Towards an MCDM-Based Evaluation Framework for Regression Algorithms

Prediction is a challenging topic and regression is one of the methods developed for this problem. Various regression algorithms produce different prediction results for different datasets. Various performance criteria are used to measure the regression algorithms performance. The Multiple Criteria Decision Making (MCDM) method can be used to synthesize measurements of the performance criteria for the algorithms to select the most appropriate algorithm for a dataset. This paper introduces a MCDM-based evaluation framework for regression algorithms selection for a dataset. Experiments are demonstrated for usability of the proposed framework.