Analysis And Research On Improved Artificial Fish Swarm Algorithm

The thesis mainly uses Artificial Fish Swarm Algorithm (AFSA) as research object, systemati-cally researches the structure principle and improvement strategy of AFSA, further more, proposes amulti-behaved AFSA which has global convergence preference. The major job and contribution de-scribed as following:(1) Detailed analyzes the structure and principle of AFSA. It gives the analysis and ex-planation of convergence principle from some ways, such as fish swarm behavior and algo-rithm parameter. Use the simulation result to prove the excellent convergence of AFSA, de-tailed analyzes the major parameter of algorithm, for instance, visual and step. It discussesthe effect of algorithm convergence performance from parameter.(2) Summarized the disadvantages of typical AFSA, proposes the methods and steps ofimprovement algorithm performance. According to the difference of improvement methods,detailed introduces and analyzes some excellent algorithm, such as quantum AFSA andchaos AFSA.(3) On the basic of former research, to improve the fault of typical AFSA, absorbs theadvantages of many improvement algorithms, proposes an improved and multi-behaviorAFSA which has global convergence preference. Synthetically improves the typical AFSAin some aspects, such as the behavior of artificial fish-swarm and the set of system parame-ter. It tests the performance of algorithm through Rastrigin function, Griewank function andRosenbrock function. Generally contrasts and researches it through many aspects such asthe speed of algorithm iteration，the accuracy of algorithm convergence and the complex ofalgorithm. Finally, compare with traditional AFSA by some data, we proved that the mul-tiple behavior global optima AFSA has better convergence performance. It is an excellentimprovement method.