使用tflearn训练文本分类模型后，用sklearn计算Precision/Recall/F1遇报错求助

问题场景 #

用TFLearn训练完成文本分类模型并通过model.evaluate(test_x, test_y)完成评估后，尝试调用model.predict(test_y)获取预测结果以计算Precision、Recall、F1值，触发ValueError，提示输入形状与模型输入张量形状不兼容。

训练数据构建代码 #

#creating our training data
training = []
output = []
#create an empty array for our output 
output_empty = [0] * len(classes)

#training set, bag of words for each sentence
for doc in documents:
  #initialize our bag of words
  bag = []
  #list of tokenized words for the pattern
  pattern_words = doc[0]
  # stem each word
  pattern_words = [stemmer.stem(word.lower()) for word in pattern_words]
  # create our bag of words array
  for w in words:
    bag.append(1) if w in pattern_words else bag.append(0)

    # output is a '0' for each tag and '1' for current tag
    output_row = list(output_empty) 
    output_row[classes.index(doc[1])] = 1

    training.append([bag, output_row])

# shuffle our features and turn into np.array
random.shuffle(training)
training = np.array(training)

# create train and test lists
train_x = list(training[:,0])
train_y = list(training[:,1])

数据集划分代码 #

from sklearn.model_selection import train_test_split
train_x, test_x, train_y, test_y = train_test_split(train_x, train_y, test_size=0.10, random_state=10, stratify=train_y)

数组转换代码 #

train_x = np.array(train_x)
train_y = np.array(train_y)
test_x=np.array(test_x)
test_y = np.array(test_y)

模型构建与训练代码 #

net = tflearn.input_data(shape=[None, len(train_x[0])])
net = tflearn.fully_connected(net, 128)
net = tflearn.fully_connected(net, len(train_y[0]), activation='softmax')
net = tflearn.regression(net)
# Define model and setup tensorboard
model = tflearn.DNN(net, tensorboard_dir='tflearn_logs')
model.fit(train_x, train_y, n_epoch=10, batch_size=8, show_metric=True )

model.evaluate(test_x, test_y)

错误触发代码 #

predict_y = model.predict(test_y)

报错信息 #

---------------------------------------------------------------------------

ValueError                                Traceback (most recent call last)

<ipython-input-56-33f24cfe9a52> in <module>
----> 1 predicted_y = model.predict(test_y)
      2 #y_hatclass = model.predict_classes(test_x)

3 frames

/usr/local/lib/python3.8/dist-packages/tensorflow/python/client/session.py in _run(self, handle, fetches, feed_dict, options, run_metadata)
   1162           if (not is_tensor_handle_feed and
   1163               not subfeed_t.get_shape().is_compatible_with(np_val.shape)):
-> 1164             raise ValueError(
   1165                 f'Cannot feed value of shape {str(np_val.shape)} for Tensor '
   1166                 f'{subfeed_t.name}, which has shape '

ValueError: Cannot feed value of shape (143, 7) for Tensor InputData/X:0, which has shape (?, 55)

问题原因 #

• 模型输入层定义为shape=[None, len(train_x[0])]，即需要接收55维的文本特征向量
• 错误地将**测试集标签test_y（形状(143,7)的one-hot编码向量）**作为输入喂给模型，维度完全不匹配，触发形状不兼容错误

解决方案 #

1. 修正预测输入 #

将model.predict(test_y)改为model.predict(test_x)，传入测试集特征而非标签

2. 转换预测结果格式 #

模型输出的是softmax概率分布，需要转换为类别索引；同时将one-hot编码的test_y也转换为类别索引，才能用sklearn计算指标

完整修正代码 #

import numpy as np
from sklearn.metrics import precision_score, recall_score, f1_score

# 用测试集特征做预测
predict_y = model.predict(test_x)
# 将softmax概率转换为类别索引
predict_classes = np.argmax(predict_y, axis=1)
# 将one-hot编码的真实标签转换为类别索引
true_classes = np.argmax(test_y, axis=1)

# 计算评估指标（average参数可根据需求调整为'macro'/'micro'/'weighted'）
precision = precision_score(true_classes, predict_classes, average='weighted')
recall = recall_score(true_classes, predict_classes, average='weighted')
f1 = f1_score(true_classes, predict_classes, average='weighted')

print(f"Precision: {precision:.4f}")
print(f"Recall: {recall:.4f}")
print(f"F1 Score: {f1:.4f}")

内容的提问来源于stack exchange，提问作者Jul