{"id":22289489,"url":"https://github.com/trgiangdo/fast_recsys","last_synced_at":"2025-03-25T21:23:09.479Z","repository":{"id":40954560,"uuid":"284442973","full_name":"trgiangdo/fast_recsys","owner":"trgiangdo","description":"Accelerated Recommendation System on the rating prediction problem using Numba library.","archived":false,"fork":false,"pushed_at":"2023-02-11T00:32:28.000Z","size":176,"stargazers_count":0,"open_issues_count":8,"forks_count":0,"subscribers_count":1,"default_branch":"master","last_synced_at":"2025-01-30T18:43:52.467Z","etag":null,"topics":["knn","movielens","netflix-prize","numba","python","recommender-system","svd"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"gpl-3.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/trgiangdo.png","metadata":{"files":{"readme":"readme.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2020-08-02T10:54:38.000Z","updated_at":"2022-02-06T08:51:31.000Z","dependencies_parsed_at":"2025-01-30T18:38:56.269Z","dependency_job_id":"d5e6c72c-dc90-4d77-90da-8b702a891eb0","html_url":"https://github.com/trgiangdo/fast_recsys","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/trgiangdo%2Ffast_recsys","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/trgiangdo%2Ffast_recsys/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/trgiangdo%2Ffast_recsys/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/trgiangdo%2Ffast_recsys/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/trgiangdo","download_url":"https://codeload.github.com/trgiangdo/fast_recsys/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":245543969,"owners_count":20632750,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["knn","movielens","netflix-prize","numba","python","recommender-system","svd"],"created_at":"2024-12-03T17:09:16.352Z","updated_at":"2025-03-25T21:23:09.456Z","avatar_url":"https://github.com/trgiangdo.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Fast Recommender System on MovieLens 20M Dataset (working in progress)\nInspired by [gbolmler implementation of SVD using numba](https://github.com/gbolmier/funk-svd).\nThis repo contains reimplementation of kNN and common matrix factorization methods using [numba](https://github.com/numba/numba) library to accelarate `numpy` operations. Numba is a cool library and you need to give this a shot for future implementation using `numpy`.\n\n## MovieLens Dataset\nThe algorithms in this repo are tested on the Movielens [20M Dataset](https://www.kaggle.com/grouplens/movielens-20m-dataset).\n\nThis is a big dataset.\nIn order to extract the dataset to get a smaller dataset, first you need to download MovieLens 20M and save it on your computer, for example, to `movielens20M` folder.\nThen you need to create a folder `movilens-sample` for the new sampling dataset.\n\nOn `utils/sample_movielens.py` you can change the parameter to your like.\n\n```python\nif __name__ == \"__main__\":\n    sample_movielens(\n       \"movielens20M\",\n       \"movielens-sample\",\n       sample_size=1000\n    )\n```\n\nwhere `\"movielens20M\"` is the folder contains MovieLens 20M Dataset, `\"movielens-sample\"` is the folder contains new extracted dataset.\nSize of the extracted dataset can be changed via `sample_size`.\n\n## Netflix Prize Dataset\nThe algorithms in this repo are also tested on the [Netflix Prize dataset](https://www.kaggle.com/netflix-inc/netflix-prize-data).\nPublished by Netflix, the dataset contains a training set of 100 million ratings, which includes a probe set of 1 million ratings.\nHowever, the qualifying dataset has not been published anywhere (to my knowledge).\n\nFor that reason, the scipt in `utils/split_netflix_dataset.py` first uses the probe set as the validation set, then split the remaining ratings into training set and testing set.\nThe output contains 3 distinct files, `rating_train.csv`, `rating_test.csv`, `rating_val.csv` just like MovieLens 20M, and can be loaded into the algorithms using `utils/DataLoader`.\n\n\n## Benchmarks\n\nFolder `/examples` contains test runs on MovieLens dataset.\n\nCompare to [NicolasHug/Surprise](https://github.com/NicolasHug/Surprise), the runtime of kNNBaseline using Pearson similarity scores is much faster (817s compared to 3166s of Surprise on MovieLens 20M dataset).\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ftrgiangdo%2Ffast_recsys","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Ftrgiangdo%2Ffast_recsys","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ftrgiangdo%2Ffast_recsys/lists"}