Elite Evolutionary Discrete Particle Swarm Optimization for Recommendation Systems
Recommendation systems (RSs) play a vital role in e-commerce and content platforms, yet balancing efficiency and recommendation quality remains challenging. Traditional deep models are computationally expensive, while heuristic methods like particle swarm optimization struggle with discrete optimiza...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-04-01
|
| Series: | Mathematics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2227-7390/13/9/1398 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850155759039414272 |
|---|---|
| author | Shanxian Lin Yifei Yang Yuichi Nagata Haichuan Yang |
| author_facet | Shanxian Lin Yifei Yang Yuichi Nagata Haichuan Yang |
| author_sort | Shanxian Lin |
| collection | DOAJ |
| description | Recommendation systems (RSs) play a vital role in e-commerce and content platforms, yet balancing efficiency and recommendation quality remains challenging. Traditional deep models are computationally expensive, while heuristic methods like particle swarm optimization struggle with discrete optimization. To address these limitations, this paper proposes elite-evolution-based discrete particle swarm optimization (EEDPSO), a novel framework specifically designed to optimize high-dimensional combinatorial recommendation tasks. EEDPSO restructures the velocity and position update mechanisms to operate effectively in discrete spaces, integrating neighborhood search, elite evolution strategies, and roulette-wheel selection to balance exploration and exploitation. Experiments on the MovieLens and Amazon datasets show that EEDPSO outperforms five metaheuristic algorithms (GA, DE, SA, SCA, and PSO) in both recommendation quality and computational efficiency. For datasets below the million-level scale, EEDPSO also demonstrates superior performance compared to deep learning models like FairGo. The results establish EEDPSO as a robust optimization strategy for recommendation systems that effectively handles the cold-start problem. |
| format | Article |
| id | doaj-art-2afc069df9ce41d6827d3adf971c15d4 |
| institution | OA Journals |
| issn | 2227-7390 |
| language | English |
| publishDate | 2025-04-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Mathematics |
| spelling | doaj-art-2afc069df9ce41d6827d3adf971c15d42025-08-20T02:24:47ZengMDPI AGMathematics2227-73902025-04-01139139810.3390/math13091398Elite Evolutionary Discrete Particle Swarm Optimization for Recommendation SystemsShanxian Lin0Yifei Yang1Yuichi Nagata2Haichuan Yang3Graduate School of Technology, Industrial and Social Sciences, Tokushima University, Tokushima 770-8506, JapanFaculty of Science and Technology, Hirosaki University, Hirosaki-shi 036-8560, JapanGraduate School of Technology, Industrial and Social Sciences, Tokushima University, Tokushima 770-8506, JapanGraduate School of Technology, Industrial and Social Sciences, Tokushima University, Tokushima 770-8506, JapanRecommendation systems (RSs) play a vital role in e-commerce and content platforms, yet balancing efficiency and recommendation quality remains challenging. Traditional deep models are computationally expensive, while heuristic methods like particle swarm optimization struggle with discrete optimization. To address these limitations, this paper proposes elite-evolution-based discrete particle swarm optimization (EEDPSO), a novel framework specifically designed to optimize high-dimensional combinatorial recommendation tasks. EEDPSO restructures the velocity and position update mechanisms to operate effectively in discrete spaces, integrating neighborhood search, elite evolution strategies, and roulette-wheel selection to balance exploration and exploitation. Experiments on the MovieLens and Amazon datasets show that EEDPSO outperforms five metaheuristic algorithms (GA, DE, SA, SCA, and PSO) in both recommendation quality and computational efficiency. For datasets below the million-level scale, EEDPSO also demonstrates superior performance compared to deep learning models like FairGo. The results establish EEDPSO as a robust optimization strategy for recommendation systems that effectively handles the cold-start problem.https://www.mdpi.com/2227-7390/13/9/1398recommendation systemmetaheuristic algorithmparticle swarm optimization algorithmelite evolution strategyneighborhood search |
| spellingShingle | Shanxian Lin Yifei Yang Yuichi Nagata Haichuan Yang Elite Evolutionary Discrete Particle Swarm Optimization for Recommendation Systems Mathematics recommendation system metaheuristic algorithm particle swarm optimization algorithm elite evolution strategy neighborhood search |
| title | Elite Evolutionary Discrete Particle Swarm Optimization for Recommendation Systems |
| title_full | Elite Evolutionary Discrete Particle Swarm Optimization for Recommendation Systems |
| title_fullStr | Elite Evolutionary Discrete Particle Swarm Optimization for Recommendation Systems |
| title_full_unstemmed | Elite Evolutionary Discrete Particle Swarm Optimization for Recommendation Systems |
| title_short | Elite Evolutionary Discrete Particle Swarm Optimization for Recommendation Systems |
| title_sort | elite evolutionary discrete particle swarm optimization for recommendation systems |
| topic | recommendation system metaheuristic algorithm particle swarm optimization algorithm elite evolution strategy neighborhood search |
| url | https://www.mdpi.com/2227-7390/13/9/1398 |
| work_keys_str_mv | AT shanxianlin eliteevolutionarydiscreteparticleswarmoptimizationforrecommendationsystems AT yifeiyang eliteevolutionarydiscreteparticleswarmoptimizationforrecommendationsystems AT yuichinagata eliteevolutionarydiscreteparticleswarmoptimizationforrecommendationsystems AT haichuanyang eliteevolutionarydiscreteparticleswarmoptimizationforrecommendationsystems |