Article Preview
Top1. Introduction
Today the Web is emerging exponentially and currently this trend is more than just a hypertext. The Web started to gain essentials of intelligence. In this context, the social web is a collection of social relationships linking people through the web. In fact, the social dimension of Web 2.0 emphasizes the communication between the Internet users with similar interests. Such interaction may be concertized through several online activities (Alfimtsev et al., 2012) such as social network (Cruz-Cunha et al., 2012), education, shopping…etc. Particularly, a social network defines the social organization between actors associated through different relationships ranging from informal contact to familiar relationships. Indeed, distinguished examples of social networks are LinkedIn, Tweeter, Facebook, Viadeo, etc.
Several works were devoted to the analysis of interest topics in social network. However, such operation remains effortful and generally does not provide satisfying results due to the large amount of managed information in this context.
Actually, the data warehouse domain delivers numerous solutions for modeling and dealing with huge amount of data. In fact, according to the interest topic, such a social structure may be modeled using multidimensional modeling and the managed information in social network context may be described using diverse dimensions and aggregated through several hierarchies using data warehouse solution.
Indeed, a data warehouse is a set of technologies aimed at enabling the analyst to manage large volumes of data extracted from production information systems in order to assist him in his decision process (Inmon, 2002). The user can interactively explore the multidimensional data by means of On-Line Analytical Processing (OLAP) paradigm (Kimball, 1997) through formulating complex MDX1 queries. These launched queries stored in OLAP log files, contain a set of useful knowledge about the analyst preferences. Indeed, a multidimensional preference is closely related to an instance of the data warehouse schema. Based on extracted preferences, several approaches of user-centric OLAP approaches are proposed: (i) OLAP recommender system involves an application predicting the user responses to options in multidimensional context (Jerbi et al, 2009),(Giacometti et al, 2009, 2011), (Khemiri et al., 2012), (Bimonte et al., 2014), (Ben Ahmed et al, 2015) and (Aligon et al., 2015); (ii) OLAP personalization tailors the presentation of the data cube to match the user’s preferences (Ravat et al., 2008), (Garrigós et al., 2009), (Jerbi et al., 2010), (Aligon et al., 2011) and (Golfarelli et al., 2011).
Generally, the personalization is known as a mechanism providing an overall customized, individualized user experience by taking into account the needs, the preferences and the characteristics of the user (Holland et al., 2003).
To the best of our knowledge, few works address the sparsity problem in web-based OLAP analysis when the MDX query generates an empty result. In this context, we distinguish two main trends of preferences: (i) frequent preference related to high frequency multidimensional component and (ii) rare or conflicting preference which has low frequency. In this context, we stress on the rare category of preferences.
In this paper, we focus on the data warehouse as a tool for representation and investigation of multidimensional social networking data. Thus, we introduce a new personalization approach dedicated to discover conflicting preferences from web-based OLAP log files. Such derived patterns will be used in the query reformulation to avoid the sparse results.
To motivate our contribution, we use, throughout this paper, an example of web-based recruitment data warehouse that we built in order to evaluate our contribution. Indeed, the web-based recruitment is the practice of human resources recruitment using electronic technologies. It renovates the recruitment landscape for both employers and job seekers. On the one hand, the companies moved their recruitment process online in order to find speedily relevant applicants. On the other hand, the job seekers may apply swiftly for job offerings. Our designed data warehouse aims to boost the web-based recruiting productivity. Its key purpose is to examine the job offerings and to enhance job opportunities for applicants.