Scholarlink

<p>Dynamic software product lines (DSPLs) must adapt to evolving conditions under conflicting objectives, often incurring high adaptation costs and computational overhead. This paper proposes an ensemble ranking-based approach integrated with standard multi-objective evolutionary algorithms (MOEAs) to enhance decision stability and reduce adaptation overhead. Unlike conventional MOEA approaches that rely solely on fitness-based dominance, the approach introduces cross-objective ranking and ensemble-based selection to balance trade-offs more effectively. The adaptation process is formulated as a multi-objective optimization problem using a linear constraint representation of feature models. Evaluated against two widely used MOEA-based DSPL benchmarks, the proposed approach demonstrated significant reductions in both adaptation cost and frequency while preserving high-quality configurations. Execution overhead remained minimal at the millisecond level. Additionally, the computational efficiency of MOEAs within the proposed scheme was analyzed. Findings confirmed the stability, efficiency, and scalability of the approach, highlighting its suitability for DSPLs operating in resource-constrained and high-performance computing environments.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Ensemble rank-based multi-objective optimization for stable and resource-efficient adaptation in dynamic software product lines

  • 期刊论文

  • 发表日期：

    2026年2月21日

查看全文
  • Alireza Azimi,
  • Mehran Mohsenzadeh,
  • Ebrahim Akbari,
  • Reza Ravanmehr,
  • Mohammad Reza Alizadeh

摘要

Dynamic software product lines (DSPLs) must adapt to evolving conditions under conflicting objectives, often incurring high adaptation costs and computational overhead. This paper proposes an ensemble ranking-based approach integrated with standard multi-objective evolutionary algorithms (MOEAs) to enhance decision stability and reduce adaptation overhead. Unlike conventional MOEA approaches that rely solely on fitness-based dominance, the approach introduces cross-objective ranking and ensemble-based selection to balance trade-offs more effectively. The adaptation process is formulated as a multi-objective optimization problem using a linear constraint representation of feature models. Evaluated against two widely used MOEA-based DSPL benchmarks, the proposed approach demonstrated significant reductions in both adaptation cost and frequency while preserving high-quality configurations. Execution overhead remained minimal at the millisecond level. Additionally, the computational efficiency of MOEAs within the proposed scheme was analyzed. Findings confirmed the stability, efficiency, and scalability of the approach, highlighting its suitability for DSPLs operating in resource-constrained and high-performance computing environments.