An Intelligent ResNet-152 CNN framework for Anticipatory Flood Forecasting

The occurrence of flood in a region is very repetitive and which causes human loss, materialistic loss and extreme damage of property and infrastructure. This is mainly due to the overflow of the river and improper structure of the drainage system. This impacts not only on human life but also on nation’s economy, property and infrastructure. Exposure to flood lead to stress, anxiety among residents and it even affects their mental health. Hence the early warning of flood, will help the residents and higher authority to be in alert and take necessary action to prevent the adverse nature of the flood. Many machine learning algorithm helps to provide the early warning to any disaster in which random forests, convolutional neural networks (CNNs), decision trees, and logistic regression algorithm outperforms and which has a struggle to capture the complex weather patterns. A novel application of the ResNet152 CNN algorithm for flood prediction shows promise in overcoming these challenges by providing more accurate and scalable models. The proposed algorithm provides 95% of accuracy in predicting the rainfall intensity and frequency in the chosen study area from the year 2024-2026. These algorithm can be used to predict the intensity of the rainfall and the frequency effectively even for the existing complex whether pattern.