Hi Atefeh,
Machine learning models usually require iteratively tuning and reinitialising parameters, along with preprocessing the dataset and trying out different models to improve the accuracy. Here are a few steps you could try for the same:
- Data Preprocessing: Ensure that the `all_feature_vectors` and `all_labels` are properly preprocessed, handling NaN values, scaling the data and any other modifications if required.
- Model Complexity: The default settings for "fitcecoc" might not be suitable for your data. Parameters such as the kernel function and regularization might need to be tuned to your specific dataset for optimal performance.
- Overfitting: If the training data is not diverse enough, or if the model is too complex, the SVM might overfit to the training data, resulting in poor generalization to new data. You can use cross-validation or regularization to try and avoid this.
- Label Encoding: Ensure that the labels ("all_labels") are correctly encoded. If there's any inconsistency in label encoding between training and prediction, it could lead to incorrect predictions.
- Class Imbalance: If the dataset is imbalanced and biased more towards one of the classes, the model might be similarly biased towards the majority class. This can be addressed by using class weights or resampling techniques.
I hope this helps!
