3 Citations (Scopus)
48 Downloads (Pure)

Abstract

The abundance of data available on Wireless Sensor Networks makes online processing necessary. In industrial applications for example, the correct operation of equipment can be the point of interest while raw sampled data is of minor importance. Classi﬿cation algorithms can be used to make state classi﬿cations based on the available data. The distributed nature of Wireless Sensor Networks is a complication that needs to be considered when implementing classi﬿cation algorithms. In this work, we investigate the bottlenecks that limit the options for distributed execution of three widely used algorithms: Feed Forward Neural Networks, naive Bayes classi﬿ers and decision trees. By analyzing theoretical boundaries and using simulations of various network topologies, we show that the naive Bayes classi﬿er is the most flexible algorithm for distribution. Decision trees can be distributed efficiently but are unpredictable. Feed Forward Neural Networks show severe limitations.
Original languageUndefined
Pages (from-to)960-967
Number of pages8
JournalProcedia computer science
Volume10
DOIs
Publication statusPublished - 2012

Keywords

  • EWI-22476
  • IR-82145
  • METIS-289771

Cite this

@article{e3ccdf9b127d42fd86e24f8644c52673,
title = "Distribution Bottlenecks in Classification Algorithms",
abstract = "The abundance of data available on Wireless Sensor Networks makes online processing necessary. In industrial applications for example, the correct operation of equipment can be the point of interest while raw sampled data is of minor importance. Classi﬿cation algorithms can be used to make state classi﬿cations based on the available data. The distributed nature of Wireless Sensor Networks is a complication that needs to be considered when implementing classi﬿cation algorithms. In this work, we investigate the bottlenecks that limit the options for distributed execution of three widely used algorithms: Feed Forward Neural Networks, naive Bayes classi﬿ers and decision trees. By analyzing theoretical boundaries and using simulations of various network topologies, we show that the naive Bayes classi﬿er is the most flexible algorithm for distribution. Decision trees can be distributed efficiently but are unpredictable. Feed Forward Neural Networks show severe limitations.",
keywords = "EWI-22476, IR-82145, METIS-289771",
author = "G.J. Zwartjes and Havinga, {Paul J.M.} and Smit, {Gerardus Johannes Maria} and Hurink, {Johann L.}",
note = "Open access. ANT 2012 and MobiWIS 2012",
year = "2012",
doi = "10.1016/j.procs.2012.06.131",
language = "Undefined",
volume = "10",
pages = "960--967",
journal = "Procedia computer science",
issn = "1877-0509",
publisher = "Elsevier",

}

Distribution Bottlenecks in Classification Algorithms. / Zwartjes, G.J.; Havinga, Paul J.M.; Smit, Gerardus Johannes Maria; Hurink, Johann L.

In: Procedia computer science, Vol. 10, 2012, p. 960-967.

Research output: Contribution to journalArticleAcademicpeer-review

TY - JOUR

T1 - Distribution Bottlenecks in Classification Algorithms

AU - Zwartjes, G.J.

AU - Havinga, Paul J.M.

AU - Smit, Gerardus Johannes Maria

AU - Hurink, Johann L.

N1 - Open access. ANT 2012 and MobiWIS 2012

PY - 2012

Y1 - 2012

N2 - The abundance of data available on Wireless Sensor Networks makes online processing necessary. In industrial applications for example, the correct operation of equipment can be the point of interest while raw sampled data is of minor importance. Classi﬿cation algorithms can be used to make state classi﬿cations based on the available data. The distributed nature of Wireless Sensor Networks is a complication that needs to be considered when implementing classi﬿cation algorithms. In this work, we investigate the bottlenecks that limit the options for distributed execution of three widely used algorithms: Feed Forward Neural Networks, naive Bayes classi﬿ers and decision trees. By analyzing theoretical boundaries and using simulations of various network topologies, we show that the naive Bayes classi﬿er is the most flexible algorithm for distribution. Decision trees can be distributed efficiently but are unpredictable. Feed Forward Neural Networks show severe limitations.

AB - The abundance of data available on Wireless Sensor Networks makes online processing necessary. In industrial applications for example, the correct operation of equipment can be the point of interest while raw sampled data is of minor importance. Classi﬿cation algorithms can be used to make state classi﬿cations based on the available data. The distributed nature of Wireless Sensor Networks is a complication that needs to be considered when implementing classi﬿cation algorithms. In this work, we investigate the bottlenecks that limit the options for distributed execution of three widely used algorithms: Feed Forward Neural Networks, naive Bayes classi﬿ers and decision trees. By analyzing theoretical boundaries and using simulations of various network topologies, we show that the naive Bayes classi﬿er is the most flexible algorithm for distribution. Decision trees can be distributed efficiently but are unpredictable. Feed Forward Neural Networks show severe limitations.

KW - EWI-22476

KW - IR-82145

KW - METIS-289771

U2 - 10.1016/j.procs.2012.06.131

DO - 10.1016/j.procs.2012.06.131

M3 - Article

VL - 10

SP - 960

EP - 967

JO - Procedia computer science

JF - Procedia computer science

SN - 1877-0509

ER -