Improved Artificial Fish Swarm Algorithm Based Cloud Model Theory

Artificial Fish Swarm Algorithm (AFSA) is a rising kind of intelligent bionic algorithm recently. This algorithm’s advantage is fast convergence rate, robust, no higher parameters requirement and so on. But its later convergence speed and optimal rate is not high, so many specialists and scholars put forward the improvement strategies. In this thesis, an improved Cloud Artificial Fish Swarm Algorithm (CAFSA) is proposed, which is based on cloud model theory and feedback mechanism. It overcomes artificial fish algorithm’s disadvantage of random generating next position strategy, but using the two dimensional normal cloud generator generates cloud drop to improve artificial fish’s next position. For the cloud model----"a single cloud drop is irregularity, however, a large number of cloud droplets keep stability tendentiousness", it is known that if the artificial fish scale many iterations and in a sufficient number of cases, the introduction of cloud model will improve the global optimal speed and precision. Additionally, this paper introduce a feedback factor which will render larger random behavior probability in the early stage of algorithm and larger drop produce probability in the later stage, and after this, convergence performance is improved greatly. Then using the limited Markov chain stability theorem demonstrates the CAFSA’s convergence. And design an comparable experiment with another two improved artificial fish algorithm and the algorithm of the paper presented, which is concluded that CAFSA has achieved good result in convergence precision. At last using AFSA and CAFSA solve the classic NP problem of TSP, and comparing with based on the pheromone of ant colony optimization algorithm (ACO algorithm), it’s concluded that the improved algorithm than original artificial fish algorithm has an advantage in solving TSP problem, this also proposes another thinking strategy to solve TSP problem.