关键词: 安全工程, 岩体质量评价, 岩体质量分级, 白鲸优化, 随机森林, 交叉验证

Abstract:

Rock mass quality classification is the foundation of initial underground engineering design and construction.In order to evaluate rock mass quality more accurately，this study used beluga whale optimization（BWO）to optimize random forest model（RF），a BWO-RF model which can be used for rock mass quality evaluation was proposed.At the same time，the rock mass quality evaluation models of sparrow search al-gorithm optimized random forest（SSA-RF），particle swarm optimization optimized random forest（PSO-RF）and non-optimized random forest（RF） were constructed for comparison.Before the models construction，a data-base containing 131 engineering cases data was established through literature review and field test data collec-tion.After writing the code of models construction，the training and testing of the four models were completed by using the database.Based on the model test results，five model evaluation indexes，accuracy，precision，recall，F1 score and AUC，were used to compare and select the best model of the four kinds of rock mass quality eva-luation models.The results show that the BWO-RF model has the best performance among the four kinds of rock mass quality evaluation models，and each evaluation indexes of model are better than the other three mo-dels，indicating that the BWO-RF model has better practicability in the evaluation of rock mass quality.Through the test set，the prediction accuracy of BWO-RF model proposed in this study is 90%，which can provide a reliable reference for practical engineering construction and has practical engineering application value.

Key words: safety engineering, rock mass quality evaluation, rock mass quality classification, beluga whale optimization, random forest, cross-validation