Rethinking Recommender Systems: Cluster-based Algorithm Selection

Andreas Lizenberger,Ferdinand Pfeifer,Bastian Polewka

2024-05-28

Abstract:Cluster-based algorithm selection deals with selecting recommendation algorithms on clusters of users to obtain performance gains. No studies have been attempted for many combinations of clustering approaches and recommendation algorithms. We want to show that clustering users prior to algorithm selection increases the performance of recommendation algorithms. Our study covers eight datasets, four clustering approaches, and eight recommendation algorithms. We select the best performing recommendation algorithm for each cluster. Our work shows that cluster-based algorithm selection is an effective technique for optimizing recommendation algorithm performance. For five out of eight datasets, we report an increase in nDCG@10 between 19.28% (0.032) and 360.38% (0.191) compared to algorithm selection without prior clustering.

Information Retrieval

What problem does this paper attempt to address?

This paper discusses the method of selecting clustering-based recommendation algorithms, aiming to optimize the performance of recommendation systems through user clustering. In the study, the authors pointed out that previous research mainly focused on the combination of a single clustering method and recommendation algorithm, while they wanted to demonstrate that algorithm selection before user clustering can improve the performance of recommendation algorithms. The paper covers 8 datasets, 4 clustering methods, and 8 recommendation algorithms, and selects the best recommendation algorithm for each cluster. Experimental results show that compared with algorithm selection without clustering, this method improves the nDCG@10 indicator on 5 datasets, with improvement ranging from 19.28% to 360.38%. The main contribution of the research is the proposal of a clustering-based recommendation algorithm selection strategy, which significantly improves the average performance of recommendation systems. The authors call this method intermediate layer algorithm selection because it operates on user groups. They use different clustering techniques, including k-Means (based on interaction count and item interaction vector) and graph-based clustering methods (Louvain and greedy modularity), and conduct experiments with the automated recommendation system tool LensKit-Auto. The paper also compares the combination effect of different clustering methods and recommendation algorithms, as well as related work with other techniques such as meta-learning, automated recommendation systems, and ensemble recommendation systems. In the end, the authors found that clustering-based recommendation algorithm selection can effectively improve recommendation performance and plan to further develop it as a long-term goal.

Rethinking Recommender Systems: Cluster-based Algorithm Selection

A Cluster-Based Incremental Recommendation Algorithm on Stream Processing Architecture

A Hybrid Recommendation Algorithm Based on Clustering and Collaborative Filtering

Human Factors Based Partitioning Versus Data Clustering for Recommendations.

Per-Instance Algorithm Selection for Recommender Systems via Instance Clustering

Review of Clustering-Based Recommender Systems

Incorporating Community Detection and Clustering Techniques into Collaborative Filtering Model

A review of systematic selection of clustering algorithms and their evaluation

Adaptive recommendation systems: A comparative analysis of KNN-based algorithms and hybrid models

Automatic Recommendation of a Distance Measure for Clustering Algorithms

User Interest Dynamics on Personalized Recommendation

A reinforcement learning recommender system using bi-clustering and Markov Decision Process

A Recommendation Algorithm Combining Clustering Method and Slope One Scheme

A careful assessment of recommendation algorithms related to dimension reduction techniques

A new fuzzy clustering-based recommendation method using grasshopper optimization algorithm and Map-Reduce

High-Performance Technique for Item Recommendation in Social Networks using Multiview Clustering

A collaborative filtering recommendation algorithm based on user clustering with preference types

Top-N Recommendation Algorithms: A Quest for the State-of-the-Art

Design of Garment Style Recommendation System Based on Interactive Genetic Algorithm

Application of k-means clustering algorithm to improve effectiveness of the results recommended by journal recommender system

ClusterExplorer: Enable User Control over Related Recommendations Via Collaborative Filtering and Clustering.