Overfitting happens when the model is too complex relative to the amount and noisiness of the training data. The possible solutions are: To simplify the model by selecting one with fewer parameters (e.g., a linear model rather than a high-degree polynomial model), by reducing the number of attributes in the training data or by constraining the model To gather more training data To reduce the noise in the training data (e.g., fix data errors and remove outliers)
