Ants Constructing Rule-Based Classifiers

This chapter introduces a new algorithm for classification, named AntMiner+, based on an artificial ant system with inherent self-organizing capabilities. The usage of ant systems generates scalable data mining solutions that are easily distributed and robust to failure. The introduced approach differs from the previously proposed AntMiner classification technique in three aspects. Firstly, AntMiner+ uses a M AX -M I N ant system which is an improved version of the originally proposed ant system, yielding better performing classifiers. Secondly, the complexity of the environment in which the ants operate has substantially decreased. This simplification results in more effective decision making by the ants. Finally, by making a distinction between ordinal and nominal variables, AntMiner+ is able to include intervals in the rules which leads to fewer and better performing rules. The conducted experiments benchmark AntMiner+ with several state-of-the-art classification techniques on a variety of datasets. It is concluded that AntMiner+ builds accurate, comprehensible classifiers that outperform C4.5 inferred classifiers and are competitive with the included black-box techniques.