Hybrid Gravitational Search Algorithm Based On Fuzzy Logic And Its Application

More and more global optimization problems are often met in our life. The classic traditional algorithm can not satisfy the needs of the people to solve these problems. Therefore many heuristic optimization algorithms have been proposed. These algorithms simulate the crowd behavior of animals or the theory of physics, and have been applied in various areas. The gravitational search algorithm(GSA) is one of them. GSA has the advantages of simple operation, high efficiency, but it has shortages of falling into local solution and the poor accuracy of solutions like other global optimization algorithm. So GSA still exist wide researching space. The research results are summarized as follows:1. In this thesis, a new hybrid gravitational search algorithm(HGSA) is proposed. Here the local search technique(LST) is incorporated into the optimization process of GSA based on fuzzy logic. The tactic makes full use of the exploration ability of GSA and the exploitation ability of LST. The hybridization is performed by using the GSA with the probability p and local search technique with the probability1-p for each agent. The probability p is controlled by fuzzy system. The HGSA is tested on 23 benchmark test functions. By comparing HGSA with GSA and other algorithms, numerical results demonstrate that the optimization ability of HGSA is superior to other GSA.2. According to the idea of minimizing cross-entropy between uncertain returns and target returns when returns and risks satisfy certain level, a portfolio selection model is proposed based on cross-entropy of uncertain variables. In this model, the objective is the cross-entropy of uncertain returns and target returns, and the returns and risks are described by expectation and variance. At last an example is given to illustrate the correction of the model, and the excellence of HGSA.