Ensemble learning using fast rule based fuzzy K –means pre clustering and classification for aquatic behavior-extracted tsunami prediction

Abstract Underwater seismic reactions have been pre-sensed by various mathematical analytical methods. Various warning systems that have deployed have sometimes proven to be effective. Tsunami-underwater seismicity is still considered as one of the major fears among mankind. In this paper we propose a Tsunami alert generation model based on Ensemble Clustering and Classification TPbEC2 for generating alerts. This model uses turtle behavioral dataset for classification of alert and No Alert. The model shown here works in two parts as: fast pre-clustering has been proposed based on Fuzzy K means modification followed by ensemble with FRBCS (Fuzzy Rule based Classification System). The first part of model method is validated in comparison to four base clustering methods with three indices. Similarly the Ensemble of MFKmeansIC along with base clustering methods (in total 25) with all five base FRBCS’s is evaluated for accuracy. To further validate, statistical test- Friedman is also performed for best ensembles from above accuracy matrix where MFKmeansIC+ FH.GBML shows the best rank.