Difference between revisions of "Top-N recommendation"
Jump to navigation
Jump to search
Zeno Gantner (talk | contribs) (Top-N recommendation is an evaluation protocol, not a measure. Back to Category: Evaluation) |
|||
| Line 1: | Line 1: | ||
'''Top-N recommendation''' is a variant of or an evaluation mode for the [[item prediction]] task. The specific task is to predict the best N items. Only those N items will be presented to the user or taken into account for computing the evaluation result. | '''Top-N recommendation''' is a variant of or an evaluation mode for the [[item prediction]] task. The specific task is to predict the best N items. Only those N items will be presented to the user or taken into account for computing the evaluation result. | ||
| − | Typical evaluation | + | Typical [[evaluation measure]]s for top-N recommendation are [[normalized discounted cumulative gain]] (NDCG) and [[precision]]. |
| − | [[Category:Task]] | + | [[Category: Task]] |
| − | [[Category: Evaluation | + | [[Category: Evaluation]] |
Latest revision as of 06:18, 17 January 2012
Top-N recommendation is a variant of or an evaluation mode for the item prediction task. The specific task is to predict the best N items. Only those N items will be presented to the user or taken into account for computing the evaluation result. Typical evaluation measures for top-N recommendation are normalized discounted cumulative gain (NDCG) and precision.
