Netflix Recommender System Project

Posted on | In

Netflix Recommender System Project (Fall 2018)

The purpose of this project is to propose several viable models by introducing concepts, analyzing advantages and disadvantages, evaluating performance, and understanding the recommendation dynamics.

The project is the final project for NYU DS-GA 1001 Intro to Data Science

Data

Data comes from kaggle MovieLens 100K Dataset.
Also uploaded, simply download everything and run the Jupyter Notebooks.

Code

Source code
Packeges used: pandas, numpy, surprise, scikit-learn, etc.

  • data-preprocessing: some data cleaning and data visualization.

  • baseling-model: implemented several baseline models: only mean of the ratings, weighed mean, SVM, gradient boosting, etc.

  • surprise: implemented several recommender system models from packege surprise: SVD, SCD++, NMF, kNN, and performed stacking.

  • evaluation: tested model accuracy, coverage, diversity through several metrics.

Authors