AN ARTIFICIAL NEURAL NETWORK MODEL FOR INCIDENT DETECTION ON MAJOR ARTERIAL STREETS

摘要

This study attempts to develop an arterial incident detection model by applying anArtificial Neural Network (ANN) with simulation data. A section of the US-1 corridor in Miami-Dade County, Florida was selected as the study area and coded in the CORSIM microscopicsimulation model. Two data sets were generated via CORSIM simulation for modeldevelopment and assessment. Multiple ANN models were designed for various scenarios. Themodel performances were evaluated using the selected measures of effectiveness (MOE),including detection rate (DR) and false alarm rate (FAR). The results showed that the ANNmodels in general could detect arterial incidents with a high DR of 90-95% and an acceptableFAR of lower than 4%. The study also identified some preferred features in the design of ANNincident detection models for this application. These include the detector configuration scheme,the selection of model input features, and the employment of data from previous cycles.