Optimized Test Data Generation for Path Testing Using Improved Combined Fitness Function with Modified Particle Swarm Optimization Algorithm

Software testing is essential for assuring the reliability and excellence of software systems. Nevertheless, already used optimization techniques, such Particle Swarm Optimization (PSO), sometimes get stuck in local optima during testing. This study suggests innovative improvements to the PSO algorithm to address and overcome this constraint. Initially, we propose a method in which every particle keeps track of a collection of superior particles and chooses a global best (gbest) at random. This approach helps to explore a wider range of solutions and reduces the likelihood of being stuck in local minima. Furthermore, we use an enhanced crowding method to specifically tackle the discrepancy between the exploration and exploitation stages. This approach prioritizes extensive exploration and exploitation during the early phases of the search, progressively shifting towards a strategy that focuses more on exploitation as the algorithm advances. We present a thorough explanation of these changes, specifically highlighting the modifications made to the pbest section and the use of a novel fitness function that enhances the search process in the given space. The method that we offer has the potential to improve software testing methods by optimizing PSO-based techniques, leading to better performance and efficiency. The experimental findings have shown that our method outperforms numerous existing evolutionary or meta-heuristic algorithms in terms of test data generation speed and achieves superior coverage with fewer evaluations. The algorithms being compared are the Adaptive Genetic Algorithm (AGA), Dandelion Optimizer (DO), Chaotic Flower-Fruit Fly Optimization Algorithm (CFFFOA), Imperialist Competitive Algorithm (ICA), Chaos Adaptive Particle Swarm Optimization Algorithm (CAPSO), Particle Swarm Optimization Algorithm with Empirical Balance Strategy (PSOEBS), and Teaching Learning-Based Optimization (TLBO).