Regression models - Evaluate and improve

To improve a model it’s good to know how evaluate it.

  1. So first evaluate your model.

  2. Then find out how to improve your model.

Evaluate regression models

  • Loss
    What’s good? You want the loss as low as possible. You want the training error to be slightly lower than test error.
    There are different error measures that you can use as loss function for a regression problem. The most common are the Mean squared error. (MSE), the Root Mean Squared Error (RMSE) and the Mean Absolute Error (MAE).

  • Error
    What’s good? A low error is good.

  • Mean Absolute Percentage Error
    What’s good? A low percentage error is good.
    The MAPE measures the mean absolute percentage error in percent.

  • R2/R-squared
    What’s good? The closer the R-squared value is to 1, the better predictions. The closer the value of r-squared is to 0, the bigger the difference between actual value and predicted value.
    The R-squared is a measurement of how well the model fits the data.

Improve regression models

  • High training and validation error
    Your model is underfitting
    Solution:

    • Run your model longer, that is,. increase the number of epochs in the Modeling view.,

    • Collect more data.

    • Try another model.

  • High discrepancy between training and validation error
    Your model is overfitting.
    Solution:

    • Try collecting more data.

    • Make sure that the data in your training and validation set are similar.

Was this page helpful?
YesNo