engleski

Selection of Classification Algorithm Using a Meta Learning Approach Based on Data Sets Characteristics

Many classification algorithms have been proposed, but not all of them are appropriate for a given classification problem. At the same time, there is no good way to choose appropriate classification algorithm for the problem at hand. In this paper, a meta learning data set classification algorithm recommendation, based on characteristics, is presented. An experimental study is performed using 128 real-world data sets and all research made has pointed to the same result: data sets characteristics significantly affect classification accuracy. Guidance in selection of classification algorithm based on data set characteristics is provided.

Data set characteristics ; meta learning ; classification accuracy ; neural networks ; decision trees ; discriminant analysis

nije evidentirano