一种复杂系统评估指标冗余识别和去除方法OA

Redundancy in indicator systems is a critical challenge in complex system evaluation.To address the limitations of existing methods that overlook causal mechanisms and rely on redundancy thresholds,a Markov blanket-based approach to indicator reduction is proposed.An optimization objective integrating the minimum redundancy criterion with the maximum information retention criterion is constructed,and both a Markov blanket identification algorithm and a genetic algorithm for indicator reduction are designed to solve the underlying combinatorial optimization problem of redundancy identification and elimination.Experimental results demonstrate the rationality of the proposed method,showing that it achieves a well-balanced trade-off among reduction rate,stability,and interpretability,and yields substantial improvements in redundancy identification accuracy compared with methods such as MIC-MAC and causal entropy.