Probabilistic Decision Trees using SVM for Multi-class Classification

Abstract

In the automotive repairing backdrop, retrieving from previously solved incident the database features that could support and speed up the diagnostic is of great usefulness. This decision helping process should give a fixed number of the more relevant diagnostic classified in a likelihood sense. It is a probabilistic multi-class classification problem. This paper describes an original classification technique, the Probabilistic Decision Tree (PDT) producing a posteriori probabilities in a multi-class context. It is based on a Binary Decision Tree (BDT) with Probabilistic Support Vector Machine classifier (PSVM). At each node of the tree, a bi-class SVM along with a sigmoid function are trained to give a probabilistic classification output. For each branch, the outputs of all the nodes composing the branch are combined to lead to a complete evaluation of the probability when reaching the final leaf (representing the class associated to the branch). To illustrate the effectiveness of PDTs, they are tested on benchmark datasets and results are compared with other existing approaches.