SamyakJ2512/collaborative-filtering

Machine learning project on Collaborative Filtering (Netflix Prize competition format)

CIL 2022 Project Repository

Team: Runtime Terror

Overview

Content

The project contains the following methods, runnable as self-contained and independent Jupyter notebooks, for collaborative filtering:

Matrix Factorization

• The baseline solution - SVD + ALS on a normalized dataset (notebook).
• The baseline solution with a k-NN initialization approach (notebook).
• The baseline solution with a Gaussian processes approach (notebook).
• Improved SVD (notebook).
• The baseline solution with Improved SVD (notebook).

Neural Networks

• Neural Net for Collaborative Filtering (notebook).
• Sparse FC (notebook).
• Knowledge Graphs (notebook).

Factorization Machines

• ...

Setup

1. Start by creating a virtual environment. Using conda,

   conda create --name cil_runtime_terror python=3.6.13

2. Activate the virtual environment. Using conda,

   conda activate cil_runtime_terror

3. Go to the project root and install the dependencies.

   pip install -r requirements.txt

Links and Data

• Kaggle competition
• data_train.csv
• sampleSubmission.csv