Decoding Informal Settlements in Core Urban Areas of Karachi: Leveraging Machine Learning Algorithms for Classification and Analysis

Abstract

The proliferation of informal settlements in developing countries marks a significant byproduct of unchecked urbanization and economic expansion, posing substantial sustainability challenges within urban systems. This complexity stresses the urgency of dissecting the nature and forces associated with such settlements to forge effective intervention strategies. Focused on Karachi’s primary urban sectors, this research enlightens the dynamics of informal settlements and their contributing factors. By utilizing published public datasets, the study evaluates the efficacy of five machine learning algorithms—K Nearest Neighbors (KNN), Neural Networks (NN), Random Forest (RF), Random Trees (RT), and XGBoost Tree—in predictive modelling of the spatial patterns and associated elements of these settlements. Random Forest distinguished itself among the assessed algorithms by delivering unparalleled precision across critical performance metrics, reaching an F1-Score of 0.80. This investigation further illuminates the critical role of several determinants, such as proximity to the central business district (CBD), railway lines, waterways, commercial zones, health facilities, educational institutions, and poverty markers, in accumulating informal settlements. The insights from this study are instrumental in predictive modeling for informed urban planning and policymaking, aiming to develop a systematic resolution of the challenges posed by informal settlements in Karachi.