A fuzzy adaptive teaching learning-based optimization strategy for generating mixed strength t-way test suites
The use of meta-heuristic algorithms as the basis for t-way (where t indicates the interaction strength) and mixed strength testing strategies is common in recent literature. Many test data generation strategies based on meta-heuristic algorithms such as Simulated Annealing (SA), Tabu Search (TS), G...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2019
|
| Subjects: | |
| Online Access: | http://umpir.ump.edu.my/id/eprint/29278/1/A%20fuzzy%20adaptive%20teaching%20learning-based%20optimization%20strategy%20for%20generating%20mixed%20strength%20t-way%20test%20suites.wm.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The use of meta-heuristic algorithms as the basis for t-way (where t indicates the interaction strength) and mixed strength testing strategies is common in recent literature. Many test data generation strategies based on meta-heuristic algorithms such as Simulated Annealing (SA), Tabu Search (TS), Genetic Algorithm (GA), Ant Colony Optimization (ACO), Particle Swarm Optimization (PSO), Harmony Search (HS), Cuckoo Search (CS), Bat Algorithm (BA) and Bees Algorithm have been developed in recent years. Although much progress has been achieved, research into new strategies is still relevant owing to the fact that no single strategy can claim dominance over other existing ones (i.e., as stipulated by the No Free Lunch Theorem). Additionally, the adoption of new parameter-free meta-heuristic-based t-way strategies has not been sufficiently explored in the scientific literature. Owing to its proven performance in many other optimization problems, the adoption of the parameter-free Teaching Learning-based Optimization (TLBO) algorithm as a new t-way strategy is deemed useful. Unlike most existing meta-heuristic algorithms, and by virtue of being parameter-free, TLBO does not have any specific parameter controls. Thus, TLBO avoids the need for cumbersome and problem specific tuning process. However, on the negative note, TLBO takes a simplistic approach of performing both global and local search sequentially per iteration. Given that exploration (i.e., globally finding the new potential region in the search space) and exploitation (i.e., locally manipulating best-known neighbourhood) are dynamic in nature depending on the current search space region, any preset division between the two can be counter-productive. Addressing these issues, this thesis proposes a new TLBO variant based on a Mamdani-type fuzzy inference system, called adaptive TLBO (ATLBO), to permit adaptive selection of its global and local search operations. The Mamdani-type fuzzy inference system of ATLBO has three inputs: Quality measure, Diversification measure and Intensification measure and one output: Selection. The three input measures capture necessary details so as to achieve optimality by guiding the search process in the right direction. Quality and Diversification measures are used to achieve solution diversity, whereas the Intensification measure is used to facilitate convergence. The Selection output of the Mamdani-type fuzzy inference system acts as an intermittent switch between global search and local search in ATLBO. The adoption of ATLBO for the mixed strength t-way test generation strategy demonstrates competitive performances in terms of obtained test suite sizes against the original TLBO and other meta-heuristic counterparts. To conclude, ATLBO-based strategy contributes to 39 new best average test suit sizes on benchmarking experiments and is the first parameter-free strategy that addresses generation for both uniform and mixed strength t-way test suites. |
|---|