'Stacking multiclass classification accuracy is lower than same model for binary classification problem?
I have developed a stacked classification model which consists of Random forest, CNN, and logistic regression as a meta-model. Its accuracy is 99% for binary classification problem and its accuracy on multiclass classification is very low.
Epoch 1/50
11527/11527 - 172s - loss: 0.4597 - accuracy: 0.0560
Epoch 2/50
11527/11527 - 153s - loss: 0.2507 - accuracy: 0.0486
Epoch 3/50
11527/11527 - 125s - loss: 0.1830 - accuracy: 0.0442
Epoch 4/50
11527/11527 - 118s - loss: 0.1447 - accuracy: 0.0442
Epoch 5/50
11527/11527 - 96s - loss: 0.1220 - accuracy: 0.0440
I am using Keras APIs kerasClassifer and stackingClassifier and also used softmax activation for output in the CNN model. What is the problem with model?
Sources
This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.
Source: Stack Overflow
| Solution | Source |
|---|