/Recommendation-Systems

Primary LanguageJupyter Notebook

Project Recommendation Systems

Part 1

In this part of the project, I tried using the data from Netflix to recommend to the user. Tried with CollabFilterOneVectorPerItem to experiment with Collaborative Filtering.

Part 2

Note

Later, I experiment with using SVD algorithm to extract embeddings as feature representation. Along with other features from movie and user information, I experimented with 2 different classification models to decide whether to recommend the movie to the user or not.

The models tested are:

  • LogisticRegression
  • GradientBoostingClassifier
  • XGBoostClassifier

Result

Achieved 0.84 held-out error during training, 0.77 mae and 0.61 balanced accuracy on the test set