| Title |
Comparative study of ensemble methods for multi-class data classification / |
| Translation of Title |
Daugiaklasių duomenų klasifikatorių apjungimo metodų palyginamasis tyrimas. |
| Authors |
Rangarajan, Jayasankar Prasanth |
| Full Text |
|
| Pages |
50 |
| Keywords [eng] |
classifier evaluation ; single class ; KNN ; SVM ; Bayes |
| Abstract [eng] |
The main Goal is to Estimate the performance (error rate) of a classifier. The lower the error, better the classifier. In this paper some sets of data's selected are set for evaluation. The data's are trained and validated by a couple of classifiers with different classifier methods, then the results are compared by performing ensemble combining approach were investigation and compare the well-known one-vs.-one and one-vs.-all decomposition strategies for multiclass data classification is done to find the best classifiers with smallest misclassification rate and also the combination method helps to improve the single classifiers result. The main motive of this research work was to discover methods for building a generalized ensemble of classifiers. As the performance on an empirical comparison of several multi classifier systems using several data sets those with different problems. Our experimental results shows that our ensemble methods on classifier will show the outperform of best state-of-art standalone ensemble methods. |
| Dissertation Institution |
Kauno technologijos universitetas. |
| Type |
Master thesis |
| Language |
English |
| Publication date |
2016 |
