# Algorithm evaluation
To be able to gauge how well your algorithm works, we have held out a total of 90,019 user-movie ratings. You are supposed to make predictions for these ratings; your predictions will be compared with the true user-movie ratings. To this end, we will measure the root mean squared error (RMSE) of your algorithm.

While developing your algorithms, you can measure their RMSE by writing your predictions to a CSV-file and uploading the resulting CSV-file to the Kaggle-in Class website. The website computes the RMSE of your algorithm on a random subset of the 90,019 held-out ratings, and puts your result on the public leaderboard. This allows you to compare your progress with that of other students.

The final score of your algorithm will be computed after the end of the competition based on the predicted ratings that were not used to compute the score on the leaderboard. This procedure is chosen because it ensures that final scores are not the result of overfitting.