Time-sensitive classification of behavioral data

Shin Ando, Einoshin Suzuki

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

2 Citations (Scopus)

Abstract

In this paper, we address a classification task under a timesensitive setting, in which the amount of observation required to make a prediction is viewed as a practical cost. Such a setting is intrinsic in many systems where the potential reward of the action against the predicted event depends on the response time, e.g., surveillance/warning and diagnostic applications. Meanwhile, predictions are usually less reliable when based on fewer observations, i.e., there exists a trade-off between such temporal cost and the accuracy. We address the task as a classification of subsequences in a time series. The goal is to predict the occurrences of events from subsequent observations and to learn when to commit to the prediction considering the trade-off. We propose an ensemble of classifiers which respectively makes predictions based on subsequences of different lengths. The prediction of the ensemble is given by the earliest confident prediction among the individual classifiers. We propose a cutting-plane algorithm for jointly training an ensemble of linear classifiers considering their temporal dependence. We compare the proposed algorithm against conventional approaches over a collection of behavioral trajectory data.

Original languageEnglish
Title of host publicationProceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013
EditorsSrinivasan Parthasarathy, Joydeep Ghosh, Zhi-Hua Zhou, Jennifer Dy, Zoran Obradovic, Chandrika Kamath
PublisherSiam Society
Pages458-466
Number of pages9
ISBN (Electronic)9781611972627
Publication statusPublished - Jan 1 2013
EventSIAM International Conference on Data Mining, SDM 2013 - Austin, United States
Duration: May 2 2013May 4 2013

Publication series

NameProceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013

Other

OtherSIAM International Conference on Data Mining, SDM 2013
CountryUnited States
CityAustin
Period5/2/135/4/13

Fingerprint

Classifiers
Costs
Time series
Trajectories

All Science Journal Classification (ASJC) codes

  • Computer Science Applications
  • Software
  • Theoretical Computer Science
  • Information Systems
  • Signal Processing

Cite this

Ando, S., & Suzuki, E. (2013). Time-sensitive classification of behavioral data. In S. Parthasarathy, J. Ghosh, Z-H. Zhou, J. Dy, Z. Obradovic, & C. Kamath (Eds.), Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013 (pp. 458-466). (Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013). Siam Society.

Time-sensitive classification of behavioral data. / Ando, Shin; Suzuki, Einoshin.

Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013. ed. / Srinivasan Parthasarathy; Joydeep Ghosh; Zhi-Hua Zhou; Jennifer Dy; Zoran Obradovic; Chandrika Kamath. Siam Society, 2013. p. 458-466 (Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013).

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

Ando, S & Suzuki, E 2013, Time-sensitive classification of behavioral data. in S Parthasarathy, J Ghosh, Z-H Zhou, J Dy, Z Obradovic & C Kamath (eds), Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013. Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013, Siam Society, pp. 458-466, SIAM International Conference on Data Mining, SDM 2013, Austin, United States, 5/2/13.
Ando S, Suzuki E. Time-sensitive classification of behavioral data. In Parthasarathy S, Ghosh J, Zhou Z-H, Dy J, Obradovic Z, Kamath C, editors, Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013. Siam Society. 2013. p. 458-466. (Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013).
Ando, Shin ; Suzuki, Einoshin. / Time-sensitive classification of behavioral data. Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013. editor / Srinivasan Parthasarathy ; Joydeep Ghosh ; Zhi-Hua Zhou ; Jennifer Dy ; Zoran Obradovic ; Chandrika Kamath. Siam Society, 2013. pp. 458-466 (Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013).
@inproceedings{7e93f146311f435ca3bdd8f52b68d73b,
title = "Time-sensitive classification of behavioral data",
abstract = "In this paper, we address a classification task under a timesensitive setting, in which the amount of observation required to make a prediction is viewed as a practical cost. Such a setting is intrinsic in many systems where the potential reward of the action against the predicted event depends on the response time, e.g., surveillance/warning and diagnostic applications. Meanwhile, predictions are usually less reliable when based on fewer observations, i.e., there exists a trade-off between such temporal cost and the accuracy. We address the task as a classification of subsequences in a time series. The goal is to predict the occurrences of events from subsequent observations and to learn when to commit to the prediction considering the trade-off. We propose an ensemble of classifiers which respectively makes predictions based on subsequences of different lengths. The prediction of the ensemble is given by the earliest confident prediction among the individual classifiers. We propose a cutting-plane algorithm for jointly training an ensemble of linear classifiers considering their temporal dependence. We compare the proposed algorithm against conventional approaches over a collection of behavioral trajectory data.",
author = "Shin Ando and Einoshin Suzuki",
year = "2013",
month = "1",
day = "1",
language = "English",
series = "Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013",
publisher = "Siam Society",
pages = "458--466",
editor = "Srinivasan Parthasarathy and Joydeep Ghosh and Zhi-Hua Zhou and Jennifer Dy and Zoran Obradovic and Chandrika Kamath",
booktitle = "Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013",

}

TY - GEN

T1 - Time-sensitive classification of behavioral data

AU - Ando, Shin

AU - Suzuki, Einoshin

PY - 2013/1/1

Y1 - 2013/1/1

N2 - In this paper, we address a classification task under a timesensitive setting, in which the amount of observation required to make a prediction is viewed as a practical cost. Such a setting is intrinsic in many systems where the potential reward of the action against the predicted event depends on the response time, e.g., surveillance/warning and diagnostic applications. Meanwhile, predictions are usually less reliable when based on fewer observations, i.e., there exists a trade-off between such temporal cost and the accuracy. We address the task as a classification of subsequences in a time series. The goal is to predict the occurrences of events from subsequent observations and to learn when to commit to the prediction considering the trade-off. We propose an ensemble of classifiers which respectively makes predictions based on subsequences of different lengths. The prediction of the ensemble is given by the earliest confident prediction among the individual classifiers. We propose a cutting-plane algorithm for jointly training an ensemble of linear classifiers considering their temporal dependence. We compare the proposed algorithm against conventional approaches over a collection of behavioral trajectory data.

AB - In this paper, we address a classification task under a timesensitive setting, in which the amount of observation required to make a prediction is viewed as a practical cost. Such a setting is intrinsic in many systems where the potential reward of the action against the predicted event depends on the response time, e.g., surveillance/warning and diagnostic applications. Meanwhile, predictions are usually less reliable when based on fewer observations, i.e., there exists a trade-off between such temporal cost and the accuracy. We address the task as a classification of subsequences in a time series. The goal is to predict the occurrences of events from subsequent observations and to learn when to commit to the prediction considering the trade-off. We propose an ensemble of classifiers which respectively makes predictions based on subsequences of different lengths. The prediction of the ensemble is given by the earliest confident prediction among the individual classifiers. We propose a cutting-plane algorithm for jointly training an ensemble of linear classifiers considering their temporal dependence. We compare the proposed algorithm against conventional approaches over a collection of behavioral trajectory data.

UR - http://www.scopus.com/inward/record.url?scp=84936983427&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84936983427&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:84936983427

T3 - Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013

SP - 458

EP - 466

BT - Proceedings of the 2013 SIAM International Conference on Data Mining, SDM 2013

A2 - Parthasarathy, Srinivasan

A2 - Ghosh, Joydeep

A2 - Zhou, Zhi-Hua

A2 - Dy, Jennifer

A2 - Obradovic, Zoran

A2 - Kamath, Chandrika

PB - Siam Society

ER -