Classification of Volatile Organic Compounds with Incremental SVMs and RBF Networks

Support Vector Machines (SVMs) have been applied to solve the classification of volatile organic compounds (VOC) data in some recent studies. SVMs provide good generalization performance in detection and classification of VOC data. However, in many applications involving VOC data, it is not unusual for additional data, which may include new classes, to become available over time, which then requires an SVM classifier that is capable of incremental learning that does not suffer from loss of previously acquired knowledge. In our previous work, we have proposed the incremental SVM approach based on Learn.MT. In this contribution, the ability of SVMLearn.MT to incrementally classify VOC data is evaluated and compared against a similarly constructed Learn.MT algorithm that uses radial basis function neural network as base classifiers.