CoCoST: A computational cost sensitive classifier

Liyun Li, Umut Topkara, Baris Coskun, Nasir Memon

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Computational cost of classification is as important as accuracy in on-line classification systems. The computational cost is usually dominated by the cost of computing implicit features of the raw input data. Very few efforts have been made to design classifiers which perform effectively with limited computational power; instead, feature selection is usually employed as a pre-processing step to reduce the cost of running traditional classifiers. We present CoCoST, a novel and effective approach for building classifiers which achieve state-of- the-art classification accuracy, while keeping the expected computational cost of classification low, even without feature selection. CoCost employs a wide range of novel cost-aware decision trees, each of which is tuned to specialize in classifying instances from a subset of the input space, and judiciously consults them depending on the input instance in accordance with a cost-aware meta-classifier. Experimental results on a network flow detection application show that, our approach can achieve better accuracy than classifiers such as SVM and random forests, while achieving 75%-90% reduction in the computational costs.

Original languageEnglish (US)
Title of host publicationICDM 2009 - The 9th IEEE International Conference on Data Mining
Pages268-277
Number of pages10
DOIs
StatePublished - 2009
Event9th IEEE International Conference on Data Mining, ICDM 2009 - Miami, FL, United States
Duration: Dec 6 2009Dec 9 2009

Other

Other9th IEEE International Conference on Data Mining, ICDM 2009
CountryUnited States
CityMiami, FL
Period12/6/0912/9/09

Fingerprint

Classifiers
Costs
Feature extraction
Decision trees
Processing

Keywords

  • Cost efficient decision tree
  • Inverse-boosting
  • Meta-classifier
  • Suppressed cost

ASJC Scopus subject areas

  • Engineering(all)

Cite this

Li, L., Topkara, U., Coskun, B., & Memon, N. (2009). CoCoST: A computational cost sensitive classifier. In ICDM 2009 - The 9th IEEE International Conference on Data Mining (pp. 268-277). [5360252] https://doi.org/10.1109/ICDM.2009.46

CoCoST : A computational cost sensitive classifier. / Li, Liyun; Topkara, Umut; Coskun, Baris; Memon, Nasir.

ICDM 2009 - The 9th IEEE International Conference on Data Mining. 2009. p. 268-277 5360252.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Li, L, Topkara, U, Coskun, B & Memon, N 2009, CoCoST: A computational cost sensitive classifier. in ICDM 2009 - The 9th IEEE International Conference on Data Mining., 5360252, pp. 268-277, 9th IEEE International Conference on Data Mining, ICDM 2009, Miami, FL, United States, 12/6/09. https://doi.org/10.1109/ICDM.2009.46
Li L, Topkara U, Coskun B, Memon N. CoCoST: A computational cost sensitive classifier. In ICDM 2009 - The 9th IEEE International Conference on Data Mining. 2009. p. 268-277. 5360252 https://doi.org/10.1109/ICDM.2009.46
Li, Liyun ; Topkara, Umut ; Coskun, Baris ; Memon, Nasir. / CoCoST : A computational cost sensitive classifier. ICDM 2009 - The 9th IEEE International Conference on Data Mining. 2009. pp. 268-277
@inproceedings{a27df686040b4d1ca1bead879a0296bc,
title = "CoCoST: A computational cost sensitive classifier",
abstract = "Computational cost of classification is as important as accuracy in on-line classification systems. The computational cost is usually dominated by the cost of computing implicit features of the raw input data. Very few efforts have been made to design classifiers which perform effectively with limited computational power; instead, feature selection is usually employed as a pre-processing step to reduce the cost of running traditional classifiers. We present CoCoST, a novel and effective approach for building classifiers which achieve state-of- the-art classification accuracy, while keeping the expected computational cost of classification low, even without feature selection. CoCost employs a wide range of novel cost-aware decision trees, each of which is tuned to specialize in classifying instances from a subset of the input space, and judiciously consults them depending on the input instance in accordance with a cost-aware meta-classifier. Experimental results on a network flow detection application show that, our approach can achieve better accuracy than classifiers such as SVM and random forests, while achieving 75{\%}-90{\%} reduction in the computational costs.",
keywords = "Cost efficient decision tree, Inverse-boosting, Meta-classifier, Suppressed cost",
author = "Liyun Li and Umut Topkara and Baris Coskun and Nasir Memon",
year = "2009",
doi = "10.1109/ICDM.2009.46",
language = "English (US)",
isbn = "9780769538952",
pages = "268--277",
booktitle = "ICDM 2009 - The 9th IEEE International Conference on Data Mining",

}

TY - GEN

T1 - CoCoST

T2 - A computational cost sensitive classifier

AU - Li, Liyun

AU - Topkara, Umut

AU - Coskun, Baris

AU - Memon, Nasir

PY - 2009

Y1 - 2009

N2 - Computational cost of classification is as important as accuracy in on-line classification systems. The computational cost is usually dominated by the cost of computing implicit features of the raw input data. Very few efforts have been made to design classifiers which perform effectively with limited computational power; instead, feature selection is usually employed as a pre-processing step to reduce the cost of running traditional classifiers. We present CoCoST, a novel and effective approach for building classifiers which achieve state-of- the-art classification accuracy, while keeping the expected computational cost of classification low, even without feature selection. CoCost employs a wide range of novel cost-aware decision trees, each of which is tuned to specialize in classifying instances from a subset of the input space, and judiciously consults them depending on the input instance in accordance with a cost-aware meta-classifier. Experimental results on a network flow detection application show that, our approach can achieve better accuracy than classifiers such as SVM and random forests, while achieving 75%-90% reduction in the computational costs.

AB - Computational cost of classification is as important as accuracy in on-line classification systems. The computational cost is usually dominated by the cost of computing implicit features of the raw input data. Very few efforts have been made to design classifiers which perform effectively with limited computational power; instead, feature selection is usually employed as a pre-processing step to reduce the cost of running traditional classifiers. We present CoCoST, a novel and effective approach for building classifiers which achieve state-of- the-art classification accuracy, while keeping the expected computational cost of classification low, even without feature selection. CoCost employs a wide range of novel cost-aware decision trees, each of which is tuned to specialize in classifying instances from a subset of the input space, and judiciously consults them depending on the input instance in accordance with a cost-aware meta-classifier. Experimental results on a network flow detection application show that, our approach can achieve better accuracy than classifiers such as SVM and random forests, while achieving 75%-90% reduction in the computational costs.

KW - Cost efficient decision tree

KW - Inverse-boosting

KW - Meta-classifier

KW - Suppressed cost

UR - http://www.scopus.com/inward/record.url?scp=77951173680&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77951173680&partnerID=8YFLogxK

U2 - 10.1109/ICDM.2009.46

DO - 10.1109/ICDM.2009.46

M3 - Conference contribution

AN - SCOPUS:77951173680

SN - 9780769538952

SP - 268

EP - 277

BT - ICDM 2009 - The 9th IEEE International Conference on Data Mining

ER -