The utilization of artificial neural networks (ANN) in developing an accident prediction model for Commonwealth Avenue

ABSTRACT: This study developed an accident prediction model for Commonwealth Avenue utilizing Artificial Neural Network (ANN). It aimed to determine the number of accidents in succeeding years for September 1, from Elliptical Road to Central Avenue; Segment 2, from Tandang Sora Avenue to Luzon Avenue; and Segment 3, from IBP Road (Sandigan) Meralco to IBP Road Litex Market. Various parameters were considered in the study, including the vehicle type, roadway characteristics, and environmental and time factors. The result showed that the most important road parameter was the vehicle type, with an importance value of 22.13%. It was followed by road width and weekdays with 18.26% and 15.61%, respectively. Meanwhile, the historical data was segmented into three portions 70% was allocated for training, while the remaining 30% was split equally between testing and validation. The correlation coefficients (R) were obtained for the three models, with values of 0.92001 for Segment 1, 0.92925 for Segment 2, and 0.94722 for Segment 3. The Neural Time Series App in MATLAB was employed to predict the number of accidents for 2024-2028. It indicated that the forecasted numbers of accidents in Segment 1 had a mix of fluctuations and relative stability. Segment 2, on the other hand, showed a tendency to slightly decrease in 2026 and slightly increase in subsequent years. While the third segment exhibited a more erratic pattern compared to the other segments, presenting mixed fluctuations and occasional spikes. It demonstrated the most variability in accident counts with less predictable trends.

Filipiniana