A new classification model based on Evidence theory

Studies have revealed that a combination of classifiers is often more accurate than an individual classifier. A multiple classifier system can take advantage of the strengths of the individual classifiers, avoid their weaknesses, and improve classification accuracy. This system can be considered as an efficient mechanism to achieve the highest possible accuracy in medical classification problem. In this paper, we propose a new method for combination of multiple classifiers using Dempster-Shafer theory of evidence combination for mining medical data. We combine the beliefs of three classifiers: Multi-Layer Perception Neural Network, K-Nearest Neighbor and Naïve Bayesian. Our experiments over the Breast Cancer Wisconsin dataset shows improvement compared to the classification results produced by the individual classifiers and other classifiers which use the combination methods

Keywords:

Medical Data Mining , Multiple Classification , Dempster-Shafer Theory.