Extracting the characteristic indicators of SF6 gas insulated switchgear that reflects equipment insulation status to evaluate its insulation state is the basis for safe operation. Aiming at this problem, this paper proposes an improved minimum-redundancy-maximum-correlation criterion method to extract the key evaluation indices of GIS (gas insulated switchgear) equipment insulation state. We extract the insulation state characteristics according to the ultra-high frequency, ultrasonic wave and SF6 during the insulation defect deterioration process. An improved minimum-redundancy-maximum-relevance criterion is proposed to form the key feature indicators in order to avoid inaccurate evaluation caused by redundant information of multi-source feature indicators. The experimental results show that the optimized feature index can lower the influence of redundant information, and higher evaluation accuracy is achieved.