Conformal Group Recommender System

Venkateswara Rao Kagita,Anshuman Singh,Vikas Kumar,Pavan Kalyan Reddy Neerudu,Arun K Pujari,Rohit Kumar Bondugula

2023-07-22

Abstract:Group recommender systems (GRS) are critical in discovering relevant items from a near-infinite inventory based on group preferences rather than individual preferences, like recommending a movie, restaurant, or tourist destination to a group of individuals. The traditional models of group recommendation are designed to act like a black box with a strict focus on improving recommendation accuracy, and most often, they place the onus on the users to interpret recommendations. In recent years, the focus of Recommender Systems (RS) research has shifted away from merely improving recommendation accuracy towards value additions such as confidence and explanation. In this work, we propose a conformal prediction framework that provides a measure of confidence with prediction in conjunction with a group recommender system to augment the system-generated plain recommendations. In the context of group recommender systems, we propose various nonconformity measures that play a vital role in the efficiency of the conformal framework. We also show that defined nonconformity satisfies the exchangeability property. Experimental results demonstrate the effectiveness of the proposed approach over several benchmark datasets. Furthermore, our proposed approach also satisfies validity and efficiency properties.

Information Retrieval

What problem does this paper attempt to address?

### What problem does this paper attempt to solve?

This paper aims to address several key issues in Group Recommender Systems (GRS):

1. **Transparency and Interpretability**:
- Traditional group recommender systems focus on improving recommendation accuracy but often overlook the transparency and interpretability of the recommendation results. Users find it difficult to understand why the system recommends specific items.

2. **Confidence Measurement**:
- Existing group recommender systems usually lack confidence measurement for the recommendation results, i.e., how confident the system is about its recommendations. While confidence measurement has been widely studied in personalized recommender systems, it is less explored in group recommender systems.

3. **Balancing Individual Preferences and Overall Recommendations**:
- Group recommender systems need to consider the preferences of each member within the group and generate a recommendation list that satisfies the entire group. Balancing individual preferences with overall recommendations is a significant challenge.

To address these issues, the paper proposes a Conformal Group Recommender System (CGRS) based on the Conformal Prediction Framework. Specifically, this method defines Nonconformity Measures to provide confidence indicators for the recommendation results, making the recommendations not only accurate but also interpretable and reliable. Experimental results show that this method performs well on multiple benchmark datasets, effectively enhancing the performance of the recommender system.