Research On Classification Based On Intelligent Optimization Algorithms

With the development of computer technology and the advent of the era of big data,classification problems have attracted more and more attention because they can help people quickly obtain effective information.Among them,SVM has become an important direction of classification research because it can solve the convex optimization problem and has the outstanding advantages of VC dimension,but the classification accuracy still needs to be improved.The traditional support vector machine parameter optimization methods are mostly difficult to implement and have a low precision.Intelligent optimization algorithms have been widely used in solving optimization problems because of their strong parallel processing ability and strong global search ability.Therefore,based on the bacterial foraging optimization algorithm,this paper proposes related research methods for low-dimensional classification problems.Based on the ant lion optimization algorithm,this paper proposes related research methods for high-dimension classification problems.Firstly,this paper proposes the search combination improved bacterial foraging optimization algorithm(SCIBFO)for low-dimensional data classification problem.The adaptive Cauchy mutation method was introduced into the bacterial foraging optimization algorithm,in addition to the idea of simulated annealing algorithm is also integrated into the bacterial foraging optimization algorithm.The simulation experiment verifies the validity of the SCIBFO algorithm by using the test function.Then the SCIBFO algorithm is applied to the classification problem of low-dimensional data,and the validity of the proposed classification algorithm is verified by some public datasets.Secondly,this paper proposes the adaptive differential evolution ant lion optimization algorithm(ADEALO)for high-dimensional data classification problem.In this algorithm,we propose an adaptive strategy for ant random walk,so that the range of random walk of ants becomes larger as the number of iterations increases,so as to enhance the search ability of the algorithm while avoiding the algorithm falling into local optimum.In addition,the idea of the differential evolution algorithm is integrated into the ant lion optimization algorithm to mutate,cross and select the ant lion individuals,and use the greedy strategy to retain better individuals for the next iteration,thereby increasing the diversity of the population and can avoid the algorithm falling into local optimum.Simulation experiments demonstrate the effectiveness of the ADADELO algorithm on high-dimensional problems using standard test functions.The ADADELO algorithm is used in high-dimensional complex classification problems,and the validity of the proposed classification algorithm is verified by some public datasets.Compared with the traditional support vector machine parameter optimization method,the search combination improved bacterial foraging optimization algorithm(SCIBFO)have higher classification accuracy and stability,and the adaptive differential evolution ant lion optimization algorithm(ADEALO)also have higher classification accuracy and stability.In addition,compared with the bacterial foraging optimization algorithm(BFO)and two other existing algorithms,the proposed SCIBFO algorithm has better classification accuracy and stability for low-dimensional classification problems.Compared with ant lion optimization algorithm(ALO)and two other existing algorithms,the ADADELO algorithm proposed in this paper has better classification accuracy and stability for high-dimensional complex classification problems.