Improvement And Application Of Backtracking Search Optimization Algorithm

The intelligent optimization algorithm has obvious advantages in solving such complex engineering optimization problems because of its unique and efficient operation mechanism,and has become a research hotspot in the field of intelligent optimization.Backtracking Search Optimization Algorithm(BSA)is a new population based intelligent optimization algorithm which is proposed in recent years.The algorithm has the advantages of simple structure,less parameters to be determined,memory history population and strong search ability.This has aroused wide concern of scholars.At present,the research of BSA is still in the initial stage.Compared with some classical optimization algorithms,the theory is not perfect enough,and the application of the algorithm has yet to be extended.The algorithm itself has a long search time,easy to fall into the local optimum and so on.To improve the global performance and extend the application domains of BSA,in this thesis,the original BSA is studied,and some improved variants of BSA are presented,and the improved algorithms are applied to different practical problems.The main contents and innovations of the thesis are as follows:(1)An improved BSA(GNBSA)based on guidance with optimal individual and niche technology is proposed for the shortcomings of the slow convergence rate of basic BSA and the easy loss of population diversity.In the method,the strategy with learning from the best individual is introduced in the mutation operator of original BSA to improve the convergence speed of it firstly.Secondly,a niche repulsing technology is designed in the thesis.The niching radius is determined according to the average minimal distance between every individual and the other individuals,and some parts individuals with high similarity are deleted,some new individuals are generated by a new mutation method which is designed with combining the worst information of current generation,and the new individuals are supplemented in the new population to maintain the constant number of population,the diversity of the population is increased by this operation.The simulation results indicate that the convergence speed and the diversity of population are fully considered in the improved BSA,which can balance the exploration and exploitation abilities of population,and the performance of original BSA is improved.(2)For the convergence accuracy of BSA is low and the learning ability of it isweak,a new learning backtracking search optimization algorithm(LBSA)is proposed in this thesis to solving these problems.The algorithm draws on the experience of learning idea from TLBO algorithm,and two new mutation operators are designed in it.On one hand,the global best information of the current generation and the historical information in BSA are combined to generate mutation individuals.On the other hand,the mutation individuals are generated by learning knowledge from the best individual and another random individual,and repulsing the worst individual of the current generation.The individuals of the population randomly chose a kind of mutation methods in each generation.The new algorithm is tested on the CEC2005 function set.The results show that the convergence accuracy of LBSA is obviously improved.(3)In order to extend the application field of BSA,the two algorithms GNBSA and LBSA,which are improved in the first two chapters,are applied to chaotic time series prediction and nonlinear system modeling respectively.Compared with other algorithms,the results show that the two improved algorithms have achieved good results,and further verify the effectiveness of the improved algorithm.In view of the problem of inventory optimization and sales forecast of rookie network logistics,a forecasting model of commodity sales based on hybrid backtracking search algorithm(HBSA)to optimize feedforward neural network is proposed.At present,there is no report on BSA in the field of sales forecast.In this model,through fusing feature extraction,neural network and hybrid optimization algorithm,achieve accurate prediction of sales of goods.The validity of the model is verified by the commodity transaction data provided by Ali Tianchi platform,which provides an effective method for commodity sales forecast.In summary,this thesis has carried on the thorough research and the analysis to the BSA not only proposed several effective improvement methods,but also extended the application field of the algorithm,which provides reference for the future research of BSA.