A hybrid model of partial least squares and neural network for traffic incident detection
Expert Systems with Applications: An International Journal
| Hi-index | 0.00 |
Most automated expressway incident detection algorithms were successfully developed using loop-based traffic occupancy from their local conditions. However, the performance of these algorithms was not satisfactory on sites that have installed a video-based detector system. Due to different traffic detector technologies and varying driving behaviors from one region to another, it is of interest to develop an algorithm that uses video-based data. This paper used a total of 160 incidents collected along the 15-km central expressway (CTE) in Singapore to develop two new dual-station algorithms: the combined detector evaluation (CODE) and the flow-based CODE algorithms. On average, the flow-based CODE algorithm yielded better performance than the CODE in terms of average reduced false alarms of about 16%. Measures were also taken to ensure that the algorithms were properly developed and assessed. It was found that the CODE algorithm can detect, on average, up to 57% of the incidents faster than those of existing detection methods on CTE.