A Wavelet‐LSTM With Attention Framework for Adaptive Oil Spill Detection in Smart Ocean Systems Using IoT and Adam Optimization

Oil spills endanger the environment and require immediate detection and proper technology for evaluation and containment. These have caused significant drawbacks to conventional detection methods in terms of resources, time, and especially limitations in scale over the extent of oceanic areas. Most conventional procedures include disturbance and aerial surveillance methods with very long‐time frames for getting evidence and broad resource allocations. Such anomalies respond very late and cover relatively small areas, making them very ineffective in spill monitoring. In this study, we present a new adaptive oil spill detection framework, Wavelet‐ Long Short ‐Term Memory (LSTM) with Attention, for smart ocean systems using the Internet of Things (IOT) and real‐time satellite data. The model involves a wavelet transform to reduce noise and extract features, an LSTM so as to model the spatiotemporal data of the environment, and an attention mechanism that will help tend toward the critical features. The whole system is optimized using the Adam algorithm, which ensures efficient learning and quicker convergence. On the basis of the results of the experiment, the model turns out to be highly efficient in terms of attaining accuracy with the following figures: 98.95%, precision: 98.45%, recall: 98.52%, and F1 score: 98.00%. Additionally, the present study confirms the efficiency of the model in terms of several other assessment criteria, namely, ROC and Precision‐Recall curves, which confirm good generalization and low rates of misclassification. Therefore, the proposed system can be employed for the purpose of autonomous, scalable, and efficient monitoring of environment protection and disaster management in marine ecosystems. The suggested approach is known as the Wavelet Transform, LSTM and Attention Mechanism.