BLASTING, Volume. 39, Issue 1, 16(2022)

Fragmentation Prediction of Rock Blasting by LOO-XGboost Model

[in Chinese]1,2, [in Chinese]3, [in Chinese]1,2, [in Chinese]1,2, [in Chinese]1,2, [in Chinese], [in Chinese], and [in Chinese]
Author Affiliations
  • 1[in Chinese]
  • 2[in Chinese]
  • 3[in Chinese]
  • show less

    In order to solve the problems of insufficient data utilization and large fluctuation of prediction accuracy in predicting rock blasting fragmentation using machine learning method under the condition of small samples,a LOO-XGboost model is built by Python 3.7,which combines the leave-one-out method(LOO) with the eXtreme Gradient Boosting(XGboost) algorithm.Firstly,31 sets of blasting data are selected for training and prediction.By calling different parameters,the optimal built-in parameters of the model are obtained as follows:tree model as solution method,the learning rate of 0.30,the number of decision trees of 50,the maximum iteration depth of the decision tree of 3,the minimum sample number of leaf nodes of 3,and the random sampling ratio of 0.8.Through comparing the prediction results with the support vector machine regression(SVR),BP neural network(BPNN),random forest(RF) model and XGboost model under 10 fold cross validation under the same conditions,the LOO-XGboost model has significantly higher prediction accuracy than the other four models.The correlation coefficient,root mean square error,and average absolute error are 0.9128,0.0587 and 0.0342,respectively.The results show that the LOO-XGboost model can not only guarantee the data utilization in the case of small samples,but also improve the prediction accuracy,and it is suitable for the prediction of rock blasting framentation.

    Tools

    Get Citation

    Copy Citation Text

    [in Chinese], [in Chinese], [in Chinese], [in Chinese], [in Chinese], [in Chinese], [in Chinese], [in Chinese]. Fragmentation Prediction of Rock Blasting by LOO-XGboost Model[J]. BLASTING, 2022, 39(1): 16

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Received: Sep. 1, 2021

    Accepted: --

    Published Online: Jan. 24, 2024

    The Author Email:

    DOI:10.3963/j.issn.1001-487x.2022.01.003

    Topics