¹û¶³Ó°Ôº

This term two module introduces state-of-the-art topics covering two broad areas: Bayesian inference (7 weeks) and Machine Learning methodologies (3 weeks) for spatial analysis applicable to the domains of social sciences, epidemiology, disaster risk reduction and geography.

In the first 3-weeks, the module provides a detailed introduction to Bayesian statistics using Stan, which is an interface to RStudio that allows state-of-the-art statistical modelling and Bayesian computation. We will introduce you to the absolute basics of Bayesian theory as well as writing your own probabilistic codes in RStudio and Stan for carrying out a broad range of multivariable models within the Bayesian framework: Generalised Linear Modelling (GLMs) and Generalized Additive Models (GAMs).

Mid-way (week 4, 5, and 6), we will then focus on Machine Learning methodologies when we have an introductory session on the various statistical learning techniques which have significant applications in image classification and pattern recognition within the social sciences and geography domain. You will be taught Deep Learning, Convolutional Neural Networks (CNNs) and GeoAI using Pytorch in Python.

Lastly, you will be shown how to formulate hierarchical, and Conditional Autoregression (CARs) models within a Bayesian framework for spatial risk prediction and uncertainty using exceedance probabilities, a technique which has significant applications to many fields such as spatial epidemiology, social sciences, or disaster risk reduction and many more.

The module will equip the students with the foundations in advanced data analytics and prepare them to tackle future research challenges in the domains of geography & social science.

By the end of the module, the students should:

• Possess a solid foundation and advanced knowledge on key principles of statistical learning and Bayesian inference;
• Have ability to apply state-of-the-art statistical modelling and machine learning techniques on social and geographic data science problems;
• Be able to perform inferential statistics on spatial data and carryout hypothesis testing for evidence-based research using the different types of regression-based models from a Bayesian framework;
• Be able to perform statistical learning for image classification and recognition using various machine learning algorithms.
• Acquire new programming language skills such as Stan (interfaced with RStudio) and Pytorch (interfaced with Python).

The course will consist of 10 lectures and 10 practical sessions.