Visualization of domain and concept descriptions

The paper presents a method for visualizing data in machine learning. A visualizing method aims at representing the space of examples (attribute-value tuples) and the corresponding concept descriptions (induced if-than rules) which are both multidimensional with a dimension depending on the number of attributes. The method is based on the parallel coordinates method which was previously used for visualizing multidimensional geometry. Visualization of examples and rules with the parallel coordinates method enables the analysis of the space of examples and induced rules. The described visualizing method is connected to the machine learning system Atris and is currently postprocessing the results of the machine learning algorithm. Results of the experiments on a real-world domain illustrate the usefulness of the method for analysing machine learning problems.