290008 PS Geographical Predictive Analytics (2024S)

Entry requirements
Knowledge of basic Python scripting is a prerequisite for this course. Knowledge of descriptive statistics or spatial statistics would be advantageous but is not required.

Aims
Geographical predictive analytics involves the creation, testing, and validation of models (i.e. abstractions of reality) for the purpose of drawing conclusions and testing hypotheses in relation to your data as well as making estimations about future and unknown samples. Some of the application domains are soil mapping, ecological mapping, crime forecasting, and geohealth (just to name a few). The types of models that are used in the class are predictive models and descriptive models. Descriptive models are used to generalize/aggregate raw geodata into groups for which extracted information becomes then meaningful and easier to interpret. Furthermore, these models can be utilized for or to develop predictive models. The predictive models, on the other hand, are used to estimate samples’ values at unknown locations and places or to forecast values of future events. The techniques that the course focuses on derive from the spatial statistics and machine learning domains, while it is important to outline the strengths and limitations of both. The geodata applications involve mainly clustering, classification, and regression tasks.

Content
Introduction to the course and its tools, location clustering, attributes clustering, regionalization, spatial regression, geographically weighted regression, decision trees, random forest, K-Nearest Neighbor, modelling workflow, model validation, and performance evaluation metrics.

Methods
The course has a fairly equal amount of both lectures and practical work, in which theory and methods precede their application. The practical work consists of exercises of multiple tasks to be solved/answered given sufficient support material. The practical work will be performed using python (geo)libraries and jupyter notebooks. During the course, students undertake an individual case study that is assisted by the practical work material. Strong emphasis is given on a) the identification of innovative parts of analysis and models, b) the critical reflection of study results, and c) scientific reproducibility.