How kunalgehlot73/hybrid-movie-recommendation-system works
This repository is a MovieLens implicit-feedback recommender: ratings are prepared into user–movie interactions, a VAE learns collaborative relevance, and genre similarity plus popularity rerank candidates into Top-10 results. The web API also supports selecting five catalog movies for recommendations and enriches movie presentation through TMDB. Evaluation, ItemKNN/BPR-MF baselines, and recommendation analysis are separate product capabilities. Source excerpts support the main recommender and web links; some offline workflow links are documented by the README rather than sampled source.
Main entry points: recommender.py Hybrid recommender · preprocess.py Prepare interactions · vae.py Variational autoencoder · training_utils.py Training utilities
Components 28
Data preparation
- Prepare interactions
preprocess.py - MovieLens loading
utils/data_loader.py - ID encoding
utils/encoder.py - Interaction matrix
utils/interaction.py - Per-user split
utils/splitter.py - Train/test matrices
VAE modeling
- Train VAE
train.py - Training utilities
utils/training_utils.py - KL annealing
callbacks/kl_annealing.py - Variational autoencoder
models/vae.py - VAE encoder
models/encoder.py - VAE decoder
models/decoder.py - Latent sampling
models/sampling.py
Recommendation
- Hybrid recommender
utils/recommender.py - Genre profiling
utils/genre_utils.py - Recommendation workflow
recommend.py
Web experience
- Recommendation API
web/backend/main.py - Recommendation service
web/backend/recommender_service.py - Movie enrichment
web/backend/tmdb_service.py
Evaluation and analysis
- Ranking metrics
utils/evaluation_metrics.py - Evaluate recommendations
evaluate.py - ItemKNN baseline
baselines/item_knn.py - BPR-MF baseline
baselines/bpr_mf.py - Dataset exploration
eda.py - EDA plots
utils/eda_plots.py - Recommendation analysis
analyze_recommendations.py
Other components
- Recommendation user
- TMDB
Connections 27
- Prepare interactions → MovieLens loading: loads dataset
preprocess.py - Prepare interactions → ID encoding: encodes IDs
preprocess.py - Prepare interactions → Interaction matrix: builds interactions
preprocess.py - Prepare interactions → Per-user split: splits interactions
preprocess.py - Prepare interactions → Train/test matrices: writes matrices
README.md - Train VAE → Train/test matrices: reads training data
README.md - Train VAE → Training utilities: uses training helpers
train.py - Training utilities → KL annealing: registers callback
utils/training_utils.py - Training utilities → Variational autoencoder: trains model
- Variational autoencoder → VAE encoder: builds encoder
models/vae.py - Variational autoencoder → VAE decoder: builds decoder
models/vae.py - VAE encoder → Latent sampling: samples latent
models/encoder.py - Recommendation user → Recommendation API: requests recommendations
- Recommendation API → Recommendation service: routes requests
web/backend/main.py - Recommendation service → Hybrid recommender: generates scores
web/backend/recommender_service.py - Hybrid recommender → Variational autoencoder: predicts relevance
utils/recommender.py - Hybrid recommender → Genre profiling: uses genre features
utils/recommender.py - Recommendation service → Movie enrichment: requests enrichment
web/backend/recommender_service.py - Movie enrichment → TMDB: fetches movie data
web/backend/tmdb_service.py - Recommendation workflow → Hybrid recommender: uses recommender
recommend.py - Evaluate recommendations → Hybrid recommender: evaluates rankings
evaluate.py - Evaluate recommendations → Ranking metrics: computes metrics
evaluate.py - ItemKNN baseline → Train/test matrices: uses shared data
README.md - BPR-MF baseline → Train/test matrices: uses shared data
baselines/bpr_mf.py - Dataset exploration → MovieLens loading: loads dataset
eda.py - Dataset exploration → EDA plots: creates plots
eda.py - Recommendation analysis → Hybrid recommender: analyzes output
analyze_recommendations.py