This paper compares performance of the Pittsburgh-style system GAssist with the Michigan-style system XCS on several datamining problems. Our analysis shows that both systems are suitable for datamining but have different advantages and disadvantages. The study does not only reveal important differences between the two systems but also suggests several structural properties of the underlying datasets. Introduction Successful data mining applications are important for modern-day learning

more »

... er systems (LCSs). Additionally, the study and comparison of different types of data miners on various data sets may enable the identification of strengths and weaknesses of the respective data miners. Several types of problem difficulty can be distinguished in data mining including data volume, search space size and type, complexity of the concept, noise in the data, the handling of missing values, or the problem of over-fitting. Successful datamining applications of learning classifier systems have been shown in the past (?) investigating and comparing performance of the accuracybased Michigan-style LCS XCS (?) and the Pittsburgh-style LCS GALE (?). Both systems showed competent performance in comparison to six other machine learning systems. Recently, new systems have appeared in the LCS field, like the Pitt-style LCS GAssist (?). Also, there are improved versions of already established systems, like the XCS with tournament selection (?). The objectives of this paper are twofold: (1) We provide further performance results of GAssist and XCS on several interesting datasets. (2) We compare and investigate performance of the two systems revealing problem dependencies, suitability of the respective approaches, as well as over-fitting or over-generalization tendencies.