The Assessment Strategy for Intelligent Algorithms Using Improved Osculating Value Method

During the past decades, dozens of intelligent algorithms (IAs), such as genetic algorithm (GA), ant colony algorithm (ACA), particle swarm optimization (PSO), emerge to solve complex optimal problems. But it is still an open problem to impartially and effectively evaluate their performances. A new assessment strategy for IAs based on improved osculating value method is firstly proposed to provide a solution for this problem. The idea of the osculating value method is that evaluation index is divided into positive and reverse index, then find out the most advantage and worst points for the evaluation index, the distance is calculated for each evaluation unit and the advantage and the worst points, the transformed distance as comprehensive index reflecting advantage and disadvantage of each sample is called close value, according to the close value, finally determine the order of each evaluation unit. The paper takes the GA and ACA for solving TSP as example, chooses solving time, algorithm complexity, solution accuracy and coding complexity as evaluation index system, it can assess algorithms from many perspectives, and avoid the disadvantage of traditional single index, by computing and analysis in certain experiment condition, it shows that ACA for TSP is better than GA.