摘要：Based on analyzing the basic characteristics of small-scale flash flood disasters, learning and con- structing a Bayesian network based on flash flood. related data in the study area, through different algo- rithms and comparisons with artificial neural net- work models and support vector machine models, tree enhancement is found. The Bayesian network performs well in risk assessment of mountain torrent disasters. It constructs dynamic critical rainfall mod. els based on multiple linear regression, dynamic crit- ical rainfall models based on BP neural network, and dynamic critical rainfall models based on support vector machines. Through a comparative analysis of the average relative error, root mean square error anddetermination coefficient of the prediction results of the three models, the results show that the dynamid critical rainfall model based on multiple linear re-gression is superior to the other two models in pre- dicting the accuracy of the dynamic critical rainfall. It is closer to the critical rainfall of the time, and the accuracy is relatively higher, which is more suitable for the prediction of dynamic critical rainfall in the basin. An intelligent rainfall prediction model based on BP artificial neural network was established which provided a new method for critical rainfal prediction. The research results of flash flood warn- ing indicators and intelligent prediction in this paper can provide certain technical support for flash flooc monitoring and early warning forecast and risk as- sessment.
