Difference between revisions of "Recommender101"
Jump to navigation
Jump to search
Ls13cstudo (talk | contribs) |
|||
| Line 1: | Line 1: | ||
| − | '''Recommender101''' is a lightweight and easy-to-use [[framework]] written in Java to carry out [[offline experiments]] for [[Recommender Systems]]. It provides the user with various [[metrics]] and common [[evaluation]] strategies as well as some example recommenders and a dataset. The framework is easily extensible and allows | + | '''Recommender101''' is a lightweight and easy-to-use [[framework]] written in Java to carry out [[offline experiments]] for [[Recommender Systems]]. It provides the user with various [[metrics]] and common [[evaluation]] strategies as well as some example recommenders and a dataset. The framework is easily extensible and allows users to quickly implement their own recommenders and metrics. On the other hand, users who only want to test pre-implemented algorithms can instantly launch the software via Ant or Eclipse. |
| − | Implemented algorithms | + | Implemented '''recommender algorithms''' include among others |
| + | * [[KNN|nearest neighbors (kNN)]], | ||
| + | * [[SlopeOne]], | ||
| + | * [[matrix factorization]] methods, e.g., [[FunkSVD]], Koren's [[Asymmetric SVD]] and [[SVD++]], | ||
| + | * [[BPR|Bayesian Personalized Ranking]], | ||
| + | * [[Factorization Machines]], | ||
| + | * [[content-based filtering]]. | ||
| − | + | Recommender algorithms can be evaluated with the help of cross-validation and '''accuracy metrics''' including | |
| + | * [[Precision]], | ||
| + | * [[Recall]], | ||
| + | * [[NDCG]], | ||
| + | * [[MAE]], | ||
| + | * [[RMSE]], | ||
| + | * [[AUC]]. | ||
| + | |||
| + | Additional metrics can be used to measure recommendation '''biases''', e.g., | ||
| + | * aggregate [[diversity]], | ||
| + | * and the [[Gini index]]. | ||
== External links == | == External links == | ||
Latest revision as of 06:03, 23 November 2015
Recommender101 is a lightweight and easy-to-use framework written in Java to carry out offline experiments for Recommender Systems. It provides the user with various metrics and common evaluation strategies as well as some example recommenders and a dataset. The framework is easily extensible and allows users to quickly implement their own recommenders and metrics. On the other hand, users who only want to test pre-implemented algorithms can instantly launch the software via Ant or Eclipse.
Implemented recommender algorithms include among others
- nearest neighbors (kNN),
- SlopeOne,
- matrix factorization methods, e.g., FunkSVD, Koren's Asymmetric SVD and SVD++,
- Bayesian Personalized Ranking,
- Factorization Machines,
- content-based filtering.
Recommender algorithms can be evaluated with the help of cross-validation and accuracy metrics including
Additional metrics can be used to measure recommendation biases, e.g.,
- aggregate diversity,
- and the Gini index.
