Machine learning programs can automatically learn to recognise complex patterns and make intelligent decisions based on data. Machine learning has become a powerful tool for data mining. A great deal of research in machine learning has focused on concept learning or classification learning. Among the various machine learning approaches that have been developed for classification, inductive learning from examples is the most commonly adopted in real-life applications. Due to non-uniform data formats and huge volume of data, it is a challenge for scientists across different disciplines to optimise the process of knowledge acquisition from data with naïve inductive learning techniques. The overarching purpose of this research is to develop a novel and efficient rule induction algorithm a learning algorithm for inducing general rules from specific examples that can deal with both discrete and continuous variables without the need for data pre-processing. This thesis presents a novel rule induction algorithm known as RULES-8 which utilises guidelines for the selection of seed examples, together with a simple method to form rules. The research also aims to improve current pruning methods for handling noisy examples. Another major concern of the work is designing a new heuristic for controlling the rule formation and selection processes. Finally, it concentrates on developing a new efficient learning algorithm for continuous output using fuzzy logic theory. The proposed algorithm allows automatic creation of membership functions and produces accurate as well as compact fuzzy sets.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:567332 |
| Date | January 2012 |
| Creators | Pham, Dinh |
| Publisher | Cardiff University |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://orca.cf.ac.uk/34591/ |
Page generated in 0.0018 seconds