The Research Of High-dimensional Data Dimension Reduction Technology Based On Shuffled Frog Leaping Algorithm

Since it was proposed,Shuffled Frog Leaping Algorithm(SFLA)has achieved remarkable results in resource scheduling,structural design,parameter regulation and combinational optimization.Compared with intelligent optimization algorithms like Genetic Algorithm(GA)and Particle Swarm Optimization(PSO)Algorithm,the application practice shows that the model optimized by the intelligent optimization algorithm under the same conditions has better performance in cost management,energy consumption saving,resource utilization and economic benefit,and the performance improvement is more obvious with the increase of model size.Feature selection and projection pursuit are dimensionality reduction technologies of complex high-dimensional data to reduce the scale and complexity of data.By analyzing the present features of low-dimensional data,the technology reveals the hidden information and value of highdimensional data,so as to simplify the scale of the problem and improve the ability of data mining.Traditional classical algorithm optimization algorithms such as Genetic Algorithm and Particle Swarm Optimization Algorithm have more research results in the two types of applications,while SFLA in the two types of applications relatively few research results,this situation at the same time gives SFLA new research direction and application space.The performance of SFLA is largely determined by its parameters and updating strategy.Reasonable algorithm parameters and improved algorithm updating strategy can improve the optimization performance of the algorithm,and further expand the application scope of the algorithm in various optimization problems.Therefore,it is of practical significance to carry out the research on parameter optimization and updating strategy of SFLA and carry out its research on Feature Selection and Projection Pursuit model.The main research work of this paper is as follows:1.The parameter optimization and updating strategy of SFLA are studied to improve the optimization performance of the algorithm.Aiming at the problem of unreasonable parameter setting of SFLA in the existing literature,a grouping-subgroup equilibrium orthogonal experimental design method was adopted to establish the factor horizontal orthogonal experimental scheme of Population,Groups Number and Maximum stride length,and the optimal parameters configuration scheme was obtained through experimental simulation.Compared with the worst parameter configuration scheme,the mean value of all parameter configuration schemes and the parameter configuration in relevant literature,the comprehensive evaluation value performance has been improved by 41.56%,28.03% and 41.00%,respectively.It is shown that reasonable parameter configuration scheme can significantly improve the optimization performance of SFLA.2.By introducing Chaotic Memory Weight Factor,Bacterial Chemokine Factor and Balanced Grouping Strategy to improve SFLA,the Chaotic Memory Weight Factor SFLA(ISFLA)and Bacterial Chemokine SFLA(BF-SFLA)are proposed.The improved strategy handles the balance between the global exploration ability and the local search ability of the algorithm,reduces the probability of falling into the local optimal,and further improves the diversity of the algorithm.Through the benchmark test functions and compared with improved algorithm in the literature,the comprehensive evaluation value of ISFLA and BF-SFLA was increased by 29.32% to 33.87%,The mean success rate of reaching optimization was increased by 29.36% to 30.23%,and the average number of iterations mean was reduced by 30.21% to 42.38%,indicating that the two improved algorithms obtained higher optimization precision and faster convergence speed,and had better stability.3.The Feature Selection method based on SFLA is proposed to improve the performance of Feature Selection in high-dimensional biological disease data.In view of the current situation that a large number of irrelevant or weakly correlated features in the data of high-dimensional biological diseases affect the efficiency of disease diagnosis,a high-dimensional biological disease data feature method based on ISFLA and BF-SFLA is constructed,which makes full use of the good optimization performance of ISFLA and BF-SFLA feature selection methods and improves the search ability of feature selection method in the feature space.By introducing Adaptive strategy,Classification accuracy,Effective features occurrence probability,Subsets number and algorithm implement time constraint factors to the feature selection method of SFLA,Adaptive Improved Shuffled Frog Leaping Algorithm(A-ISFLA),Adaptive Bacterial Foraging Shuffled Frog Leaping Algorithm(ABF-SFLA)and Adaptive Bacterial Foraging Shuffled Frog Leaping Algorithm(ABF-SFLA)were is proposed.Three feature selection methods of adaptive adjustment in the process of iterative solution space dimension,feature subset selection more representative,feature selection by biological disease high dimension data validation,respectively.Compared with improved feature selection method in the literature of,ISFLA and BF-SFLA feature selection method for nine biological disease high dimension data set to obtain the mean average classification accuracy was increased by 3.4% to 7.0%,the mean average standard deviation was reduced by 26.9% to 46.7%,and the mean average number of feature subset mean was reduced 3.0% to 52.1%,on the whole have achieved higher classification accuracy,better stability and less number of feature subsets,short the time of feature selection,raise the efficiency of related disease diagnosis and prediction.Compared with ISFLA,BF-SFLA and SFLA,A-ISFLA,ABF-SFLA and A-SFLA feature selection methods,on the premise that the classification accuracy,the algorithm running time were shortened by 67.8%,67.6% and 67.9%,respectively.The three feature selection methods not only ensure the classification accuracy,but also greatly shorten the execution time and improve the efficiency of diagnosis and prediction.4.A Projection Pursuit model based on improved SFLA was proposed to improve its evaluation performance in agricultural drought vulnerability assessment and urban heavy metal pollution assessment.Compared with the improved algorithm in relevant literatures,ISFLA and BF-SFLA were used to optimize the projection tracking evaluation model for agricultural drought vulnerability,the projection index function value was increased by 4.9% to 5.1%,and the standard deviation was reduced by 47.0% to 78.5%.ISFLA and BF-SFLA were used to optimize the projection tracking evaluation model of heavy metal pollution in cities,the projection index value was increased by 6.8%to 7.8%,and the standard deviation was reduced by 10.7% to 68.0%.The higher the projection index function value and the lower the standard deviation,the more accurate and reasonable the index weight obtained,which can further reduce the impact of unreasonable and inaccurate index weight on the evaluation results,and enhance the accuracy of the comprehensive evaluation using the projection pursuit model.