Study And Application Of Backtracking Search Optimization Algorithm

Swarm intelligence optimization algorithm is an important research direction in optimization, as optimization is a very important research area in applied mathematics. With development of computer technology, Swarm intelligence optimization algorithm is applied widely in many science and technology areas and attach much attention now; Backtracking search optimization algorithm is a new evolutionary algorithm based on population, its framework is similar to the differential evolution algorithm, while in the mutation and crossover operation it is essentially different from differential evolution algorithm, the mutation and crossover operation of BSA is novel and efficient, which can well balance exploration and exploitation ability,However, BSA still has some deficiencies, first, as mutation scale factors’ deviation is big, too large or too small value will be generated during evaluation, that can easily cause bad convergence or increase the possibility of tracking into local optimal, second, the way of random call between two cross strategy cause some blindness to search, which reduce the efficiency of algorithm; To solve these problems, we propose several improvements of BSA and BSA-based hybrid evolutionary algorithm, and apply some of them to solve a series of practical problems; the main work is as follows:This paper choose evolutionary algorithm as the starting point and introduce background of intelligent optimization algorithms and popular intelligent optimization algorithms and their variants in recent years firstly. introduce BSA algorithm and compare it with other intelligent optimization algorithms in performance. Aim at the lack in exploitation of BSA, an improved backtracking search algorithm(IBSA) is presented; to overcome shortcomings of mutation operator, we propose a new mutation operator which can effectively improve the exploit capabilities and accelerate BSA’s convergence speed, compare the performance of the new mutation operator with a variety of classic mutation operators from differential evolution algorithm and evaluate them; Consider BSA’s joint crossover characteristics, design an intelligent strategy based on fitness for calling two cross strategies, which increases stability of BSA. In the end, to solve Parameter Estimation for Frequency-Modulated Sound Waves(FM),propose a hybrid algorithm based on differential evolution algorithm and backtracking search algorithm, and add intelligence-based adaptation strategies to call the two algorithms’ mutation operator; To solve Absolute Value Equation, propose an improved BSA based on fitness Euclidean-distance ratio, which choose leading individual according to fitness Euclidean-distance ratio, that enhance the performance of global search; To System identification method for Hammerstein model, propose a compact BSA can quickly solve the problem.