Name
Abstract | ||
|---|---|---|
Voting ensemble method combines results of single classifiers aiming to offer improved classification performance. However, it is intuitively accepted that the combined classifiers during voting should be both diverse and accurate. In this study, we used the unsupervised method of cluster analysis in four datasets related to medical diagnosis in order to differentiate the single classifiers according to their individual results. Using this information we selected the most accurate among similar classifiers proposing the optimal classifier combination for each dataset. The results show that the estimated combination was actually the best performing during voting training for two of the datasets while in the other two it was one of those that outperformed single classifiers. The proposed methodology is a quick and easy tool for estimating classifier combinations that outperforms the single classifiers during voting. |
Year | DOI | Venue |
|---|---|---|
2015 | 10.1145/2797143.2797156 | EANN Workshops |
Field | DocType | Citations |
Voting,Pattern recognition,Random subspace method,Computer science,Cascading classifiers,Artificial intelligence,Classifier (linguistics),Machine learning,Medical diagnosis | Conference | 1 |
PageRank | References | Authors |
0.34 | 18 | 4 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Androniki Tamvakis | 1 | 4 | 1.42 |
| Christos-Nikolaos Anagnostopoulos | 2 | 1034 | 91.30 |
| George E. Tsekouras | 3 | 235 | 20.14 |
| George C. Anastassopoulos | 4 | 82 | 16.82 |