The main objective of the present thesis is the development of automatic programming models based on ant colony optimization (ACO) to address classification problems.
Among several data mining (DM) tasks, classification aims to predict the class to which a specific instance belongs to, taking into account the values of its attributes (predictive attributes). To this end, a model or classifier is inferred from a training set. Then, it can be employed later to classify new instances or patterns that are not labelled.
Many algorithms and techniques have been employed to address the classification task. Recently, ant colony optimization metaheuristic has tackled this task successfully. ACO is a nature inspired optimization metaheuristic which mimic the behavior and self-organization of ant colonies in their search for food. On the other hand, Genetic Programming (GP), a particular type of automatic programming where genetic algorithms are used as search technique, also has demonstrated to obtain good results for classification. In contrast, another type of automatic programming known as Ant Programming (AP), which uses ACO instead of genetic algorithms as search technique, has never applied to classification. Considering the good results obtained by both ACO and GP for classification, we consider that it would be interesting to explore the application of AP to this task.
The main objective of this thesis can be broke down in the following subobjectives:
- Carry out a theoretical study of the existent ACO-based algorithms for classification rule mining.
- Perform a bibliographic review of the several proposals of AP presented to date.
- Develop an AP model based on the use of a context-free grammar for the extraction of classification rules.
- Address the classification task from a multi-objective perspective. Adapt the previous model to this approach.
- Evaluate the implemented models over different problems of actual interest by using standard UCI data:
- Imbalanced data
- Intrusion detection systems
- Text categorization
The development of this thesis was supported by:
- Spanish Ministry of Science and Technology, project TIN2011-22408.
- Regional Government of Andalusia, project P08-TIC-3720.
PUBLICATIONS ASSOCIATED WITH THIS THESIS
- J.L. Olmo, J.R. Romero and S. Ventura. Using Ant Programming Guided by Grammar for Building Rule-Based Classifiers.IEEE Transactions on Systems, Man and Cybernetics-Part B: Cybernetics, vol. 41(6), pp. 1585-1599. 2011.
- J.L. Olmo, J.R. Romero and S. Ventura. Classification rule mining using ant programming guided by grammar with multiple Pareto fronts. Soft Computing, vol. 16(12), pp. 2143-2163. 2012.
- J.L. Olmo, J.M. Luna, J.R. Romero and S. Ventura. Mining association rules with single and multi-objective grammar guided ant programming. Integrated Computer Aided Engineering, vol. 20(3), pp. 217-234. 2013.
- J.L. Olmo, J.M. Luna, J.R. Romero and S. Ventura. An Automatic Programming ACO-Based Algorithm for Classification Rule Mining. Trends in Practical Applications of Agents and Multiagent Systems. Advances in Intelligent and Soft Computing, vol. 71/2010, pp. 649-656. 2010.
- J.L. Olmo, J.R. Romero and S. Ventura. A grammar based ant programming algorithm for mining classification rules.Proceedings of the 2010 IEEE Congress on Evolutionary Computation (CEC’10), pp. 225-232. 2010.
- J.L. Olmo, J.M. Luna, J.R. Romero and S. Ventura. Association rule mining using a multi-objective grammar-based ant programming algorithm. Proceedings of the 11th International Conference on Intelligent Systems Design and Applications (ISDA’11), pp. 971-977. 2011.
- J.L. Olmo, J.R. Romero and S. Ventura. Multi-Objective Ant Programming for Mining Classification Rules. Proceedings of the 15th European Conference on Genetic Programming (EuroGP’12), LNCS Volume 7244/2012, pp. 146-157. 2012.
- J.L. Olmo, A. Cano, J.R. Romero, and S. Ventura. Binary and Multiclass Imbalanced Classification Using Multi-Objective Ant Programming. Proceedings of the 12th International Conference on Intelligent Systems Design and Applications (ISDA’12), pp. 70-76. 2012.