Comparing Estimation Achievements by Determining Ideal Training Iteration Numbers in Supervised Machine Learning Algorithms


Bekmezci S., YİĞİT T.

2017 International Conference on Computer Science and Engineering (UBMK), Antalya, Turkey, 5 - 08 October 2017, pp.649-654 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Volume:
  • Doi Number: 10.1109/ubmk.2017.8093490
  • City: Antalya
  • Country: Turkey
  • Page Numbers: pp.649-654

Abstract

Data analysis has become important with the ever-increasing and diversified data from the past to the present. Supervised machine learning algorithms have been one of the preferred methods with time because of the ability to produce fast and effective solutions in data analysis. One of the most important problems of supervised machine learning algorithms is the inability to achieve the ideal predicted values which are obtained by adversely affecting the test values of little or over training. Therefore, determining the ideal number of training iterations for each algorithm is important for algorithm estimation success. In this study, it has been tried to obtain the best results in the estimation performance with the help of an application being developed in determining the ideal training iteration numbers of the supervised machine learning algorithms on a sample dataset and the prediction performances of the algorithms are compared with each others.