Natural disaster prediction optimized light gradient boosting enabled hybrid convolutional neural network and bidirectional long short-term memory

The prediction of a natural disaster involves developing tools to forecast the timing, strength, and effect of sudden violent event such as earthquakes, hurricanes, floods, and tsunamis. Existing models are often confronted by challenges such as lack of data availability, uncertainty about the processes involved, and the complexity of how environmental factors interact with each other. Therefore this new, more advanced model Proposed Canary Timber Fox Optimization-based light Gradient boosting Machine-enabled hybrid Convolutional Neural Network and Bidirectional Long Short-Term Memory (CTFO-light GBM-enabled CNN-BiLSTM) intends to address all those problems faced by the existing models. CFTO represents the model with adaptive and collaborative features so that this strategy is similar to natural survival tactics. Hence, it can efficiently react to dynamic conditions in an environment. The use of a light GBM in the model facilitates the efficiency and speed of the model which is critical for handling large datasets containing high-dimensional features leading to higher prediction accuracy through iterative learning. The CNN block allows the model to excavate intricate spatiotemporal features from different environmental data sources, thus expressing the complex patterns that are essential for prediction. Also, the BiLSTM architecture is designed as bidirectional which means that the model can process the data sequentially in both directions always retaining the information from both previous and future time steps thus successfully learning the long-term dependencies and temporal dynamics present in natural disaster occurrences. This incorporation enables the model to be adaptable, efficient, and capable of feature extraction as well as time comprehension. Similarly, based on achievements employing k-fold 8 leads to a model by a lower Mean Square Error (MSE) of 3.36 and enhanced accuracy of 94.01%, sensitivity of 92.27%, and specificity of 95.74% respectively.