Overlap Function Based Fuzzified Aquatic Behaviour Information Extracted Tsunami Prediction Model

Abstract

Past natural hazards have produced numerous biological and physical indicators that can be used to predict similar instances in the future. These indicators can be sensed dynamically underwater or on land to generate real time alerts. This article proposes the first validated fuzzified system to predict tsunamis (FABETP) using an overlap-based algorithm. This proposed algorithm can predict seismicity based on underwater marine animal's anomalous behavior, characterized and implemented as biological indicators (i.e., aquatic animal behavioral attributes). Relevant information is extracted from these attributes and used to design fuzzy rules that generate opinion-based alerts. More precisely, the proposed algorithm, Overlap-based Fuzzified rated Marine Behavior, (OBF_MB), derives alert rules when executed on a sea turtle behavior dataset obtained from an online repository. The deployed underwater sensor-collected dataset includes the following measurements: induced electromagnetic field, undersea turtle count, and angle of deviation (in terms of the turtles' navigation direction formulated per month and per day). These values are used as the inputs to the proposed system. To generate an opinion, an information gain-based opinion score is used to calculate the opinion deviations from the generated opinions of the default rule. For future data values, 2004 is used here as the default opinion year and the scenarios is the default rule. This paper formulates three classes of opinions using the proposed algorithm: Alert, Pre-Alert and No-Alert. These opinions can be used in the future to generate real-time alerts based on aquatic animal behavior.