Rockburst Intensity Prediction Based on African Vultures Optimization Algorithm-Random Forest Model

Rockburst is a common geological disaster in geotechnical engineering construction, rockburst prediction is closely related to geotechnical engineering and has great significance. Therefore, this paper proposed an African vultures optimization algorithm and random forest combined with AVOA-RF model to achieve the better performance of rockburst prediction. Six key parameters about microseismic, i.e., cumulative event number, event rate, cumulative release energy logarithm, energy rate logarithm, cumulative apparent volume logarithm, and apparent volume rate logarithm are selected to constitute rockburst prediction index system. A data set of 78 rockburst cases is constructed by collecting the monitoring data of rockburst microseismic of Jinping II Hydropower station, and used for train and test the proposed AVOA-RF model. In the process of model building, the average error rate obtained by 10-fold cross validation is used as the fitness value in the African vultures optimization algorithm. The model's optimal parameters were mtry=2 and ntree=41. Then, the accuracy, precision, recall, F1-score, macro-average, micro-average, and AUC are selected to evaluate the model's prediction performance. The results showed that AVOA-RF model has good performance in rockburst data of test sets and new engineering projects, the accuracy on the test set is 94.4% and the model of AUC is 0.9974. The feature importance obtained by the AVOA-RF model indicated that cumulative release energy logarithm plays the most important role in rockburst. Besides, the proposed model is compared with support vector machine, decision tree, random forest, and probabilistic neural network. The comparison results show that the proposed model has the better performance.

Keywords: Rockburst classification, Machine learning, Random forest, African vultures optimization algorithm.