Exercises done during the discipline of Ensemble Learning (Machine Learnig and Signal Processing PhD)
In this exercise, I implemented and evaluated the Bagging ensemble method. Two data sets from UCI were used for evaluation. Three kinds of classifiers were used: Decision Trees, k-Nearest Neighbors (kNN), and Multilayer Perceptrons (kNN). More details of the experiments are described in the report.
I implemented an Ensemble Pruning method via Individual Contribution ordering (EPIC). I compared the performance of this method with the performance of Bagging, using a dataset from UCI. More details of the experiments are described in the report.
In this exercise, I implemented two methods of Dynamic Classifier Selection from a pool of classifiers: selection via Overall Local Accuracy (OLA) and via Local Class Accuracy (LCA). the performance of these methods with the performance of Bagging, using a dataset from UCI. More details of the experiments are described in the report.
The project is to replicate the results reported in the following paper:
[Guo, L. and Boukir, S.] Margin-based ordered aggregation for ensemble pruning, Pattern Recognition Letters, v. 34, pp. 603-609, 2013.
More details are described in the report.