Data Mining and
Machine Learning for Predicting and Managing Flooding Disasters in Coastal and Riverine
Areas
Tamilselvan K., Jena Suvendu Kumar, Prabu S., Leopauline S., Deepika J. and Srinivasan
J M.E.
Disaster Advances; Vol. 18(12); 71-79;
doi: https://doi.org/10.25303/1812da071079; (2025)
Abstract
Coastal lowlands and river floodplains still rank among the places most likely to
swallow lives and highways alike during a single storm. Scientists who track these
disasters continue to notice that climate change has nudged both the frequency and
intensity of inundations toward increasingly extreme levels. Because of that shift,
the old siren systems and water gauges no longer give people enough notice or confidence.
This research study asks whether data-mining routines and the newly fashionable
family of machine-learning tricks can fill that warning gap. Rain maps, river sticks,
hillside surveys, sod moisture snapshots and centuries of logbook commas are fed
into algorithms in hopes they will whisper a danger sign long before the water rises.
Alongside the supervised staples, Random Forests, SVMs, etc., the K-Means unsupervised
stage sifts the output to categorize neighborhoods into low, medium and high risk.
A quick-hitter feature examination, borrowed from information theory, enables engineers
to rank the flood drivers worth keeping and the noise worth ditching.
That shortlist shuffles into geo-geo-dashboards that paint crimson where tomorrow's
traffic could grind to a standstill. Field tests on Caribbean watersheds and midwestern
floodplains push overall hit rates past the 90-percent mark and give emergency crews
a map rather than a mantra. Analysis of the experimental results indicates that
the fusion of data mining techniques with advanced predictive analytics markedly
improves the readiness of emergency services, optimizes the distribution of limited
resources and bolsters the resilience of affected communities. Such work feeds into
the expanding literature on intelligent disaster-management systems and highlights,
once more, the role of machine-learning algorithms in curbing the economic and social
toll wrought by natural hazards on at-risk regions.
