Bayesian inference and prediction in complex social systems

在许多系统中，复杂的行为源于局部相互作用的代理之间的相互作用，从而产生一种涌现现象。复杂系统在人类行为中普遍存在;有组织犯罪网络和城市零售结构是两个具有实际意义的例子。传统的分析方法是有限的，当试图模拟这种行为，和新的方法需要开发预测模型。这项工作主要是在复杂的，典型的多尺度，系统的推理和预测的新的计算方法的调查感兴趣。目前的机器学习和数据科学的兴趣已经产生了广泛的印象，这些方法能够解决大多数问题，而不需要科学调查。同样，尽管数学建模和计算科学取得了巨大进步，但复杂系统的模型往往具有有限的预测能力。在这项工作中，我们探讨机械和概率数学模型，并期待在这些模型中的数据同化技术和推理。我们探索犯罪行为和城市零售结构的模型，并使用真实世界的数据估计参数值。此外，我们感兴趣的新方法，使贝叶斯推理在复杂系统中可行。

Complex behaviour in many systems arise from interactions between locally interacting agents, resulting in an emergent phenomenon. Complex systems are prevalent in human behaviour; two examples of practical interest are organised criminal networks and urban retail structure. Traditional analytical approaches are limited when trying to model such behaviour, and novel approaches are required to develop predictive models. This work is largely interested in the investigation of new computational approaches for inference and prediction in complex, typically multiscale, systems.The current interest in machine learning and data science has generated the widespread impression that such methods are capable of solving most problems, without the need for scientific inquiry. Likewise, despite enormous advances in mathematical modelling and computational science, models of complex systems tend to have limited predictive capability. In this work we explore mechanistic and probabilistic mathematical models and look towards data assimilation techniques and inference in these models. We explore models of criminal behaviour and urban retail structure and estimate the parameter values using real-world data. Furthermore, we are interested in new methods that make Bayesian inference feasible in complex systems.