proqwest/collaborative-filtering-recommendation-engine

Survey of collaborative filtering techniques and their effectiveness as a movie recommendation system

Collaborative Filtering Recommendation Engine

This project is a survey of neighborhood-based collaborative filtering techniques and their effectiveness in a movie recommendation system. Similarity metrics we are surveying include:

• Pearson Correlation Coefficient
• Spearman Rank Correlation Coefficient
• Mean-Squared Distance
• Cosine Similarity

For a detailed understanding of the study, see the writeup attached to this repo.

Things To Do

• Establish Data Pipeline
• Setup prediction/testing environment
• Write evaluation script
• Write/test Pearson Correlation Coefficient
• Write/test Spearman Rank Correlation Coefficient
• Write/test Mean-Squared Distance
• Write/test Cosine Similarity
• Write final analysis

Running the Code

All of these commands should be issued from the main directory of the repository.

Setup

pip install -r requirements.txt

Training

python Code/runner.py --mode train --algorithm insert_algorithm_here --model-file algorithm's_name.model --data Data/ratings.csv

Predicting

python Code/runner.py --mode test --algorithm insert_algorithm_here --model-file algorithm's_name.model --num-neighbors default_is_five --data Data/ratings.csv --predictions-file insert_algorithm_here.predictions

Evaluation

python Code/eval.py Data/ratings.csv algorithm's_name.predictions

Running Testing Suite

python test.py

Algorithm Names (for command line arguments)

• pearson
• spearman
• mean_squared
• cosine