I agree with Dr. Samundra Gurung

There are different ways to improve the model accuracy:

1) The most important is to get more data.

2) Do not forget to remove the outliers, missing data, and normalize the input (in some cases).

3) You can use principal component analysis also to reduce the dimension of your data.

4) You can tune the hyperparameter if you are using learning algorithms such as support vector machine, decision trees, random forest, neural networks and so on.

Thanks a lot Mr. Samundra Gurung for your guidance.

You are right, in this case, I have few data (approximately 110 inputs) and also I did not normalize the input data.

I will try to improve the accuracy of my model with your important points.

Thankful.

I totally agree with Dr. Samundra Gurung

Thanks Mr. Ahmed Elbeltagi

I am agreeing with Dr. Samundra Gurung . Besides your may try to add some more features to see what is the accuracy is, also add divide the data to train, cross validation and test to train the model more and to reduce the error in the test stage.

Thank you Mr. @Ali Mokhtar for your guidance.

I agree with Dr. Samundra Gurung

Thank you Mr. @Abdelghani Rouini.

You should divide your dataset into three subsets including training, validation, and testing datasets. then use training and validation set to obtain parameters and hyper-parameters. The most suitable model appears when the validation accuracy is on its peak. Using normalization, early stopping, gathering more data point, and removing outliers are some ways to improve your model performance.

Thank you very much dear @Amin Bemani for your valuable guidance.

1) You write that you have great accuracy for training data, but not for test data. The model is based on the training data, so it is obvious that the results are good. The quality of the model, however, is demonstrated by the accuracy for the test data.

2) You write that you have about 110 inputs of data. It is not enough. It is important how you divide the data into those for training and those for testing. Maybe use cross-validation as part of building your model.

3) And why Matlab?

4) Your question is quite general, so the answers will also be quite general :)