Please use this identifier to cite or link to this item:
|Title:||GIS-Based Multi-Hazard Susceptibility Assessment using Machine Learning|
Verma, A. K.
|Keywords:||Remote Sensing;Geographic Information System;Hazard Susceptibility Mapping;Machine Learning;Particle Swarm Optimization;Evolutionary Optimization|
|Abstract:||The present century has been the most unfortunate in the context of the casualties and losses caused on account of the disasters induced by natural hazards globally. Words fall short in describing the scale of devastation and havoc brought on Indian soils by floods, landslides, forest fires, avalanches, etc. while the mankind remained a mute spectator in such crucial times. All the while the advancements in science and technology has coined terms such as machine learning and artificial intelligence in its miraculous projectile of evolution. These sciences have empowered us to be better prepared for handling aforementioned adversities caused by natural disasters through indigenous techniques of hazard susceptibility prediction and assessments. These efforts have been enhanced by an incandescent flourishing of technologies like Remote Sensing and Geographic Information Systems. This thesis is an attempt in this direction of predicting a region’s susceptibility towards a specific hazard of which the given region has been a victim of. This task involves the use of satellite images to extract characteristic details of the stretch of land under study. The aforementioned specific characteristics are the causative factors which play a role in the unfolding of the specific hazard. The spatial coordinates of the locations which in the past have witnessed the disasters on various scales are mapped with their corresponding values of their causative factors/traits. This problem of mapping is reduced to that of mere pattern detection by identifying the traits and their values that have played a pivotal role in the past episodes of disaster and predicting the susceptibility of any such events in future. This is where the expertise attained in the field of machine learning steps in to rescue. The susceptibility for the region is interpolated for the entire expanse of the region selected, to delineate the regions in terms of their susceptibility on the scale of very highly susceptible to least susceptibility. The accuracy in delineations has been enhanced by virtue of state-of-the-art optimizations and ensembling of a variety of machine learning models. The final outcome is a hazard susceptibility map describing the corresponding region’s achieved delineations that could aid in framing a contingency plan, along with, strengthening the efforts of disaster mitigation and management.|
|Description:||Master of Engineering -CSE|
|Appears in Collections:||Masters Theses@CSED|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.