Semi-supervised Data Mining Tool Design with Self-tuning Optimization Techniques

摘要

Due to its wide applicability, the problem of semi-supervised classification is attracting increasing attention in machine learning. Presented in this article are semi-supervised artificial neural network- (ANN) and support vector machine- (SVM) based classifiers designed by the self-configuring genetic algorithm (SelfCGA) and the fuzzy controlled meta-heuristic approach Cooperation of Biology Related Algorithms (COBRA). Both data mining tools are based on dividing instances from different classes using both labelled and unlabelled examples. A new collective bionic algorithm, namely fuzzy controlled cooperation of biology-related algorithms, which solves constrained optimization problems, COBRA-cf, has been developed for the design of semi-supervised SVMs. Firstly, the experimental results obtained by the two types of fuzzy controlled COBRA are presented and compared and their usefulness is demonstrated. Then the performance and behaviour of the proposed semi-supervised SVMs and semi-supervised ANNs were studied under common experimental settings and their workability was established. Then their efficiency was estimated on a speech-based emotion recognition problem. Thus, the workability of the proposed meta-heuristic optimization algorithms was confirmed.