MCMARS: Hybrid Multi-criteria Decision-Making Algorithm for Recommender Systems of Mobile Applications

Abstract

A mobile application (app) recommender system needs to support both developers and users. Existing recommender systems in the literature are based on single-criterion analysis, which is insufficient for producing better recommendations. Moreover, recommendations do not reflect the user’s perspectives. To address these issues, in this paper, we present a Multi-Criteria Mobile App Recommender System (MCMARS) that assists developers in improving their apps and recommends the top-performing apps to users. We define the performance score of an app based on four criteria attributes: risk assessment score, functionality score, user rating, and the app’s memory size. We define the risk assessment score for each app using multi-perspective analysis and the functionality score by assigning preference weights to the services of apps in the same category. We evaluate optimal weights of the criteria by integrating the entropy method and the extended Best-Worst method (BWM) using Hesitant-Triangular-Fuzzy information with group-decisions. Finally, the TOPSIS uses these weights to assess the app’s performance. To validate our MCMARS, we prepared a dataset of 124 government-approved COVID-19 Android apps from 80 countries and made it available on GitHub for the research community. Finally, we perform a fine-grained analysis of the app’s performance based on the criteria attributes that help the developers to improve their apps. The experimental results show that two independent attributes, “risk assessment score” and “functionality score”, significantly measure the app’s performance. According to our findings, only 12.5% of the apps in the experimental dataset provide high-performance, high-functionality, and low-risk.