Flooding has been a major causative aimed at social and also economic losses, and also human life’s loss. Flood flow’s accurate forecasting is a vital necessity to decrement the flooding’s risk and is essential to plan and manage the water resources systems, however, due to water flood stage analysis that is an intricate dynamic procedure characterized by temporal and spatial differences it remains to be quite challenging. Aimed at overcoming the challenge, the work has proposed a two-mode operation i.e. Data Visualization (DV) and Data Analysis (DA) for forecasting flood aimed at maintaining an accurate prediction. The main motive of the proposed framework is to forecast the flood in location wise by using the location features. Initially, DV is processed, which is initiated with the data’s quality enhancement for decrementing the error. In aid to fully utilize the database with important data, this work has initiated feature extraction (FE), and thereafter, the features are clustered into a group aimed at evading data misallocation utilizing the levy flight K-means clustering (LF-K-Means Clustering) technique. The clustered data’s visualization is executed utilizing Gaussian Kernel-Adaptive Neuro-Fuzzy Interface System (GK-ANFIS). After that, DA is executed and is initiated with data’s preprocessing followed by flood forecasting (FF) utilizing Conjugate gradient deep neural network (CGDNN) centred on K-Fold cross-validation train test split. Experiential outcomes exhibited that the framework proposed yielded 96.66% prediction accuracy and outshined the existent top-notch method.
Read full abstract