Predicting Movie Ratings by using Matrix Factorization:
- Machine Learning - ML.NET Library
Steps used by the Matrix Factorization Algorythm in the MLService:
- The LoadDataFromDb method reads the training data from a CSV file and returns an IDataView. The schema for the training data is defined by type MovieRating.
- The TransformData method encodes the two features userId and movieId. These encoded features will be provided as input to the MatrixFactorizationTrainer.
- The GenerateOptionsObject method specifies the options for the MatrixFactorizationTrainer.
- The GeneratePipeline method creates the training pipeline.
- The TrainAndReturnModel method creates the training pipeline by appending the data processing pipeline and MatrixFactorizationTrainer. It fits the model to the training dataset and returns the trained model.
- The EvaluateModelPerformance method loads the test dataset, performs the prediction, and evaluates the model's performance. The model evaluation metrics include RootMeanSquaredError.
- The GetMovieRatingPrediction method makes a single movie rating prediction for a specific user. It uses the trained model to predict the rating for a particular movie for a user.
recommendation-movies.csv (9743 movies)
| movieId | title | genres |
|---|---|---|
| 1 | Toy Story(1995) | Adventure Animation Children Comedy Fantasy |
| 2 | Jumanji (1995) | Adventure Children Fantasy |
| 3 | Grumpier Old Men (1995) | Comedy Romance |
recommendation-ratings-test.csv (20 records – just used as test values)
| userId | movieId | rating | timestamp |
|---|---|---|---|
| 1 | 1097 | 5 | 964981680 |
| 1 | 1127 | 4 | 964982513 |
| 1 | 1136 | 4 | 964981327 |
recommendation-ratings-train.csv (62369 records – used to train the model)
| userId | movieId | rating | timestamp |
|---|---|---|---|
| 1 | 1 | 4 | 964982703 |
| 2 | 3 | 3 | 964981247 |
| 3 | 4 | 5 | 964982224 |