Article Preview
Top1. Introduction
Social networking systems (SNS) (Chen et al., 2009) are usually communicate between the system who provide services and the user. Due to inaccessible information of systems, users may face the problems of risk, because they cannot trial products before purchase. In addition of these, SNS mainly suffer from problem of sparsity (Hu et al., 2017) in which the user’s rating data or other information is not sufficient for analyzing or predicting the user’s future choice. Scalability (Banda & Bharadwaj, 2012) is one of the problems occur in Collaborative Filtering (CF) (Mu & Zeng, 2018) when data grow in web content. It is difficult to manage data, and this problem plays a major role in many SNSs. Cold start problem (Duricic et al., 2018) occurs when a new user/item enters into websites, it is then difficult for recommending or predicting user’s behavior. To overcome these problems, tagging systems were introduced in Recommendation Systems (RS) (Adomavicius & Tuzhilin, 2005).
Tagging (Golder & Huberman, 2006) is a part of SNS, and this is broadly categorized as Taxonomies and Folksonomies (Lohmann & Diaz, 2012). Wherein taxonomies, only authorized person can add or edit the description of content. The only shortcoming of taxonomy is, content cannot be changed by users and thus it is not user-friendly. Examples are Wordpress and Blog categories. To overcome these limitations of taxonomies, folksonomies are introduced. Folk means people, and it is defined as user generated tags which are categorized by administrators. This folksonomy may suffer from problems like: polysemy-same tag with different meaning, synonyms-different tags with same meaning, plural and singular words, acronym and abbreviations and language barrier. Tagging is used in both taxonomies and folksonomies in SNS. In this area, we deal with major methods in Recommender Systems and their types, Collaborative Tagging (CT) (Banda & Bharadwaj, 2014b) and Tag cloud (Sinclair & Cardew-Hall, 2008).
Recommender system is recommending an item to user and this is divided into: (i) Content-based (Cantador et al., 2010), in which user’s past behavior is considered; (ii) Collaborative Filtering computes the similarities between users; and (iii) Hybrid Filtering (Zhang et al., 2018) is mixture of content-based and collaborative filtering.
Collaborative tagging provides tagging information in websites, so that prediction and recommendation done accurate. Tagging is divided into static tagging, in which only liking information is considered and dynamic tagging gives description of each item is also called as sentiment analysis (Pang & Lee, 2008).
A tag cloud generates popular tags in an image form. These tag clouds are maintained in several images such as rectangle, circle and it may be of any image. These tag clouds help users in finding information in web sites easily.
This paper develops a new RS method based on Gradual Decay approach (Banda & Bharadwaj, 2014a) for webpage recommendation using item-based, tag-based, genre-based, diffusion similarity (Banda & Bharadwaj, 2014a) and a genetic algorithm for predicting. The following section 2 will present literature survey so that the proposed work can be clearly demonstrated in section 3. The validation and conclusions are covered in the two last sections.