NEURAL NETWORK MODELING OF 85TH PERCENTILE SPEED FOR TWO-LANE RURAL HIGHWAYS

摘要

The present study was undertaken to develop neural network (NN) models to predict 85th percentile speed(V_(85)) for two-lane rural highways in Oklahoma. Several input parameters, namely physical characteristicsof road, traffic parameters, pavement condition indices, and accident data were considered in developingthe NN models. The physical characteristics of road include surface width (SW), shoulder type (ST), andshoulder width (SHW). The traffic parameters cover average daily traffic (ADT) and posted speed (PS).The pavement condition parameters include skid number (SN) and international roughness index (IRI).The location collision rate, statewide collision rate (overall, fatal, and injury), and percentage unsafespeed drivers (USD) were covered in the accident data. Four different models - Model 1, Model 2, Model3, and Model 4 - were developed. Model 1 included physical characteristics of road and traffic parametersincluding PS, while Model 2 covered all the parameters included in Model 1 except for PS. Similarly,Model 3 considered accident data with all the parameters included in Model 1. Model 4 used all theparameters included in Model 3 excluding PS. It was observed that Model 1 and Model 3 give the highestlevel of accuracy compared to Model 2 and Model 4. The parametric study of the model parameters showthat V_(85) decreases with an increase in the accident rate, ADT, SN, and IRI. Similarly, an increase in SWand SHW resulted in a higher V85. It is expected that developed NN models would be an effective tool forthe Oklahoma Department of Transportation and other DOTs to enhance traffic safety.