Random projection ensemble classifiers

Alon Schclar, Lior Rokach

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

48 Scopus citations

Abstract

We introduce a novel ensemble model based on random projections. The contribution of using random projections is two-fold. First, the randomness provides the diversity which is required for the construction of an ensemble model. Second, random projections embed the original set into a space of lower dimension while preserving the dataset's geometrical structure to a given distortion. This reduces the computational complexity of the model construction as well as the complexity of the classification. Furthermore, dimensionality reduction removes noisy features from the data and also represents the information which is inherent in the raw data by using a small number of features. The noise removal increases the accuracy of the classifier. The proposed scheme was tested using WEKA based procedures that were applied to 16 benchmark dataset from the UCI repository.

Original languageEnglish
Title of host publicationEnterprise Information Systems - 11th International Conference, ICEIS 2009, Proceedings
PublisherSpringer Verlag
Pages309-316
Number of pages8
ISBN (Print)9783642013461
DOIs
StatePublished - 1 Jan 2009
Event11th International Conference on Enterprise Information Systems, ICEIS 2009 - Milan, Italy
Duration: 6 May 200910 May 2009

Publication series

NameLecture Notes in Business Information Processing
Volume24 LNBIP
ISSN (Print)1865-1348

Conference

Conference11th International Conference on Enterprise Information Systems, ICEIS 2009
Country/TerritoryItaly
CityMilan
Period6/05/0910/05/09

Keywords

  • Classification
  • Ensemble methods
  • Pattern recognition
  • Random projections

ASJC Scopus subject areas

  • Management Information Systems
  • Control and Systems Engineering
  • Business and International Management
  • Information Systems
  • Modeling and Simulation
  • Information Systems and Management

Fingerprint

Dive into the research topics of 'Random projection ensemble classifiers'. Together they form a unique fingerprint.

Cite this