Improving the prediction accuracy of recommendation algorithms: Approaches anchored on human factors

TLDR

Recommender systems predict user interest from past ratings or item features, yet factors such as personality and lifestyle—potential determinants of behavior—are rarely incorporated. The paper aims to demonstrate that integrating lifestyle information can improve recommendation accuracy by mitigating limited data availability. The authors propose two methods: a lifestyle‑only model and a lifestyle‑augmented nearest‑neighbor model. Both methods outperform existing nearest‑neighbor approaches in most cases when evaluated on personalized television advertisement recommendations.

Abstract

Recommender systems are a special class of personalized systems that aim at predicting a user’s interest on available products and services by relying on previously rated items or item features. Human factors associated with a user’s personality or lifestyle, although potential determinants of user behavior are rarely considered in the personalization process. In this paper, we demonstrate how the concept of lifestyle can be incorporated in the recommendation process to improve the prediction accuracy by efficiently managing the problem of limited data availability. We propose two approaches: one relying on lifestyle alone and another integrating lifestyle within the nearest neighbor approach. Both approaches are empirically tested in the domain of recommendations for personalized television advertisements and are shown to outperform existing nearest neighborhood approaches in most cases.