Due to significant societal and environmental impacts, obtaining a more informed understanding of wildfire activities is always important. This work uses historical data to focus on wildfire pattern recognition, prediction, and subsequent uncertainty quantification. We propose an interpretable and flexible marked spatio-temporal point process model to accomplish the tasks and adopt recent advances in time-series conformal prediction. Through extensive real-data experiments, we demonstrate the effectiveness of our methods against competing baselines.
翻译:由于重大的社会和环境影响,对野火活动取得更知情的了解总是十分重要的,这项工作利用历史数据侧重于野火模式的识别、预测和随后的不确定性量化。我们提出了一个可解释和灵活、有标记的时空点进程模型,以完成任务,并采纳时间序列一致预测的最新进展。通过广泛的实际数据实验,我们展示了我们方法相对于相互竞争的基线的有效性。