Grey residual error model of highway traffic accident forecast

Grey residual error model(GREM) was proposed to improve the prediction accuracy of highway traffic accidents with high randomicity.GREM was used to depict the accident prediction values piecewise with the linear superposition of multiple exponential functions,and it overcomed the shortcomings of grey model(GM).GM was only able to describe the monotonous change process and only could be used to estimate the sequences consistent with obvious exponential law.The manifold factors affecting GREM were analyzed and concluded in detail,including amending objects,data selection,data pretreatment and amending methods.Then,4 practical GREMs were summed up.Analysis result indicates that in comparison with GM,all GREMs' predicting errors reduce by 70%～80%,and the one modeled by residual error directly and restored only one time has low complexity and its predicting error is less than 5%.6 tabs,4 figs,19 refs.