基于历史命中率的时空重排扫描最大搜索半径选取方法及应用实验

Method and Application Experiment for Selecting Maximum Search Radius of Space-Time Permutation Scan Based on Historical Hit Ratio

最大搜索半径是控制时空重排扫描统计方法性能的关键参数,而搜索半径过大会造成资源浪费。为此,该文提出了一种基于历史命中率的时空重排扫描最大搜索半径选取方法。通过迭代不同最大搜索半径对历史性时空数据集进行时空重排扫描统计,根据不同最大搜索半径所得扫描结果计算预警命中率,选取命中率最高的最大搜索半径作为前瞻性扫描的最大搜索半径,避免了因最大搜索半径过大导致的资源浪费,提升了模型的性能;基于旧金山地区火灾数据集的验证结果证明了该方法的有效性。

The maximum search radius is a key parameter to control the performance of the space-time permutation scanning statistical method.But the oversized search radius will lead to a resource loss when the study area is too large.Therefore,this paper proposed a method to select the maximum search radius of space-time permutation scan based on the historical hit ratio.Firstly,the historical space-time dataset was scanned with the space-time permutation scanning statistical method by iterating different maximum search radii.Then,early warning hit ratios were calculated with the scan results.At last,the largest search radius with the highest hit ratio was selected as the maximum search radius for the prospective scans.This method avoided the resource waste caused by the oversized search radius and improved the performance of the model.This method was validated according to the fire dataset in San Francisco area.