Empirické porovnání komerčních systémů dobývání znalostí z databází

The presented work "An Empirical Comparison of Commercial Data Mining Tools" deals with data mining tools from world's leading software providers of statistical solutions. The aim of this work is to compare commercial packages IBM SPSS Modeler and SAS Enterprise Miner in terms of their specification and utility considering a chosen set of evaluation criteria. I would like to achieve the appointed goal by a detailed analysis of selected features of the surveyed software packages as well as by their application on real data. The comparison is founded on 29 component criteria which reflect user's requirements regarding functionality, usability and flexibility of the system. The pivotal part of the comparative process is based on an application of the surveyed data mining tools on data concerning meningoencephalitis. Results predestinate evaluation of their performance while analyzing small and large data. Quality of developed data models and duration of their derivation are stated in reference to the use of six comparable data mining techniques for classification. Small data more likely comply with IBM SPSS Modeler. Although it produces slightly less accurate models, their development times are much shorter. Increasing the amount of data changes the situation in favor of competition. SAS Enterprise Miner manages better results while analyzing large data. Considerably more accurate models are accompanied by slightly shorter times of their development. Functionality of the surveyed data mining tools is comparable, whereas their usability and flexibility differentiate. IBM SPSS Modeler offers apparently better usability and learnability. Users of SAS Enterprise Miner have a slightly more flexible data mining tool at hand.