Towards intelligent spatial decision support for policy-making

Abstract

In their daily workload, policy and decision makers located in governments and organisations throughout the world need to rapidly assess complex situations where numerous factors may affect the current decision scenario. If any of these factors is geographically-referenced over the study area, the decision problem to be evaluated is said to be a spatial decision problem.

Spatial Decision Support Systems (SDSSs) are the computing solution to approach spatial decision problems. However, most SDSSs are designed as case-specific, deterministic methodologies, so there is an increasing demand for generic SDSSs able to (i) approach a wide range of spatial decision domains, (ii) perform comparative studies between different SDSSs and (iii) represent the elements that are frequently related to real-case spatial decision problems namely: data, expert knowledge, actors, goals, alternatives as well as the uncertainty related to them. To fill this task, this research takes advantage of a specialised rule-based expert system, named RIMER+.

In real-case spatial decision scenarios, data and expert knowledge are usually voluminous and collected from different sources, so situations of inconsistency and incompleteness are likely to arise, especially in data-driven rule-based decision models. This research proposes an efficient and effective algorithm to deal with these issues, so results should be more accurate and realistic.

It is also important to note that in spatial decision problems, geographic information is another crucial aspect to consider since it can help modelling the problem in more detail and obtain better predictions by studying the spatial relationships among different elements. This research proposes mechanisms to automatically analyse and integrate it within the decision-making process, instead of the common practice of just visually evaluate the spatial patterns of data. Moreover, the proposed model aggregates the information available in a rational way, using state-of-the-art inference algorithms to produce decision results.

To demonstrate the value of the methodologies presented in this thesis in terms of flexibility, applicability, visualisation, accuracy and time complexity, a generic, stand-alone SDSS software prototype, named Geo-RIMER+, was designed, implemented and evaluated.

Date of Award	Jan 2014
Original language	English
Supervisor	Jun Liu (Supervisor), Hui Wang (Supervisor) & Anil Kashyap (Supervisor)