A Novel Hybrid Supervised Machine Learning Model for Real Time Risk Assessment of Floods using Concepts of Big Data
Amongst various natural catastrophes, flood is one of the most destructive natural disasters, which has severe impact over the economy of a country and its people. Modelling and analyzing the flood is a complex task. The advancement in technologies helps to predict the impact of floods through an essential metric, namely, risk assessment, which drastically reduces the loss of human life,property damage,through preventive measures taken well in advance. Risk assessment modelling refers to combinatorial development of identification and assessment of potential for occurrence of an event which causes a negative impactonanentityofinterest.Withrecentadvancesindataacquisitionandarchivalmethods, concepts of big data have been a great boon to risk assessment development. It is primarily due to the fact that, the accuracy of risk assessment(RA) relies on the volume of historical data analysed. Based on this, a risk assessment model is designed as a hybrid model using differentialevolutionandadaptiveneurofuzzyinferencesystemtoassessriskinrealtime.The performance ability of proposed hybrid model is compared with conventional ANFIS and neural network models by analysing the rainfall status in India. Based on the monthly, annual and seasonal data,theriskassessmentisperformedthroughvariousfactorslikewetnessindex, land slope angle, stream power, stream density, rainfall, curvature and distance. Data from the expert systems are collected by analysing various case study areas from India to validate the performance of proposed hybrid system.
Flood risk assessment, Prediction models, Big Data, Machine learning, Hybridization, Prediction accuracy