Consistency and reproducibility of attribute extraction by different machine learning systems

Machine learning systems as tools for intelligent data analysis are used for extracting attributes relevant for prediction of defined outcomes. The aim of the paper was to compare two machine learning (ML) systems and propose method for intra- and inter-testing of consistency and reproducibility in attributes extracted from real dataset. See5 and FMLS extracted relevant attributes from real medical dataset. Comparison of results of both systems shows that: accuracy and sensitivity are nearly the same for both FMLS and See5 when FMLS was forced by attributes extracted by See5, and little bit lower when FMLS used its own extracted attributes; specificity is nearly the same for both FMLS and See5 when FMLS was forced by attributes extracted by See5, and much more higher when FMLS used its own extracted attributes; both of ML systems show intra-testing consistency and does not show any inter-testing consistency.