CAJ | 학술논문

从文献资料中收集并整理了45组各类危险边坡数据实例，结合粗糙集理论的数据挖掘功能和BP神经网络理论的非线性映射功能，建立了基于粗糙集-BP神经网络( RS-BPNN)理论的边坡稳定性预测模型.利用粗糙集对离散化后的数据进行了属性约简，利用神经网络对约简前后的数据进行了网络训练和仿真，并对其中五组边坡的安全系数和稳定状态进行了预测.结果表明，未经约简的BP网络安全系数预测的平均误差率为14.51％，约简后的RS-BP网络预测的平均误差率为7.24％，且经过粗糙集约简后边坡的预测状态与边坡的实际状态更加吻合.
종문헌자료중수집병정리료45조각류위험변파수거실례，결합조조집이론적수거알굴공능화BP신경망락이론적비선성영사공능，건립료기우조조집-BP신경망락( RS-BPNN)이론적변파은정성예측모형.이용조조집대리산화후적수거진행료속성약간，이용신경망락대약간전후적수거진행료망락훈련화방진，병대기중오조변파적안전계수화은정상태진행료예측.결과표명，미경약간적BP망락안전계수예측적평균오차솔위14.51％，약간후적RS-BP망락예측적평균오차솔위7.24％，차경과조조집약간후변파적예측상태여변파적실제상태경가문합.
Forty-five sets of various data of dangerous slope were collected and arranged from literature in this paper, and prediction model of slope stability based on RS-BPNN Theory was established by using data mining functions of Rough Set Theory and nonlinear mapping function of BP Neural Network Theory. BP Neural Network was employed to train and simulate the discrete data,before and after that had been reduced by using attribute re-duction based on Rough Set. The results show that the average error rate of safety coeffi-cient prediction of BP Network and RS-BP Network is 14 . 51% and 7 . 24% with or without attribute reduction,respectively,and the predicted state of the slope is more consistent with the actual state through the use of attribute reduction based on Rough Set.