Designing A New Genetic-Fuzzy Type ۲ Approach to Evaluate Self-Adaptive Systems by Software Quality Indicators

چکیده مقاله

This paper introduces a novel genetic-fuzzy type-2 approach for evaluating the quality of self-adaptive systems using software quality indicators. Self-adaptive systems, which adjust their behavior based on environmental and internal changes, present unique challenges for quality assessment, as their quality indicators are often uncertain, non-mathematical, and linguistically expressed by experts. To address these challenges, the authors propose a hybrid framework that combines interval type-2 fuzzy logic with a genetic algorithm. Type-2 fuzzy logic is employed to model the inherent uncertainty and linguistic nature of quality indicators, while the genetic algorithm is used to optimally tune the fuzzy weights and membership function parameters for each indicator.

A key innovation of the proposed method is its dual-dimensional evaluation, which considers both the conventional software quality aspects and the specific self-adaptive properties of the system. This holistic approach allows for a more comprehensive assessment compared to existing models, which typically focus on only one dimension. The framework operates through seven main stages, starting with the identification of relevant quality indicators and metrics, followed by the fuzzy ranking of indicators and metrics, assignment of fuzzy weights, and finally, defuzzification to produce a quantitative quality score. The genetic algorithm optimizes a chromosome encoding the parameters of the fuzzy membership functions, with the fitness function minimizing the error between the system's estimated quality and expert-provided ground truth values.

The proposed approach was validated using a real-world case study: the InSync adaptive traffic control system (ATCS). Through questionnaires, 32 quality indicators and 47 underlying metrics were identified and evaluated. The results demonstrated the method's effectiveness, flexibility, and comprehensiveness. Notably, the approach eliminates the need for complex scenario generation, can be extended to all software quality parameters, and is simpler to apply than many existing evaluation models. Comparative analysis with other quality models and recent research showed that the proposed method covers a wider range of quality indicators and better accommodates the uncertain, multi-faceted nature of self-adaptive systems.

Overall, this research makes a significant contribution by providing a structured, adaptive, and uncertainty-aware framework for quality evaluation in self-adaptive systems. The integration of type-2 fuzzy logic with evolutionary optimization offers a robust solution to a long-standing challenge in the field, paving the way for more reliable and insightful quality assessments in complex, dynamic software environments.