An Incremental Classification Based-On Mahalanobis Measurement
Sunantha Sodsee, Maytiyanin Komhao, Phayung Meesad · ITC-CSCC :International Technical Conference on Circuits Systems, Computers and Communications · 2007
We propose a new classification classifier for supervised learning. The proposed classifier, which is called an Incremental Learning Algorithm Based-On Mahalanobis distance (ILM). ILM has various features, which are an incremental learning, hard and soft decision, and a classification using an ellipsoid shape. It uses Mahalanobis distance to measure a similarity and dissimilarity among data; the Mahalanobis Gaussian function is used to fuzzify the distance into a degree of membership. To evaluate the performance of ILM, five popular benchmark datasets: Iris Plants Dataset, Wisconsin Breast Cancer Dataset, Sonar-Mines-Rocks Dataset, Vowel Recognition Dataset, and StatLog DNA Dataset are used to compare its performance with other existing classifiers. The results can be concluded that ILM can classify the patterns in the supervised learning effectively.