Clustering Based Sampling for Learning from Unbalanced Seismic Data Set

Thisarticledescribeshowsomestratumcontainastressconcentrationzones,andwhilethestress increases andexceedsahighvalueor socalledcriticalvalue, it destroys rocks.This causes the emissionofseismictremorsofdifferentenergies.Seismologyconsistsofthestudyoftheeffectsof seismicwaves,andpredictingtheseismichazardstorocksandlongwallcoals.Thisisalongsidethe mainproblemoccurredinthisfield,theunbalanceddatathatlackscausewhenstudyingtheseismic hazards.Learningfromunbalanceddataisconsideredasoneofthemostdifficultissuestosolve nowadays,thisarticlepresentsaninformedsamplingmethodthatisbasedonaclusteringapproach forthepredictionofseismichazardsinPolishcoalmines.Theideaisbasedonthedividingofnonhazardousexampleswhichrepresentsmorethan90%ofthereal-lifecasesintosubsetsofexamplesin ordertobalancetheclasses.Thisactionfacilitatesthelearningfromtherecordeddata.Forevaluation, theauthorshaveevaluatedthesystembasedonthepredictionofseismichazardswherepositive resultshavebeenreviewedcomparedtotheclassificationofexampleswithoutbalancingthecases. KEywoRDS Clustering, Data Mining, Machine Learning, Seismic Hazards Detection, Supervised Classification, Unbalanced Data