Interest-based recommendations for business intelligence users. (December 2019)

• Record Type: Journal Article
• Authors: Drushku, Krista; Aligon, Julien; Labroche, Nicolas; Marcel, Patrick; Peralta, Verónika
• Is Part Of: Information systems. Volume 86(2019)
• Page Start: 79
• Page End: 93
• Publication Date: 2019-12
• Subjects: User interest -- Feature construction -- Clustering -- BI analyses -- Collaborative recommender systems; Database management -- Periodicals; Electronic data processing -- Periodicals; Bases de données -- Gestion -- Périodiques; Informatique -- Périodiques; Database management; Electronic data processing; Periodicals; 005.7
• DOI: 10.1016/j.is.2018.08.004
• Languages: English
• ISSNs: 0306-4379

Abstract

Abstract: It is quite common these days for experts, casual analysts, executives and data enthusiasts, to analyze large datasets through user-friendly interfaces on top of Business Intelligence (BI) systems. However, current BI systems do not adequately detect and characterize user interests, which may lead to tedious and unproductive interactions. In this paper, we propose a collaborative recommender system for BI interactions, specifically designed to take advantage of identified user interests. Such user interests are discovered by characterizing the intent of the interaction with the BI system. Building on user modeling for proactive search systems, we identify a set of features for an adequate description of intents, and a similarity measure for grouping intents into coherent clusters. On top of these automatically identified interests, we build a collaborative recommender system based on a Markov model that represents the probability for a user to switch from one interest to another. We validate our approach experimentally with an in-depth user study, where we analyze traces of BI navigation. Our results are two-fold. First, we show that our similarity measure outperforms a state-of-the-art query similarity measure and yields a very good precision with respect to expressed user interests. Second, we compare our recommender system to two state-of-the-art systems to demonstrate the benefit of relying on user interests. Highlights: A simple formal model of BI interactions. The learning of a similarity measure based on features characterizing BI user interests. An approach to automatically discover user interests based on our measure. A recommender system designed to take advantage of the discovered user interests. An extensive set of experiments for the tuning and validation of our approach through a user study.