def predict(self, query_text, predictor_list):
preds = []
selected_models = []
print(predictor_list)
for key in predictor_list:
print(key)
if key in self.__trained_models.keys():
tmp_model = self.__trained_models[key]
selected_models.append(tmp_model)
print(selected_models)
for (tokenizer, model) in selected_models:
print(tokenizer)
print(model)
query_vectorized = Utils.create_document_term_matrix(query_text, tokenizer)
predictions = model.predict(query_vectorized)
prediction_vectorized = np.argmax(predictions, axis=1)
prediction_raw = Utils.devectorize_y(prediction_vectorized, self.__label_encoder)
preds.append(prediction_raw)
if len(preds) == 0:
return 'xx'
preds = zip(*preds)
print('preds', preds)
prediction = Utils.most_common(preds)
return prediction
def predict_batch(self, X_test_raw, predictor_list):
preds = []
if(len(predictor_list) > 1):
selected_models = operator.itemgetter(*predictor_list)(self.__trained_models)
else:
selected_models = [self.__trained_models[0]]
for (tokenizer, model) in selected_models:
X_test_vectorized = Utils.create_document_term_matrix(X_test_raw, tokenizer)
predictions = model.predict(X_test_vectorized)
Y_test_predicted_vectorized = np.argmax(predictions, axis=1)
Y_test_raw = Utils.devectorize_y(Y_test_predicted_vectorized, self.__label_encoder)
preds.append(Y_test_raw)
preds = zip(*preds)
Y_test_predicted = map(lambda x: Utils.most_common(x), preds)
return Y_test_predicted
