)據(jù)集，這些數(shù)據(jù)標(biāo)識有情感標(biāo)簽（正面/負(fù)面）。數(shù)據(jù)包含50000條電影評論，其中有25000條訓(xùn)練數(shù)據(jù)及25000條用于測試，有著相同數(shù)量的正面與負(fù)面評論。LSTM情感分類準(zhǔn)確率:0.8422LSTM情感分類混淆矩陣Predicted01Accuracy01093715631238110119一、數(shù)據(jù)加載max_features=15000 #設(shè)置最大特征為15000max_len=300 #單個句子最大長度為300batch_size=64 #每個批次大小#加載數(shù)據(jù)(x_train,y_train),(x_test,y_test)=imdb.load_data(num_words=max_features)print(len(x_train),'trainobservations')print(len(x_test),'testobservations')二、數(shù)據(jù)預(yù)處理#使得數(shù)據(jù)長度保持一致x_train_2=sequence.pad_sequences(x_train,maxlen=max_len)x_test_2=sequence.pad_sequences(x_test,maxlen=max_len)print('x_trainshape:',x_train_2.shape)print('x_testshape:',x_test_2.shape)y_train=np.array(y_train)y_test=np.array(y_test)三、模型構(gòu)建與訓(xùn)練#構(gòu)建模型model=Sequential()model.add(Embedding(max_features,128,input_length=max_len))model.add(LSTM(64))model.add(Dense(1,activation='sigmoid'))model.compile('adam','binary_crossentropy',metrics=['accuracy'])#訓(xùn)練模型model.fit(x_train_2,y_train,batch_size=batch_size,epochs=4,validation_split=0.2)三、模型構(gòu)建與訓(xùn)練Trainon20000samples,validateon5000samplesEpoch1/1020000/20000[==============================]-37s2ms/step-loss:0.5066-accuracy:0.7582-val_loss:0.3708-val_accuracy:0.8484...Epoch9/1020000/20000[==============================]-33s2ms/step-loss:0.0608-accuracy:0.9804-val_loss:0.4903-val_accuracy:0.8666Epoch10/1020000/20000[==============================]-33s2ms/step-loss:0.0952-accuracy:0.9656-val_loss:0.4883-val_accuracy:0.8562四、模型評估#模型預(yù)測y_train_predclass=model.predict_classes(x_train_2,batch_size=100)y_test_predclass=model.predict_classes(x_test_2,batch_size=100)y_train_predclass.shape=y_train.shapey_test_predclass.shape=y_test.shape#輸出模型評估結(jié)果print(("\nLSTMBidirectionalSentimentClassification-Testaccuracy:"),(round(accuracy_score(y_test,y_test_predclass),3)))print("\nLSTMBidirectionalSentimentClassificationofTestdata\n",classification_report(y_test,y_test_predclass))print("\nLSTMBidirectionalSentimentClassification-TestConfusionMatrix\n\n",pd.crosstab(y_test,y_test_predclass,rownames=["Actuall"],colnames=["Predicted"]))四、模型評估LSTM情感分類準(zhǔn)確率:0.8422LSTM情感分類混淆矩陣Predicted

0

1Accuracy