Modeling the 85th percentile speed on Oklahoma two-lane rural highways via neural network approach

摘要

Traffic operations on two-lane rural highways and setting of realistic speed limits are some of the difficult tasks faced by state transportation agencies including the Oklahoma Department of Transportation (ODOT). Most traffic engineers believe that speed limits should be posted to reflect the maximum speed considered to be safe and reasonable by the majority of drivers using the roadway. Modeling traffic speeds using the classical mathematical approach does not frequently yield reliable results because of the human factors involved in driving. In recent years neural network (NN) approach has emerged as a powerful tool in solving many engineering problems where the physics of the problem are poorly understood or are not known because of the complexities involved. A neural network (backpropagation architecture) model is presented in this paper for the prediction of 85th percentile speed on Oklahoma two-lane rural highways. The model predicted the 85th percentile speed with an average degree of accuracy of about 96%.