Application of different objective weighting methods in rockburst - RASIM2022

Aiming at the subjective impact caused by subjective weighting method in the establishment of rockburst risk attribute identification and prediction model, the combined weighting method was introduced to eliminate the influence of unreasonable weight of each index. Eight discriminant indexes affecting the risk of rockburst were selected. Through the literature review, relevant rockburst cases were collected and 8 index rockburst sample sets were established. A variety of machine learning methods were used to analyze the feature importance of discriminant indexes in the samples, and then to determine the objective weight values of different methods. Combined with the subjective weights determined by the Analytic Hierarchy Process, the combination of weights were successively carried out. Based on the attribute recognition theory, the combination weights with different objective weight methods were established to identify the rockburst risk attributes, and verified by the confidence criterion. The results show that the objective weight value determined by the feature importance analysis of machine learning reduces the subjectivity of the weight of Analytic Hierarchy Process method, improves the prediction accuracy, and makes the prediction result closer to the fact.