Improvement DACS3 Searching Performance using Local Search

Currently, Ant Colony Optimization (ACO) metaheuristic becomes the most prominent techniques applied in TSP. It is based on the cooperation of a complex society of ants through a chemical substance called pheromone. Several versions of metaheuristic ACOs’ have been developed through several improvement processes to produce better algorithm. Past research has proposed a Dynamic Ant Colony System with Three Level Updates (DACS3) algorithm that embedding a Malaysian House Red Ant behavior into current ACS. The algorithm consists of three levels of pheromone updating rules such as local, intermediate and global pheromone updates. Embedding such behavior has improved quality of solution and time taken to reach the solution. However, the algorithm performance is reduced for large data set. Therefore this research aims to improve the algorithm using various supportive and improvement strategies. Three supportive strategies are used such as dynamic candidate list, elitist ants and pheromone trail smoothing whereas improvement strategies used local search. The capability of DACS3 is measured based on quality of solution, time taken to reach the solution and algorithm performance. Moreover, ROC test was carried out between DACS3 and ACS on TSP datasets ranging from 100 to 318 cities. The result shows that applying the above strategies improve the quality of solution on few data and remain on others. Nonetheless it also improves the time taken to reach the solution by 4%-90%. The ROC test result shows that DACS3 is more sensitive then ACS. This research proves that applying various supportive and improvement strategies has improves the DACS3 performance.