Discrimination of Mine Seismic Events and Blasts Using the Fisher Classifier, Naive Bayesian Classifier and Logistic Regression

L. Dong, Johan Wesseloo, Yves Potvin, X. Li

Research output: Contribution to journalArticle

78 Citations (Scopus)

Abstract

© 2015, Springer-Verlag Wien. Seismic events and blasts generate seismic waveforms that have different characteristics. The challenge to confidently differentiate these two signatures is complex and requires the integration of physical and statistical techniques. In this paper, the different characteristics of blasts and seismic events were investigated by comparing probability density distributions of different parameters. Five typical parameters of blasts and events and the probability density functions of blast time, as well as probability density functions of origin time difference for neighbouring blasts were extracted as discriminant indicators. The Fisher classifier, naive Bayesian classifier and logistic regression were used to establish discriminators. Databases from three Australian and Canadian mines were established for training, calibrating and testing the discriminant models. The classification performances and discriminant precision of the three statistical techniques were discussed and compared. The proposed discriminators have explicit and simple functions which can be easily used by workers in mines or researchers. Back-test, applied results, cross-validated results and analysis of receiver operating characteristic curves in different mines have shown that the discriminator for one of the mines has a reasonably good discriminating performance.
Original languageEnglish
Pages (from-to)183-211
JournalRock Mechanics and Rock Engineering
Volume49
Issue number1
DOIs
Publication statusPublished - 2016

Fingerprint

Discriminators
Logistics
logistics
Classifiers
probability density function
Probability density function
Testing
parameter

Cite this

@article{204f0e1d5dc24235b38487b87e732c94,
title = "Discrimination of Mine Seismic Events and Blasts Using the Fisher Classifier, Naive Bayesian Classifier and Logistic Regression",
abstract = "{\circledC} 2015, Springer-Verlag Wien. Seismic events and blasts generate seismic waveforms that have different characteristics. The challenge to confidently differentiate these two signatures is complex and requires the integration of physical and statistical techniques. In this paper, the different characteristics of blasts and seismic events were investigated by comparing probability density distributions of different parameters. Five typical parameters of blasts and events and the probability density functions of blast time, as well as probability density functions of origin time difference for neighbouring blasts were extracted as discriminant indicators. The Fisher classifier, naive Bayesian classifier and logistic regression were used to establish discriminators. Databases from three Australian and Canadian mines were established for training, calibrating and testing the discriminant models. The classification performances and discriminant precision of the three statistical techniques were discussed and compared. The proposed discriminators have explicit and simple functions which can be easily used by workers in mines or researchers. Back-test, applied results, cross-validated results and analysis of receiver operating characteristic curves in different mines have shown that the discriminator for one of the mines has a reasonably good discriminating performance.",
author = "L. Dong and Johan Wesseloo and Yves Potvin and X. Li",
year = "2016",
doi = "10.1007/s00603-015-0733-y",
language = "English",
volume = "49",
pages = "183--211",
journal = "Journal Rock Mechanics and Rock Engineering",
issn = "0723-2632",
publisher = "Springer",
number = "1",

}

TY - JOUR

T1 - Discrimination of Mine Seismic Events and Blasts Using the Fisher Classifier, Naive Bayesian Classifier and Logistic Regression

AU - Dong, L.

AU - Wesseloo, Johan

AU - Potvin, Yves

AU - Li, X.

PY - 2016

Y1 - 2016

N2 - © 2015, Springer-Verlag Wien. Seismic events and blasts generate seismic waveforms that have different characteristics. The challenge to confidently differentiate these two signatures is complex and requires the integration of physical and statistical techniques. In this paper, the different characteristics of blasts and seismic events were investigated by comparing probability density distributions of different parameters. Five typical parameters of blasts and events and the probability density functions of blast time, as well as probability density functions of origin time difference for neighbouring blasts were extracted as discriminant indicators. The Fisher classifier, naive Bayesian classifier and logistic regression were used to establish discriminators. Databases from three Australian and Canadian mines were established for training, calibrating and testing the discriminant models. The classification performances and discriminant precision of the three statistical techniques were discussed and compared. The proposed discriminators have explicit and simple functions which can be easily used by workers in mines or researchers. Back-test, applied results, cross-validated results and analysis of receiver operating characteristic curves in different mines have shown that the discriminator for one of the mines has a reasonably good discriminating performance.

AB - © 2015, Springer-Verlag Wien. Seismic events and blasts generate seismic waveforms that have different characteristics. The challenge to confidently differentiate these two signatures is complex and requires the integration of physical and statistical techniques. In this paper, the different characteristics of blasts and seismic events were investigated by comparing probability density distributions of different parameters. Five typical parameters of blasts and events and the probability density functions of blast time, as well as probability density functions of origin time difference for neighbouring blasts were extracted as discriminant indicators. The Fisher classifier, naive Bayesian classifier and logistic regression were used to establish discriminators. Databases from three Australian and Canadian mines were established for training, calibrating and testing the discriminant models. The classification performances and discriminant precision of the three statistical techniques were discussed and compared. The proposed discriminators have explicit and simple functions which can be easily used by workers in mines or researchers. Back-test, applied results, cross-validated results and analysis of receiver operating characteristic curves in different mines have shown that the discriminator for one of the mines has a reasonably good discriminating performance.

U2 - 10.1007/s00603-015-0733-y

DO - 10.1007/s00603-015-0733-y

M3 - Article

VL - 49

SP - 183

EP - 211

JO - Journal Rock Mechanics and Rock Engineering

JF - Journal Rock Mechanics and Rock Engineering

SN - 0723-2632

IS - 1

ER -