Research On Parameter Tuning Of Intelligent Optimization Algorithm Based On Variance

In the era of industrialization, many control problems of the complex system are concerned with the combinational optimization problems, most of which are NP-Hard problems, for example, the famous traveling salesman problem, and the problems of protein folding and prediction structure in biological calculation. Traditional optimization algorithm in most optimization problems can hardly reach the ideal goals,which means that no effective algorithm is dealing with the optimization problem. At this time, intelligent algorithm can be used to solve this kind of optimization problems as an effective tool. But the performance of the intelligent algorithm will be directly affected by different setting parameters which are the heated research topic in optimization problems,data mining and network security, and there haven’t been a series of solutions at home and abroad. Based on the theory and analysis of Cuckoo search algorithm and Bat algorithm,this thesis not only proposed the improved algorithm according to the variance of fitness value and the parameter selection, but also applied such an algorithm into objective function optimization problem. The contents of this research consist of four parts.(1) This thesis illustrated current background, relevant theory and development process of the intelligent optimization algorithm at home and abroad, and the basic ideas and steps of Cuckoo search algorithm and Bat algorithm.(2) By studies of the relationship between different algorithm parameters and the searching ability of algorithm, and based on the variance of fitness values, this thesis firstly,analyzed the relationship of Cuckoo search algorithm parameters and the variance of fitness values; Secondly, this thesis established the adjustment strategy of parameters according to the searching ability of algorithm, improved the searching ability of the algorithm by adjusting the parameters, and avoided being involved into the local optimum;finally, this thesis made a simulation comparison between the improved Cuckoo search algorithm and the original one through five test functions. And it showed that the improved Cuckoo search algorithm could effectively jump out of local optimum.(3) As it is easy for bat algorithm to converge and fall into local optimum, bat algorithm needs to be improved. In the iterative process of bat algorithm, firstly, this thesis considered the population diversity by the variance of fitness value, and improved thesearching ability of algorithm by adjusting the parameters when the diversity is small;secondly, this thesis selected the best fitness and put forward an improved bat algorithm that changed the bat position through Gaussian disturbance and Lévy flight disturbance.Finally, this thesis made a simulation comparison between the improved bat algorithm and the original through seven typical test functions. And the results showed that the improved bat algorithm was better than the bat algorithm.(4) The improved cuckoo search algorithm is applied to deal with the function optimization problems.