How to test model developed using the Regression Learner App

I am working on some oil data trying to build a model for future predictions. My challenge has come with the testing of the chosen GPR model. Having split my original data into the train and test datasets, and trained, i find that my response when testing with the test dataset is poor. How can I handle this?

Answer 1

[Omega]

Hi Samuel,

I understand that you are encountering poor accuracy with the test dataset on your prediction with GPR model. You can try the strategies mentioned below which might help you improve the test accuracy.

• Check for Overfitting and Simplify the Model: If the model is too complex, it might be capturing noise instead of the underlying pattern. You an try reducing the number of features or simplifying the kernel function in GPR.
• Feature Engineering: You can try adding new features or removing less relevant ones as sometimes certain features don't contribute much to the prediction. You can also try transforming features (e.g., log, square root) to better capture relationships.
• Data Preprocessing: Ensure your data is scaled properly. GPR can be sensitive to the scale of the input. Also, try to remove or treat outliers in your dataset as they can skew results.
• Use Cross Validation: Use cross validation dataset to tune the Hyperparameters which will help in increasing the generalisation.

If you are still not able to get the desired accuracy, explore different models like SVMs, Neureal networks.