Experimental Comparisons of Multi-class Classifiers.
Lin Li,Yue Wu,Mao Ye
2015-01-01
Abstract:The multi-class classification algorithms are widely used by many areas such as machine learning and computer vision domains. Nowadays, many literatures described multi-class algorithms, however there are few literature that introduced them with thorough theoretical analysis and experimental comparisons. This paper discusses the principles, important parameters, application domain, runtime performance, accuracy, and etc. of twelve multi-class algorithms: decision tree, random forests, extremely randomized trees, multi-class adaboost classifier, stochastic gradient boosting, linear and nonlinear support vector machines, K nearest neighbors, multi-class logistic classifier, multi-layer perceptron, naive Bayesian classifier and conditional random fields classifier. The experiment tested on five data sets: SPECTF heart data set, Ionosphere radar data set, spam junk mail filter data set, optdigits handwriting data set and scene 15 image classification data set. Our major contribution is that we study the relationships between each classifier and impact of each parameters to classification results. The experiment shows that gradient boosted trees, nonlinear support vector machine, K nearest neighbor reach high performance under the circumstance of binary classification and minor data capacity; Under the condition of high dimension, multi-class and big data, however, gradient boosted trees, linear support vector machine, multi-class logistic classifier get good results. At last, the paper addresses the development and future of multi-class classifier algorithms.