Floods are natural disasters that often occur and are among the most destructive because they have significant economic and social impacts. Accurate flood predictions are essential to manage risk and organize emergency response planning effectively. This research uses Deep Neural Network (DNN) to build a flood forecasting model that relies on rainfall index indicators and captures complex and ever-changing patterns obtained from rainfall index data. Using historical information from flood disaster events in Kerala, India, an analysis was conducted to assess the impact of various factors, particularly in learning rate and optimizer type, on model performance. The experimental results show that the type of optimizer is a crucial factor in determining the model's effectiveness, as shown in the ANOVA statistics with a P-value of 0.008493, much lower than the general threshold of 0.05. This is because this type of optimizer can significantly improve prediction accuracy. With the Adam optimizer type, the learning rate range is between 0.1 and 0.4, showing an accuracy level of up to 100%. However, the choice of learning rate does not significantly impact, indicating that the main emphasis on parameter adjustment should be determined accurately. Therefore, by carrying out appropriate parameter adjustments and thorough validation to find the optimal configuration that can increase accuracy in predicting flood disasters based on rainfall indices, the DNN model has the potential to become a tool that can assist in flood risk planning and management.
Copyrights © 2024