Learning Probabilistic Simulation Models for Rare Event/Condition Occurrence

过去，人们研究了从给定数据和背景知识中学习概率模型的各种方法，但对于稀有/特殊情况下的概率模型学习的研究非常有限。在这项研究中，我们开发了一种有效而准确的方法，通过使用给定的数据集及其相关的背景知识来学习罕见/特殊条件下的概率模拟模型。此外，通过对一次罕见和大规模自然灾害的应用，我们展示了一个新的框架，用于对罕见/特殊事件和情景的概率估计和预测。

Various approaches for learning probabilistic models from given data and background knowledge have been studied in the past, however, studies on the probabilistic model learning for rare/special conditions have been very limited. In this study, we developed an efficient and accurate approach to learn probabilistic simulation models for the rare/special conditions by using a given data set and its associated background knowledge. Moreover, we demonstrated a novel framework for the probabilistic estimation and prediction of rare/special events and scenarios through its applications to a rare and large scale natural disaster.