The subject of the dissertation is the techniques for reducing training data (Data Reduction Techniques \u2013 DRTs) in classification problems. Specifically, these techniques, known as Prototype Selection (PS) algorithms and Prototype Generation algorithms, constitute a preprocessing step for training data with the aim of effective instance-based classification. Generally, these techniques are applied to training data that is organized in rows and columns and are capable of reducing the size of the data while maintaining high classification accuracy and reducing computational cost.<\/p>\n
The aim of the dissertation is to develop effective techniques for reducing complex training data, namely data related to data streams [1], distributed data [2],[3], data in non-metric spaces [4], data in complex structures (e.g., graphs), etc. Additionally, the development of data reduction techniques that take into account the phenomenon of concept drift [5] that appears in data streams is also a field of research. This phenomenon occurs when, over time, the distribution of data in space and classes changes.<\/p>\n