基于嵌入式机器学习的路面积水量预测

Research on Machine Learning RegressionAlgorithm in Rainfall Prediction

此研究以局部加权线性回归算法(LWLR)采集某一固定区域内的单位时间降雨量及积水量数据,构建地面积水量预测模型,并不断优化预测模型.结果表明,单位时间降雨量与单位时间雨水流水量呈明显的正相关,当降雨量逐渐增大过程中,单位时间雨水流入量增加的速率逐渐降低.路面现有积水深度与单位时间流出量总体呈现负相关,在局部出现正相关的情况.路面积水量总测试误差约为15.85%,均方差为5.09 cm,测试样本的误差相对稳定.

In this paper,the local weighted linear regression algorithm(LWLR)is used to collect the data of rainfall and accumulated water per unit time in a fixed area,and the prediction model of land area water quantity is constructed,and the prediction model is continuously optimized to ensure the reliability of the prediction data.The results show that there is a significant positive correlation between rainfall per unit time and rainfall flow per unit time.When rainfall increases gradually,the increasing rate of rainfall inflow per unit time decreases gradually.There is a negative correlation between the existing road surface accumulated water and the outflow per unit time,but there is a positive correlation in some places.The total test error of road area water quantity is about 15.85%,the mean square error is 5.09cm,and the test sample error is relatively stable.