Mining Underground Alert Signals for Seismic Detection using Wireless Sensor Nodes

Since ages scientists are trying to predict Seismic disturbances such as Earthquake, Tsunami based on physical parameters determined by properties of a seismic cycle. The belief that strange animal behaviour can predict seismic disturbances has been a belief amongst researchers. This paper proposes a novel conjunct framework for detecting underwater and underground seismic activity based on impact of physical parameters identified on animal behaviour. The framework is further implemented for observed differences in soil animal behaviour based on dataset collected using wireless sensor nodes deployed in active seismic areas of Netherlands. The implementation is a data mining technique to generate real time alert where the technique makes use of identified parameters that defines the soil and underground condition in terms of magnetic field, CO2 concentration, soil plasticity, peroxide as well salt level and sea bed silt concentration respectively. The implementation method filters the alert value of parameters based on threshold identified using decision tree method. In reference to the analysis presented here, we propose a novel algorithm for pre-seismic condition prediction using Earthworm as biological sensor dataset. The dataset used for mining the information has been retrieved from public free resource which on 10 cross fold validation gives an error of 0.047 % with a time of 0.03 seconds required to build a test model.