An App-Based Recommender System Based on Contrasting Automobiles

Abstract

Product recommendation systems are essential for enhancing customer experience, and integrating them with mobile apps is crucial for improving usability and fostering user engagement. This study proposes a hybrid approach that utilizes comparative facts from pairwise comparison data and comparison lists, with association rules as the method to formulate the recommendation system. The study employs a dataset from the New-Cars Database app, comprising 30,867 vehicle comparisons made by 5327 users across 40 car brands and 870 cars from 30 January 2015 to 2 April 2015. Two metrics are developed to measure the system’s output under varying support and confidence thresholds. The findings suggest that adjusting the support and confidence values can improve the breadth and depth of product recommendations. In addition, the unit of analysis can affect the recommendation system’s output, with comparison lists supplementing and expanding the exploration of potential outcomes. The proposed hybrid approach aims to provide more reliable and comprehensive product recommendations by combining both approaches and has implications for both academic and managerial contexts by facilitating the development of effective recommendation systems.