the accuracy of the training set is about 79%. How can I improve the accuracy of the model?

We'd like to know how to improve the accuracy of the *Classification learner* (CL) results in Matlab R2021a. We are attempting to use (CL) to categorise Arabic speech errors into five classes, and the highest accuracy we have achieved is 79%. Including data with a size of 20000*1100 & a features vector with a size of 1100*33 Is there anything else that can be considered in order to maximize the resultant accuracy ? And is it possible to suggest a better tool to use instead of (CL) to complete this task ? Thanks a lot ,,,

Answer 1

[Udit06]

Hi,

To boost classification accuracy, you can refine your feature set using PCA or LDA. Normalize your data and reduce noise for cleaner inputs. Experiment with different algorithms such as SVM, k-NN, or decision trees, and fine-tune them using cross-validation. If class imbalance is an issue, try balancing methods like oversampling or undersampling.

You can also consider using a deep learning approach like LSTM which are suitable for speech and sequence data. You can refer to the following documentations for the same:

• https://www.mathworks.com/help/deeplearning/ug/long-short-term-memory-networks.html
• https://www.mathworks.com/discovery/rnn.html