Question

我训练了一个模型，该模型显示所有4个预测并以json格式保存。当我尝试加载它并进行预测时，它仅显示一个预测。可能会发生什么？

我的代码：

test = pd.read_csv('./Data/test.tsv', sep="\t")
from nltk.tokenize import word_tokenize
from nltk import FreqDist
from nltk.stem import SnowballStemmer,WordNetLemmatizer
stemmer=SnowballStemmer('english')
lemma=WordNetLemmatizer()
from string import punctuation
import re
testing = test.Phrase.apply(lambda x: x.lower())
tokenizer = Tokenizer(num_words= 10000)
X_test = tokenizer.texts_to_sequences(testing.values)

X_test = sequence.pad_sequences(X_test, maxlen=48)
json_file = open('model1.json', 'r')
loaded_model_json = json_file.read()    
json_file.close()
loaded_model = model_from_json(loaded_model_json)

# Load weights into new model
loaded_model.load_weights('model1.h5') 
loaded_model.compile(loss='categorical_crossentropy',
                     optimizer=Adam(lr=0.001),
                     metrics=['accuracy'])

prediction = model.predict_classes(X_test,verbose=1)
model.summary()#while training
Layer (type)                 Output Shape              Param #   
=================================================================
embedding_1 (Embedding)      (None, None, 100)         1373200   
_________________________________________________________________
lstm_1 (LSTM)                (None, None, 64)          42240     
_________________________________________________________________
lstm_2 (LSTM)                (None, 32)                12416     
_________________________________________________________________
dense_1 (Dense)              (None, 5)                 165       
=================================================================
Total params: 1,428,021
Trainable params: 1,428,021
Non-trainable params: 0
print(X_test.shape)
(66292, 48)

Answer 1

如果我正确理解了您的问题，则问题出在这里： predict_classes将为您返回最终的预测标签，而不是概率。它将返回概率最高的四个标签之一。如果您需要每个类别的概率，则可能应该使用predict_proba或predict，它们是相同的，例如：

prediction = model.predict(X_test,verbose=1)

Answer 2

错误已解决，无法正确调整测试值。在以下命令中删除错误

tokenizer.fit_on_texts(testing.values)

Keras model.predict（）仅给出一个预测

2 个答案: