Evolutionary Computation in Data Mining

" Evolutionary Computation in Data Mining " edited by Ashish Ghosh and Lakhmi C. Jain, published in the series on Studies in Fuzziness and Soft Computing by Springer contains twelve skillful papers, which show the advances and state of the art in the domain of evolutionary methods' application in data mining. Data mining is a relatively broad domain of automatic knowledge discovery from databases and is now one of the most developed fields in the area of artificial intelligence. A rapid growth of data collected in various realms of human activity and their potential usefulness in these realms requires appropriate and efficient tools to extract and use of the potentially gathered knowledge. Evolutionary algorithms are very powerful tools and can be utilized in many ways to make use of knowledge hidden in terabites of data collected in databases all over the world. In the Chapter 1 by Ashish Ghosh some essential aspects of application of evolutionary computation methods for rule generation are discussed. focuses on data preprocessing methods meant to minimize the size of data to be processed. Data reduction is obtained using a CHC evolutionary algorithm with stratification strategy to select representative instances of the processed data set. The presented evolutionary method outperforms the non-evolutionary ones and gives better instance reduction rates, higher accuracy and lower resources consumption. The next chapter by Matthew G. Smith and Larry Bull presents also a preprocessing method, where genetic programming is used to feature construction (inferring new features enables to find hidden relationships between them) and a genetic algorithm for feature selection from the set produced by the first stage. The preprocessing method with the C4.5 classification algorithm works significantly better than the single C4.5 method. Also the fourth chapter by Riyaz Sikora presents a multi-agent based method of feature selection and inductive learning. Genetic algorithms are treated as learning agents to discover solutions of small subproblems into which the whole data set is decomposed.