Detection of vulnerable communities in East Africa via novel data streams and dynamic stochastic block models
Publication Date
December 8, 2021
Creator
Abstract
In developing countries it is challenging to collect data on poverty and its associated community health characteristics. Data collection in this context is impractically laborious and resource greedy. Additionally due to the sensitive nature of these themes the data is often unreliable. There is a need for alternative methods of detection of vulnerable communities. However, promising opportunities arise via novel rich data streams such as Call Data Records stemming from the ubiquitous use of mobile phones. Despite the growth of Call Data Record data there has been limited previous application to problems of poverty and development. This thesis makes three main contributions: (i) Methods of collecting ground truth data in Developing areas; (ii) Best practices in application to detect vulnerable regions; (iii) Development of new applications of statistical approaches to the problem via the stochastic block model. This work is focused on Dar es Salaam in Tanzania. Having more reliable and easily accessible truths on these vulnerabilities can have a high potential impact for policy makers and NGOs trying to make positive changes to reduce devastating effects of poverty. This thesis produces comprehensive results to amend the current knowledge gaps, via rigorous fine-grained data collection processes surveying the 452 subwards in Dar es Salaam in relation to poverty and social vulnerability.
Item Type
ethesis
Thesis Type
PhD
Supervisors
Subjects (LC)
Associated Schools / Departments
Nottingham University Business School (UK)
eprints ID
67041
UoN Repository URI
Except where otherwise noted, this item's license is described as
File(s)![Thumbnail Image]()
Name
Maddy_Thesis (2).pdf
Type
Full-text
Description
Examined
Size
89.14 MB
Format
Adobe PDF
Checksum (MD5)
2bcb87778baaa1bc34686b02310bf7ca