System Of Systems Effectiveness Simulation Analysis Based On Decision Trees

System of Systems(SoS)effectiveness simulation can be said to evaluate SoS effectiveness via simulation and modeling.The steps of SoS effectiveness simulation include analyzing the need,establishing simulation model,simulating with computers and processing simulation results,which aims to have a research relationship between SoS effectiveness and input factors,analyze SoS essential characteristics and guide analysts to evaluate different experimental program.Because of the large-scale and heterogeneity of SoS,SoS simulation experiment data has a characteristics of large-scale,high-dimensional and high-randomness,which leads to a poor performance of traditional mathemetics method.Thus,data mining is adopted to analyze complex simulation data.In this thesis,we mainly study the following two aspects of SoS effectiveness simulation:(1)Analyze the relationship between input factors and SoS effectiveness;(2)Find the important factors interval,that is,to explore how input factors range when the system effectiveness value reaches a certain level.The main contents of this thesis are as follows:(1)This paper studies SoS effectiveness simulation analysis method.In this paper,SoS effectiveness simulation analysis method and framework are studied based on the concept of SoS.To explore the relationship between input factors and SoS effectiveness,decision tree algorithm is chosen comparing to common simulation data analysis methods.(2)This paper presents an improved algorithm of decision tree.The thesis also introduces several classic decision tree algorithms—ID3 algorithm,C4.5 algorithm and CART algorithm.According to characteristics of different algorithms,the thesis give the advantages and disadvantages.Fuzzy sets and rough sets are introduced to to provide a theoretical basis for improved algorithms.In view of the multiple types of experimental data,ie,continuous data and discrete data in data sets,weighted fuzzy classification and regression tree algorithm(WF-CART)with the aim of improving fuzzy CART is proposed.The performance of the proposed algorithm is analyzed and validated by data sets Iris,Diabetes and Wine.Experimental results show that the improved algorithm can reduce the complexity of the model under certain classification accuracy.Considering the characteristics of the interactions among multiple factors in the same unit,The thesis also introduces a hybrid intelligent systems technique: weighted multivariate fuzzy based on rough sets Decision tree(FR-MDT)with the aim of improving decision tree algorithm based on rough set and the weighted fuzzy CART algorithm.FR-MDT has also demonstrated its efficiency and effectiveness as compared with a conventional Multivariate Decision Tree and weighted fuzzy Decision Tree.(3)Algorithm applicationAccording to a report from United States Naval Postgraduate School about NUCAS system,this thesis designs the scenario and experimental design scheme.Then,SoS effectiveness simulation experiment is carried out on SEAS platform.FR-MDT algorithm is used to analyze the data which are obtained from the experiment.Through analyzing the dataset using the Intelligent Systems Techniques,it has been shown that the relationship between factors and SoS effectiveness can be found or can be classified through combining those techniques together.What's more,suggestions can be provided for the actual combat plan formulation and the SoS construction of combat mission.