基于XGBoost算法的胶凝砂砾石劈拉强度预测分析

Predictive Analysis of Cemented Sand and Gravel Splitting Tensile Strength Based on XGBoost Algorithm

摘要:

将水泥质量浓度、砂率、水胶比和粉煤灰质量浓度设为输入变量， 28 d劈拉强度设为输出变量，用极端梯度提升树（XGBoost）算法对胶凝砂砾石（CSG）的劈拉强度进行预测，并与随机森林（RF）算法的预测结果进行对比，以决策系数（R²）、均方根误差（RMSE）、平均绝对误差（MAE）和平均百分比误差（MAPE）作为评估标准对2种算法进行对比分析.结果表明： XGBoost算法的R²为0.968 1，具有高度的预测准确性；相比表现良好的RF算法，XGBoost算法测试集的RMSE和MAE均降低了0.003， MAPE降低了0.32%，表明XGBoost算法能够对CSG劈拉强度进行更为精准的预测.

Abstract:

The XGBoost algorithm was used to predict the splitting tensile strength of cemented sand and gravel （CSG） with mass concentration of cement， sand ratio， ratio of water to binder and mass concentration of fly ash as input variables and 28 d splitting tensile strength as output variables. Additionally， the result of XGBoost algorithm was compared with that of random forest （RF） algorithm， and the coefficient of determination （R²）， root mean square error （RMSE）， mean absolute error （MAE） and mean absolute percentage error （MAPE） were used as the evaluation criteria. The results show that the value of R² by XGBoost algorithm is 0.968 1， which means XGBoost algorithm has a high degree of prediction accuracy. Compared with the well-performing RF algorithm， the values of RMSE and MAE of the testing set of XGBoost algorithm are both reduced by 0.003， and the value of MAPE is reduced by 0.32%. XGBoost algorithm is more accurate in predicting the splitting tensile strength of CSG than RF algorithm.