Using ABC algorithm for classification and analysis on effects of control parameters

In this study, Artificial Bee Colony (ABC) algorithm based classifier is used. Also, in order to improve the effectiveness of ABC algorithm, some modifications are done. New method is called MABC algorithm. Both methods are applied on various real life data sets such as IRIS, WINE, PIMA, BUPA, ECG and results are compared. Those datasets are obtained from UCI Machine Learning Repository and MITBIH ECG database. In addition to it, validity indices and effects of some control parameters such as MCN, Limit are examined. It is observed that, selected features have significiant effect on classification success rate of classifier. If there is high overlap between the classes, success rate of classifier decreases. However observed results indicate that ABC algorithm can successfully be used for classification of multi dimensional datasets. By means of SCTR control parameter, MABC algorithm based classifier provides higher classification success rates versus ABC algorithm, independent from Limit and MCN values.