Research And Improvement Of Gravitational Search Algorithm Based On Data Feature Analysis

Gravity search algorithm is a heuristic optimization algorithm inspired by Newton’s law of universal gravitation.Because of its robustness,universality and ease of use,it has been widely applied in solving complex optimization problems,and has become one of the research hotspots in the field of optimization algorithms.In recent years,there have been many researches on the improvement of gravitational search algorithm and its application.However,in the partical application,it still faces the following problems:(1)In some optimization scenarios,the algorithm is easy to fall into the local optimal solution,which may lead to the deterioration of the optimization performance;(2)The optimization performance of the algorithm in high-dimensional scenarios drops sharply,which leads to the need to find effective improvement methods for high-dimensional optimization scenarios.Therefore,aiming at the above-mentioned problems,this paper conducts research on the improvement and application of the optimization performance of the gravitational search algorithm.The main contributions of this paper are as follows:First of all,for the problem that the algorithm is easy to fall into the local optimal solution,a new research method is adopted in this paper.From the view of data,by selecting classic benchmark test function set and performing simulations on the it,the test functions can be divided into two parts,according to whether the algorithm can find the global optimal solution or fall into the local optimal solution.Then,according to the characteristics of the two parts of the problem set,constructing some typical test functions to verify and analyze the factors that affect the optimization performance of the algorithm.Through the above analysis,two important factors affecting the optimization performance of gravitational search algorithm are found.One is the Kbest mechanism that makes particles with better fitness values may not be able to effectively guide the direction of the population movement.Particles are easily attracted by the local optimal solutions,which can make the algorithm easily falling into the local optimal solution.The second one is that the gravitational search algorithm has the center convergence characteristics,which leads to the algorithm easy to fall into the local optimal solution,especially in the scenario where the global optimal solution is located near the boundary and the suboptimal solution with strong interference exists in the center of the region.Secondly,aiming at the two major factors that affect the performance of the algorithm,a balanced gravity search algorithm is proposed.In the algorithm,the balance operator is introduced.In the process of algorithm optimization,the particles are balanced with the historical optimal solution as the center.After the balance operation,the particles are centered on the historical optimal solution,and the central convergence characteristic can be utilized to make the population search near the historical optimal solution to better make use of its information.In addition,this can also provide an additional particle generation mechanism,giving the particles in the local optimal region an extra chance to reborn in the global optimal area.By this way,the occupation of the particles in the global optimal region may be raised and the drawbacks of Kbest mechanism in GSA could be compensated.Subsequently,aiming at the problem of a sharp decline in high-dimensional scenarios,a manifold-guided dimension division strategy is proposed.By introducing the strategy of manifold learning,the particle dimensions are divided accroding to the comparing results of the dimensional differences of adjacent particles in the high-dimensional and low-dimensional spaces.Compared with the random dimension division strategy,it can avoid the blindness and inefficiency of random dimension division to a certain extent,reduce the number of times of re-division in the optimization process,and improve the optimization performance.Finally,the improved gravitational search algorithm is applied to the rural leisure tourist flow scheduling problem.Taking the number of people dispatched from the overloaded operation point to the underloaded operation point as the independent variable,the objective function is to maximize the overall tourist experience value and operating profit of the region.The simulation results show that the algorithm proposed in this paper can effectively solve the abovementioned passenger flow scheduling problem,and has better advantages than other heuristic algorithms.