Performance of artificial neural networks for incident detection in ITS

In this paper we introduce a new incident detection model based on a modified form of the Probabilistic Neural Network (PNN) that utilizes the concept of statistical distance. Results using both simulation and real incident data demonstrate its competitiveness with the Multi Layer Feed Forward (MLF) architecture which was found in previous studies to yield superior incident detection performance. The PNN performance was competitive with the MLF in terms of Detection Rate (DR), False Alarm Rate (FAR), and average Time To Detection (TTD). It is also capable of learning in real time. Thus, the PNN may be transferable without the need for retraining in the new site as its performance is enhanced with time in service. Moreover, the PNN is also very flexible to fit varying requirements in different locations and in different times of the day by dynamically varying simple parameters.