Context-Aware Recommendation Systems in the IoT Environment (IoT-CARS)–A Comprehensive Overview

Abstract

An essential goal of recommendation systems is to provide users with accurate and personalized recommendations that meet their preferences. With the rapid growth of IoT-connected sensors, the availability of contextual information has increased, and this has necessitated the fast development of Context-Aware Recommendation Systems (CARS). Context-Aware recommenders are different from traditional recommenders because of their ability to predict the ratings of target users/items by exploiting the knowledge of contextual information. Context-aware recommenders define the context as any information that characterizes the situations of items and users at a particular interaction. They are essential for some contexts where prediction can be more precise in generating specific personalized recommendations. This paper provides a comprehensive review of context-based recommendation systems in IoT environments, namely IoT-CARS, and sheds light on their requirements, characteristics, and applications. We characterize context-aware recommenders in terms of the different IoT contexts and how these contexts are modeled. We also highlight the used metrics to evaluate the performance of various context-based recommenders.

Retracted: Graph‐based content matching for web of things through heuristic boost algorithm

“…In this study, nine fatal diseases are forecasted using seven classification techniques in ML. And four quantifiable metrics, such as precision and adaptability, curve area, and sensitivity were being used to examine the effectiveness of the suggested approach [17]. To unload the IoT-sensor apps tasks in a fog computing environment, a metaheuristic scheduler called Smart-Ant-Colony-Optimization offloading method is proposed in this paper.…”

Relevancy or Diversity?

“…CARS is a multidimensional (i.e., user, item, and context) system and hence brings a challenge to learn and predict users' preferences (Suhaim and Berri 2021). The existing studies on CARS mainly concentrate on methods and strategies to improve recommendation accuracy (Kulkarni and Rodd 2020, Morgan, Paun and Ntarmos 2020, Colomo-Palacios et al 2017, Nawara and Kashef 2021. In general, three CARS methods are mainly discussed in previous research: Interest-relevancy, contextrelevancy and content-diversity (Jain, Singh and Dhar 2020, Wu et al 2020, Werneck et al 2021.…”

“…Except for the demographic variables and information on the app use, they measured all constructs on a seven-point Likert scale from 1 as "strongly disagree" to 7 as "strongly agree". The degree of multi-context of the app use in this study refers to the number of contexts (or activities) in which a user is usually engaged when thsy read news feed in daily life (Suhaim and Berri 2021, Nawara and Kashef 2021, Carlarne 2011, Dey 2001. Therefore, in this study, the authors focused on the user's daily routines to study the degree of multi-context, which is consistent with the previous studies (Jin et al 2019, Porter et al 2010.…”

Deep Learning Model for Predicting Consumers’ Interests of IoT Recommendation System

“…Several research works have recognized the significance of recommendation systems for IoT environments and discussed different recommendation techniques including: i) contentbased, ii) collaborative filtering, iii) knowledge-based, and other techniques [16,17,18,19,20]. Other researchers conducted more specific survey papers on IoT recommendation systems related to the fields of context-aware [21], trust [22], and mobile health (m-health) [23]. Some research works used a content-based technique to recommend IoT devices and services.…”