Kokten Ulas

This is an early access version, the complete PDF, HTML, and XML versions will be available soon. Open AccessArticle by Goksu Tuysuzoglu , Kokten Ulas Birant and Derya Birant * Department of Computer Engineering, Dokuz Eylul University, Izmir 35390, Turkey * Author to whom correspondence should be addressed.

4. Experimental Studies The experiments were conducted on 18 datasets to evaluate the efficiency and validity of the proposed SSS method. To show the superiority of the proposed holo-training approach, we compared it with the traditional self-training approach. We also investigated the impact of different ratios of labeled data (10%, 25%, 50%, and 75%) on the performance of SSS.

4. Experimental Studies To demonstrate the effectiveness of the proposed MOMIL approach, the experiments were carried out on a publicly available LoL dataset. The algorithms in the multi-instance learning package in WEKA [50] were used with default parameters. As a base classifier, the C4.5 Decision Tree (DT) algorithm was preferred due to its efficiency. This algorithm calculates entropy, which is a powerful measure to determine how a tree node splits data.