IoT-based flood disaster early detection system using hybrid fuzzy logic and neural networks

A flood stands as one of the most common natural occurrences, often resulting
in substantial financial losses to property and possessions, as well as affecting
human lives adversely. Implementing measures to prevent such floods becomes
crucial, offering inhabitants ample time to evacuate vulnerable areas before flood
events occur. In addressing the flood issue, numerous scholars have put forth
various solutions, such as the development of fuzzy system models and the establishment of suitable infrastructure. However, when applying a fuzzy system,
it often results in a loss of interpretability of the fuzzy rules. To address this
issue effectively, we propose to reframe the optimization problem by incorporating stage costs alongside the terminal cost. Results show the proposed model
called hybrid fuzzy logic and neural networks (NNs) can mitigate the loss of
interpretability. Results also show that the proposed method was employed in
a flood early detection system aligned with integrating into Twitter social media. The proposed concepts are validated through case studies, showcasing their
effectiveness in tasks such as XOR-classification problems.