MurutiFelix/Predictive-Model-for-power-theft-detection-in-Kisauni-Sub-County-Mombasa
Domain:
geospatialRecord type:
project
Creator:
Mur
Host:
#-- BUILDING A PREDICTIVE MODEL FOR POWER THEFT HOTSPOTS USING GIS AND MACHINE LEARNING IN KISAUNI SUB COUNTY MOMBASA
# 1. Introduction
Electricity theft presents a significant challenge to power utilities in developing countries, resulting in revenue losses, operational inefficiencies, and reduced service reliability. Conventional electricity theft detection methods are largely reactive, relying on manual inspections and rule-based auditing, which are often costly and spatially inefficient.
Advances in Geographic Information Systems (GIS), Earth Observation data, and Machine Learning (ML) techniques provide an opportunity to enhance electricity theft detection through spatially informed, data-driven approaches. Integrating spatial infrastructure data with billing and anomaly records enables the identification of high-risk areas and supports proactive decision-making.
This project develops a spatially explicit machine learning framework for analysing and predicting electricity theft hotspots within Kisauni Sub-County, Kenya.
---
# 2. Research Objectives
The main objective of this study is to develop and evaluate a machine learning–based spatial model for predicting electricity theft hotspots.
The specific objectives are to:
1. Analyse the spatial and temporal distribution of electricity theft incidents.
2. Integrate geospatial infrastructure and billing data to derive predictive features.
3. Develop and train machine learning models for electricity theft prediction.
4. Visualize predicted theft hotspots using GIS-based mapping techniques.
---
# 3. Research Gaps
Despite increasing application of machine learning in utility management, existing approaches exhibit several limitations:
- Limited integration of spatial context in electricity theft detection models.
- Minimal use of GIS-based proximity and density analyses.
- Lack of spatial decision-support outputs for operational planning.
This study addresses these gaps by combining spatial analysis, mac …