A Weighted Minimum Distance Classifier Based on Relative Offset

In many classifiers based on supervised learning, the Minimum Distance Classifier (MDC) is widely used because of its simplicity and effectiveness. But it has the disadvantage of the low accuracy on classification. To improve the performance of the MDC, this paper proposed the Weighted Minimum Distance Classifier based on Relative Offset (WMDCRO). Under the condition of the Fisher criterion, this method defined the weights of each category attribute in terms of the relative distance. It analyzed and judged the degree of clustering about data in the feature space by using the relative offset. The experiment results show that the accuracy of this classifier is higher than that of the traditional MDC. Compared with the Naive Bayes classifier, the classification stability of this classifier is better.