Vol.:(0123456789) 1 3 Evolving Systems https://doi.org/10.1007/s12530-019-09296-3 ORIGINAL PAPER An aggregation approach to multi‑criteria recommender system using genetic programming Shweta Gupta 1  · Vibhor Kant 1 Received: 2 December 2018 / Accepted: 6 August 2019 © Springer-Verlag GmbH Germany, part of Springer Nature 2019 Abstract Recommender system is one of the emerging personalization tools in e-commerce domains for suggesting suitable items to users. Traditional collaborative fltering (CF) based recommender systems (RSs) suggest items to users based on the over- all ratings to fnd out similar users. Multi-criteria ratings are used to capture user preferences efciently in multi-criteria recommender systems (MCRSs), and incorporation of criteria ratings can lead to higher performance in MCRS. However, aggregation of these criteria ratings is a major concern in MCRS. In this paper, we propose a multi-criteria collaborative fltering-based RS by leveraging information derived from multi-criteria ratings through Genetic programming (GP). The proposed system consists of two parts: (1) weights of each user for every criterion are computed through our proposed modifed sub-tree crossover in GP process (2) criteria weights are then incorporated in CF process to generate efective recommendations in our proposed system. The obtained results present signifcant improvements in prediction and recom- mendation qualities in comparison to heuristic approaches. Keywords Collaborative fltering · Genetic programming · Multi-criteria ratings · Recommender system 1 Introduction The information overload problem occurs due to the increas- ing growth of Web data, which makes it difcult for users to discover relevant information, products or services accord- ing to their needs and preferences. Recommender system has contributed to a great deal to solve this overload problem by providing the relevant items to users based on their pref- erences. RS became an important research area in the last two decades (Adomavicius and Tuzhilin 2005; Jannach et al. 2012). Recommender systems have applied successfully for diferent types of e-commerce platforms such as Netfix and MovieLens (movie RS), Amazon.com (book RS), YouTube. com (video RS), Entree (Restaurant RS), TripAdvisor (Hotel recommendation), Last.fm (music RS) (Adomavicius and Kwon 2007; Adomavicius and Tuzhilin 2005) From mid-90’s, several fltering techniques have been explored to build an efective recommender system. The widely used techniques in the feld of recommender sys- tems are content-based fltering (CBF), collaborative flter- ing (CF) (Al-Shamri 2014; Wang et al. 2014) and hybrid fltering (Adomavicius and Tuzhilin 2005). In CF technique, items are recommended to users based on users who have similar preferences. While CBF technique Al-Shamri and Bharadwaj (2008), recommends items to users based on the similar contents which they have preferred in the past. Hybrid fltering technique, combine techniques of both CF and CBF technique to solve some problem inherent to each of them in isolation. A comprehensive survey was presented by Robin Burke (2002) on hybrid RS where various hybrid methods were proposed to build efective RS. Recommender system works on the two basic entities namely, user U and item I, in which user u ∈ U provides his/her preferences on experienced items. In RS, numerical rating R o represents the preference of the user for an item. The goal of recommender system is to predict the R o of an unknown item that can be defned as where, R o is calculated through overall ratings given by other users who have similar preferences to an active user i.e. to whom the item is to be recommended. In Angelov (1) R ∶ U × I → R o * Vibhor Kant vibhor.kant@gmail.com 1 Department of Computer Science and Engineering, The LNM Institute of Information Technology, Jaipur, Rajasthan 302031, India