Automatic Support for the Development of Classification Systems
Buch in deiner Nähe kaufen
Beschreibung
Developing pattern recognition systems is a complex task. Despite the availability of diverse data mining toolboxes, non-experts are usually not able to select the proper classification algorithm, suitable pre-processing steps, and optimal parameter values for their dataset at hand. The objective of this thesis is to enable non-experts to develop high-accuracy classification systems and to avoid time consuming trial-and-error processes. Additionally, experts are supported to reduce the development time of new classification systems and to minimize the risk of bad solutions. Several approaches incorporating experience gathered from previous learning tasks and applying machine learning algorithms for improving major steps of developing a classification system are presented. Novel methods providing automatic support in algorithm selection, feature selection, and parameter optimization for the users dataset are proposed.
Autor*in
Matthias Reif