An Improved Batch Classifier with Bands and Dimensions

This paper presents a new version of a branching batch classifier that has added fixed value ranges through bands, for each column or feature of the input dataset. Each layer branches like a tree, but has a different architecture to the current classifiers. Each branch is not for a feature, but for a change in output category. Therefore, each classifier classifies its own subset of data rows and categories, using averaged values only and with decreasing numbers of data row in each new level. When considering features however, it is shown that some of the data can be correctly classified through using fixed value ranges, while the rest can be classified by using the classifier technique. Tests show that the method can successfully classify benchmark datasets to better than the state-of-the-art. Fixed value ranges are like links and so the paper discusses the biological analogy with neurons and neuron links.