Construction of composite indicator system based on simulation data mining

The indicator system is the foundation and emphasis in the effectiveness evaluation of system of systems (SoS). In the past, indicator systems were founded based on qualitative methods, and every indicator was mainly determined by the expert with experience. This paper proposed a brand-new method to construct indicator systems based on the repeated simulation of the scenario space, and calculated by quantitative data. Firstly, the selection of key indicators using the Gini indicator importance measure (IIM) is calculated by random forests (RFs). Then, principal component analysis (PCA) is applied when we use the selected indicators to construct the composite indicator system of SoS. Furthermore, a set of rulesare is developed to verify the practicability of the indicator system such as correlation, robustness, accuracy and convergence. Experiment shows that the algorithm achieves good results for the construction of composite indicators of SoS.