Classification Model Accuracy Using Classification Learner

Hello, I am a beginner in machine learning.

I train the data using a classification learner and I choose the SVM with the highest accuracy.

Blind testing the generated model with new data does not yield the expected performance.

In the confusion matrix, the classification accuracy of a specific class is 90%, but it cannot be predicted even by 40% in a blind test.

Even though the classification model created for model validation predicted the same data as the data used for training, the accuracy did not match the classification table of errors.

What's the problem? please answer about my question. thankyou

Answer 1

[Drew]

It sounds like the model is overfitting on the training data. Some techniques to avoid overfitting can be seen here: https://www.mathworks.com/campaigns/offers/common-machine-learning-challenges.html