Difference between revisions of "Discounted cumulative gain"
Jump to navigation
Jump to search
Zeno Gantner (talk | contribs) |
Zeno Gantner (talk | contribs) |
||
| Line 2: | Line 2: | ||
For comparing different lists of recommendations with various lengths, '''normalized Discounted Cumulative Gain''' ('''nDCG'''/'''NDCG''') is used. | For comparing different lists of recommendations with various lengths, '''normalized Discounted Cumulative Gain''' ('''nDCG'''/'''NDCG''') is used. | ||
It is computed by dividing the DCG by the '''Ideal Discounted Cumulative Gain''' or '''IDCG'''. The higher the nDCG, the better ranked list. | It is computed by dividing the DCG by the '''Ideal Discounted Cumulative Gain''' or '''IDCG'''. The higher the nDCG, the better ranked list. | ||
| + | |||
| + | == Literature == | ||
| + | * Markus Weimer, Alexandros Karatzoglou, Quoc Viet Le, Alex Smola: ''[http://books.nips.cc/papers/files/nips20/NIPS2007_0612.pdf CofiRank - Maximum Margin Matrix Factorization for Collaborative Ranking]'', NIPS 2007 | ||
== External links == | == External links == | ||
* [[Wikipedia: Discounted cumulative gain]] | * [[Wikipedia: Discounted cumulative gain]] | ||
* [http://github.com/zenogantner/MyMediaLite/blob/master/src/MyMediaLite/Eval/ItemPredictionEval.cs C# implementation of DCG, IDCG, and nDCG] (part of the [[MyMediaLite]] library) | * [http://github.com/zenogantner/MyMediaLite/blob/master/src/MyMediaLite/Eval/ItemPredictionEval.cs C# implementation of DCG, IDCG, and nDCG] (part of the [[MyMediaLite]] library) | ||
| + | * [http://www.cofirank.org/ CofiRank] contains an implementation of an algorithm that optimizes NDCG. | ||
[[Category:Evaluation measure]] | [[Category:Evaluation measure]] | ||
Revision as of 10:09, 10 August 2011
Discounted Cumulative Gain (DCG) is a measure for ranking quality and measures the usefulness (gain) of an item based on its relevance and position in the provided list. For comparing different lists of recommendations with various lengths, normalized Discounted Cumulative Gain (nDCG/NDCG) is used. It is computed by dividing the DCG by the Ideal Discounted Cumulative Gain or IDCG. The higher the nDCG, the better ranked list.
Literature
- Markus Weimer, Alexandros Karatzoglou, Quoc Viet Le, Alex Smola: CofiRank - Maximum Margin Matrix Factorization for Collaborative Ranking, NIPS 2007
External links
- Wikipedia: Discounted cumulative gain
- C# implementation of DCG, IDCG, and nDCG (part of the MyMediaLite library)
- CofiRank contains an implementation of an algorithm that optimizes NDCG.
