An Overview of Recommender Systems and Machine Learning in Feature Modeling and Configuration

Abstract

Recommender systems support decisions in various domains ranging from simple items such as books and movies to more complex items such as financial services, telecommunication equipment, and software systems. In this context, recommendations are determined, for example, on the basis of analyzing the preferences of similar users. In contrast to simple items which can be enumerated in an item catalog, complex items have to be represented on the basis of variability models (e.g., feature models) since a complete enumeration of all possible configurations is infeasible and would trigger significant performance issues. In this paper, we give an overview of a potential new line of research which is related to the application of recommender systems and machine learning techniques in feature modeling and configuration. In this context, we give examples of the application of recommender systems and machine learning and discuss future research issues.

Knowledge-based recommender systems: overview and research directions

“…Intensional representations of solution spaces (itemsets) as presented in Table 5 can be implemented with different knowledge representations ranging from constraint satisfaction problems (CSPs) (Rossi et al, 2006 ; Felfernig and Burke, 2008 ) and Boolean satisfiability problems (SAT problems) (Biere et al, 2021 ; Felfernig et al, 2021 ) to less frequently used recommendation knowledge representations such as answer set programming (ASP) (Eiter et al, 2009 ; Teppan and Zanker, 2020 ) and ontology-based knowledge representations , for example, description logics (DL) (Lee et al, 2006 ; McGuinness, 2007 ). In addition, database queries can be applied in such a way that intensionally formulated recommendation knowledge is encoded directly in database queries—see, for example, Felfernig et al ( 2006 , 2023a ).…”

“…KBR systems support the determination of recommendations specifically in complex and high-involvement item domains [domains where suboptimal decisions can have significant negative consequences, for example, when investing in high-risk financial services (Felfernig et al, 2006 )] where items are not bought on a regular basis (Aggarwal, 2016 ). Example item domains are financial services (Felfernig et al, 2007 ; Musto et al, 2015 ), software services (Felfernig et al, 2021 ), apartment or house purchasing (Fano and Kurth, 2003 ), and digital cameras (Felfernig et al, 2006 ). These systems are able to take into account constraints (e.g., high-risk financial services must not be recommended to users with a low preparedness to take risks) and provide explanations of recommendations also in situations where no solution could be identified.…”

Resource Optimisation in Cloud Computing: Comparative Study of Algorithms Applied to Recommendations in a Big Data Analysis Architecture

“…Hybrid systems combine two or more techniques to achieve better performance, such as content-based filtering and collaborative filtering. The limitations of one technique can be overcome by another technique Today, recommender systems have been studied in several areas such as smart cities [14], education [15], e-commerce [16], e-learning [17].…”

An overview of machine learning techniques in constraint solving

“…Highly-variant products, services, and systems allow for a potentially huge number of possible configurations (Felfernig et al, 2014;Felfernig et al, 2021). The corresponding configuration space could also allow faulty configurations simply due to the fact that some relevant constraints have not been integrated into the configuration model.…”

“…For an overview of the application of machine learning and recommender systems techniques in configuration scenarios (especially in feature-modeling related scenarios), we refer to Felfernig et al (2021). The example depicted in Table 3 sketches a basic approach for value prediction in configuration scenarios.…”