Author(s): Yu-Zhao Xie; Xiang-Ju Cheng; Ze-Hai Chen
Linked Author(s): Xiangju Cheng
Keywords: Tidal river drifting trajectory random forest
Abstract: As the core of Guangzhou's waters, the former channel of the Pearl River carries multiple functions such as commerce, culture, tourism, etc. Behind its prosperity is often plagued by drowning incidents. The former channel of the Pearl River is the tidal river, which has complex hydraulic conditions. It is difficult for the police to determine the specific location of the drowning victim fell into the water. Consequently, we used a dummy drift simulation experiment and a random forest model to realize the traceability of the drift trajectories of bodies in the former channel of the Pearl River and help the police determine the location of the drowned person in the water. The results show that the R-square of the random forest model in the prediction of the x-coordinate and y-coordinate of the overboard position reaches 0.997 and 0.981, respectively, and the proportion of samples in which the model predicts the position and the actual position with the distance error of less than 1km is more than 99%, which achieves a satisfactory prediction accuracy. Our model can narrow down the range of most drowning victims to 1km, effectively reduce the workload of the police in searching for clues, and improve the efficiency of handling cases. The construction method of the traceability model is also applicable to other tide-sensitive river sections, which can make the drifting trajectory of the corpse measurable, and provide convenience and reference for the corpse salvage work and the police's case processing.
Year: 2025