Probabilistic Agent-Based Modelling for Predicting School Attendance

在英国，16岁的孩子在完成考试后，必须再接受两年的教育、就业和培训。然而，大量儿童没有参加他们义务教育的最后阶段，成为“NEET”(不在教育、就业或培训中)。有许多事件可能使一个人在16岁时成为NEET，这些事件通常发生在非常年轻的年龄，所以确定最重要的原因是极其复杂的。因此，设计旨在识别并在理想情况下防止这些负面影响的政策可能非常具有挑战性。在对导致NEET结果的个人和系统进行建模方面，一种可能有希望的方法是基于代理的建模(ABM)。ABM的特点是直接代表个人的模型，模拟他们随着时间的推移的行为和行动。因此，这种方法可能是探索在很长一段时间内(从出生到16岁)所有影响年轻人的事件、人和系统的理想方法，目的是确定可能与在以后的生活中成为NEET相关的关键事件或行为。这样的预测本质上是非常不确定的，所以这个项目还将首次开发一个基于概率代理的模型，以概率的方式表示模型元素(变量、函数、输出等)，直接捕捉系统中的不确定性。该项目将设在利兹大学，附属于沃尔夫森应用健康研究中心。作为该中心的一部分，学生将能够访问世界领先的假匿名安全数据集，其中包含该区域所有学童的记录，其中包含丰富的信息，这些信息对于建立和验证基于代理的模型至关重要。此外，该项目将与当地政府的专家合作，确保项目的设计、实施和成果能够直接用于制定有可能改善许多年轻人生活的政策。重要的是，该项目将探索导致儿童成为NEET的一般因素；通过研究将不会确定真正的儿童。

In the UK, after finishing their exams at age 16 children must undertake a further two years of education, employment and training. However, large numbers of children do not participate in this final stage in their compulsory education, becoming 'NEET' (not in education, employment, or training). There are a wide range of events that may predispose someone to becoming NEET at 16, and these can often occur at a very young age, so identifying the most important causes is extremely complicated. As such, designing policies that are aimed at identifying and, ideally, preventing these negative influences can be extremely challenging.A method that might hold promise in modelling the individuals and systems that cause NEET outcomes is that of Agent-Based Modelling (ABM). ABM is characterised by models that represent individuals directly, simulating their behaviour and actions over time. Hence such a method might be ideal for exploring the events, people, and systems that all influence a young person over a long time period (from birth to age 16) with the aim of identifying pivotal events or behaviours that may correlate with becoming NEET in later life. Such predictions are inherently extremely uncertain, so this project will also develop, for the first time, a probabilistic agent-based model that represents model elements (variables, functions, outputs, etc) in a probabilistic way that captures the uncertainty in the system directly.The project will be based at the University of Leeds and affiliated with the Wolfson Centre for Applied Health Research. As part of the Centre the student will have access to a world-leading, pseudo-anonymised, secure dataset that contains records for all school children in the region with rich information that will be essential for building and validating the agent-based model. In addition, the project will collaborate with experts in local government to ensure that the project design, implementation and results can be of direct use for developing policies that will have the potential to improve the lives of many young people. Importantly, the project will explore the factors that lead to children becoming NEET in general; no real children will be identified through the research.