Feature Selection and Ensemble Meta Classifier for Multiclass Imbalance Data Learning
The aim of this paper is to investigate the effects of combining feature selection and ensemble classifiers on the prediction performance in addressing the multiclass imbalance data learning .This research uses data obtained from the Malaysian medicinal leaf images shape data and three other large...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Conference or Workshop Item |
| Language: | English |
| Published: |
2018
|
| Subjects: | |
| Online Access: | http://repo.uum.edu.my/25208/1/KMICE%202018%20134%20139.pdf http://repo.uum.edu.my/25208/ http://www.kmice.cms.net.my/ProcKMICe/KMICe2018/toc.html |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The aim of this paper is to investigate the effects of combining feature selection and ensemble classifiers on the prediction performance in addressing the multiclass imbalance data learning .This research uses data obtained from the Malaysian medicinal leaf images shape data and
three other large benchmark data sets in which six
ensemble methods from Weka machine learning
tool were selected to perform the classification task.These ensemble methods include the AdaboostM1, Bagging, Decorate, END, MultiboostAB, and Rotation Forest.In addition, five base classifiers were used; Naïve Bayes, SMO, J48, Random Forest, and Random Tree in order to examine the performance of the ensemble methods. There are
two feature selection approaches implemented
which are filter-based (CfsSubsetEval, ConsistencySubsetEval and FilteredSubsetEval)
and wrapper-based (WrapperSubsetEval). The
results obtained from the experiments show that
although the performance accuracy is not much
improved, however, with less number of attributes,
the classifiers are able to achieve similar accuracy or slightly improved with less processing time.In knowledge management, the findings provide
important insight of which algorithm is suitable for decision making when dealing with high dimensional and large data. |
|---|