Predicting Floods with Deep Learning Neural Networks and Analyzing Their Socioeconomic Impacts with OLS Regressionn

The ability to predict flood occurrences and gauge their socio-economic impacts is highly essential for flood management.This research collated one of the largest datasets of historical floods (4000+ entries) ever used in Australian flood research and mapped flood locations to raster datasets of hydrological flooding factors created using ArcGIS software.The resulting dataset underwent unsupervised clustering to display statistical significance and was used to train a Deep Learning Neural Network with an 85-90% accuracy in predicting flood likelihoods given coordinate inputs.A K-nearest-neighbors optimisation algorithm was used to determine the most relevant hydrological factors in flood prediction.Seeking to address current research gaps in estimating the societal impacts of floods, we merged the historical floods dataset with Australian census data and performed Ordinary Least Squares regression, accounting for time & space fixed effects through categorical regression variables.We discovered that floods disproportionately impact Indigenous Australians as flooded areas see a disproportionate increase in Indigenous Australian population, hinting towards a dangerous possibility of environmental redlining.This research also quantified how floods affect select census statistics such as household income, unemployment, and migration.This research's regression model can be extended to gauge the impact of floods on any census statistic.