Intrusion detection with evolutionary learning classifier systems

Supervised Learning Classifier System for Grid Data Mining

“…When new input enters into the system, it matches the input data with current population of classifiers. If there is some classifiers in the population that matched with condition and action parts of the new input data, those matched classifiers are stored in the correct set (C) (H. H. Dam, 2008, A Orriols_Puig, et al, 2007, K. Shafi, et al, 2007. Otherwise, correct set would be empty.…”

“…Genetic Algorithm (GA) and covering are the two methods for the learning process. The classifiers, whose condition part is correctly matched with the condition value of the training data, compose the match set (M) (H. H. Dam, 2008, A Orriols_Puig, et al, 2007, K. Shafi, et al, 2007. Similarly the classifiers, which are correctly matched with condition and action part of the training data are considered as the correct set (C) (H. H. Dam, 2008, A Orriols_Puig, et al, 2007, K. Shafi, et al, 2007.…”

“…The classifiers, whose condition part is correctly matched with the condition value of the training data, compose the match set (M) (H. H. Dam, 2008, A Orriols_Puig, et al, 2007, K. Shafi, et al, 2007. Similarly the classifiers, which are correctly matched with condition and action part of the training data are considered as the correct set (C) (H. H. Dam, 2008, A Orriols_Puig, et al, 2007, K. Shafi, et al, 2007. The covering is executed only when the correct set becomes empty; otherwise, if the average experience of classifier in the correct set is greater than the GA_ Threshold, then GA is executed (H. H. Dam, 2008, A Orriols_Puig, et al, 2007, K. Shafi, et al, 2007.…”

“…Similarly the classifiers, which are correctly matched with condition and action part of the training data are considered as the correct set (C) (H. H. Dam, 2008, A Orriols_Puig, et al, 2007, K. Shafi, et al, 2007. The covering is executed only when the correct set becomes empty; otherwise, if the average experience of classifier in the correct set is greater than the GA_ Threshold, then GA is executed (H. H. Dam, 2008, A Orriols_Puig, et al, 2007, K. Shafi, et al, 2007. Covering is the process of generating new classifiers into the population.…”

“…Grid computing is applied in this work, because of the compatibility for data mining in grid platform. Supervised classifier system is a newly introduced, successive implementation of learning classifier system (K. Shafi, et al, 2007). The following subsections give detailed explanation of supervised classifier system, distributed data mining, grid data mining and details of Gridclass system.…”