Data mining‐based tag recommendation system: an overview. (9th April 2015)

• Record Type: Journal Article
• Authors: Vairavasundaram, Subramaniyaswamy; Varadharajan, Vijayakumar; Vairavasundaram, Indragandhi; Ravi, Logesh
• Is Part Of: Wiley interdisciplinary reviews. Volume 5:Number 3(2015)
• Page Start: 87
• Page End: 112
• Publication Date: 2015-04-09
• Subjects: Data mining -- Periodicals; 006.31205
• DOI: 10.1002/widm.1149
• Languages: English
• ISSNs: 1942-4787

Abstract

Abstract : The advent of high‐speed Internet connections has revolutionized the way research is being carried out to obtain relevant information. Conversely, retrieving pertinent information from the copious resources available is not only difficult but also time consuming. In the recent years, tagging activity has been perceived as a potential source of knowledge on personal preferences, interests, targets, goals, and other attributes. Tags allow users to effectively annotate resources using keywords to personalize their recommendations and organize the resources for easy retrieval. However, the preference of users varies extremely resulting in tagging being counterproductive. These shortcomings reduce the application of the tagging system for filtering as well as retrieval of information. The tag recommendation system becomes useful by suggesting a set of relevant keywords to annotate the resources. This paper presents a review of the tag recommendation systems and the constraints that affects the available tag recommendation systems. Furthermore, we propose the use of spreading activation algorithm to study the role of constructed topic ontology for efficient tag recommendations. This approach is founded on the assumption that tags that are recommended to the user are predicted from the extracted keywords from the existing blogs and the topics in constructed topic ontology. We have also proposed a tag classification system, namely Correlation‐based Feature Selection–Hybrid Genetic Algorithm and classifier HGA‐SVM (support vector machine), and have compared the results with results produced by other existing feature selection methods. The results obtained from the experiments have been presented. WIREs Data Mining Knowl Discov 2015, 5:87–112. doi: 10.1002/widm.1149 This article is categorized under: Algorithmic Development > Web Mining Technologies > Classification Technologies > Computational Intelligence