基于改进KNN模型的城市道路事故持续时间预测

Prediction of Urban Road Accident DurationBased on Improved KNN Model

文中提出一种改进的KNN(K-Nearest neighbor)预测模型.通过对训练集的多元回归建模分析得到对事故持续时间占主要影响地位的多个因素及其beta值,并代入到传统KNN模型中,改善模型所用欧氏距离,设计出一种改进的KNN预测模型.结果表明:这种改进的KNN预测模型能够较为准确的预测城市道路环境下的事故持续时间,在K值取不同大小时,测试集预测结果的平均相对误差为15.79%~16.24%,在K值均取最优值9时,模型的平均绝对误差相对于传统KNN模型降低了42.15 s,平均相对误差降低了10.3%.

An improved KNN(K-Nearest neighbor)prediction model was proposed.Through the multivariate regression modeling analysis of the training set,several factors that mainly affect the accident duration and their beta values were obtained.The values were substituted into the traditional KNN model to improve the Euclidean distance used in the model,and an improved KNN prediction model was designed.The results show that the improved KNN prediction model can accurately predict the accident duration in urban road environment.The average relative error of the prediction results of the test set is 15.79%~16.24%when the K value is different.When K values are all taken as the optimal value of 9,the average absolute error of the model is reduced by 42.15s and the average relative error is reduced by 10.3%compared with the traditional KNN model.