Evaluation metrics

The evaluation metric is used by Netflix challenge is the root mean squared error (RMSE). The RMSE is calculated by the root of the averaging all squared errors between the true user ratings in the test set and the corresponding predicted values.

$$RMSE = \sqrt{\frac{{\sum {(r_i - \hat{r_i})^2} }}{n}}$$

where $r_i$ and $\hat{r_i}$ are the actual and predicted rating values respectively; and $n$ is the size of test set.
Another evaluation metric is widely used to compare the performance of different recommendation methods is MAE (Mean Absolute Error). The MAE is calculated by averaging all absolute errors between the true user ratings in the test set and the corresponding predicted values.

$$MAE = \frac{{\sum {\vert r_i - \hat{r_i}\vert} }}{n}$$

We use both of these metrics to evaluate the performance of our approach on the Netflix subset.