No abstract

No abstract

Abstract. Discounted Cumulative Gain (DCG) is a well-known ranking evaluation measure for models built with multiple relevance graded data. By handling tagging data used in recommendation systems as an ordinal relevance set of , , , we propose to build a DCG based recommendation model. We present an efficient and novel learning-to-rank method by optimizing DCG for a recommendation model using the tagging data interpretation scheme. Evaluating the proposed method on real-world datasets, we demonstrate that the method is scalable and outperforms the benchmarking methods by generating a quality top-item recommendation list. Keywords:Tagging data · Tag-based item recommendation · Discounted cumulative gain · Top-recommendation IntroductionIn a tag-based recommendation system, users annotate items of their interest with freely-defined tags, hence the ternary relation of , , is naturally formed. This system must interpret the observed and non-observed entries in tagging data efficiently in order to generate quality recommendations [1]. The observed, or positive, entries reveal the user interest by indicating that the user has annotated an item using certain tags. The non-observed entries can reveal two types of information: (1) negative entries that indicate users are not interested with the items; or (2) null values that indicate users might be interested in them in the future and they need to be predicted [2]. Accordingly, tagging data can be labelled using the ordinal relevance set of , , for a tuple of , , . The task of a tag-based recommendation system is to generate the list of items, which may be of interest to a user, by learning the users past tagging behavior. The list of recommended items is ordered in descending order based on the predicted preference score. Users usually show more interest to the fewer items at the top of the list than the ones further down the list [3]. The order of items in the recommendation list is crucial and, therefore, the recommendation task can be considered as a ranking problem in which the item preference score needs to be determined and sorted for

No abstract

No abstract