Applying Machine Learning to Understand Water Access Inequality in Underserved Colonias Communities

Keywords: machine learning, clustering, classification, water accessibility, underserved communities
Session Type: Virtual Paper Abstract
Day: Monday
Session Start / End Time: 2/28/2022 11:20 AM (Eastern Time (US & Canada)) - 2/28/2022 12:40 PM (Eastern Time (US & Canada))
Room: Virtual 28

Abstract

This paper introduces our research in applying machine learning to understand the water accessibility issues in the underserved communities called colonias. These are communities of people who live near the US-Mexico border and have poor living conditions in terms of access to basic infrastructure and utilities. We analyze nearly 2000 such colonias communities using data from RCAP (Rural Community Assistance Partnership) and apply an adaptive affinity propagation algorithm to automatically group colonias into different clusters. Gower distance is introduced to make the algorithm capable of processing both categorical and numerical attributes. We then go one step further to apply a decision tree analysis algorithm to understand the driving factors that make these clusters similar and identify common characteristics in each colonias cluster. This machine learning pipeline allows us to gain a deeper insight on the water access and quality issues in these underserved communities, contributing therefore to the future development of the coupled social, physical and cyberinfrastructure to address severe water issues. Our methodology also offers a new way of data-driven analysis which combines cluster and classification analysis together to gain a more comprehensive view and a previously uncovered perspective on domain problems beyond exploratory analysis.