Title: User interests profiling using fuzzy regression tree

Authors: Abd El Heq Silem; Hajer Taktak; Faouzi Moussa

Addresses: Laboratory LIPAH LR11ES14, Faculty of Sciences of Tunis, University of Tunis El Manar, 2092 Tunis, Tunisia ' Laboratory LIPAH LR11ES14, Faculty of Sciences of Tunis, University of Tunis El Manar, 2092 Tunis, Tunisia ' Laboratoire LCOMS, Université de Lorraine, France

Abstract: User modelling is an essential process for recommender systems. In user interests modelling, two kinds of approaches can be found. The first uses the text extracted from the user's browsing history to predict his interest degree, while the second, besides the extracted text, adds the user actions as input. Both groups predict an incorrect interest degree due to the use of text-only or an incorrect assessment of the user's actions. This paper proposes an architecture that employs a fuzzy regression tree algorithm to model the user's interests. The architecture detects the user's interests by evaluating his behaviour based on these factors: scrolling speed, time spent, location, number of visits, and clicked links. It also uses fuzzy logic to ensure the correct interpretation of each factor value according to the user's habits. Finally, the results show that the proposed architecture has the minimum RMSE (≈0.06) compared to the existing solutions.

Keywords: user modelling; human-computer interaction; ubiquitous computing; user profiling; recommender systems; context-awareness; machine learning; regression tree; fuzzy logic; user behaviour.

DOI: 10.1504/IJAHUC.2022.126114

International Journal of Ad Hoc and Ubiquitous Computing, 2022 Vol.41 No.3, pp.191 - 203

Received: 31 Mar 2021
Accepted: 08 Dec 2021
Published online: 11 Oct 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article