Using Interactive Decision Tree Modelsin Artillery Fuse Testing

Dariusz AMPUŁA; Dariusz AMPUŁA

Pobierz

Ta publikacja jest chroniona prawem autorskim. Dostęp do jej cyfrowej wersji jest możliwy po zalogowaniu.

Tytuł: Using Interactive Decision Tree Modelsin Artillery Fuse Testing ; Using Interactive Decision Tree Modelsin Artillery Fuse Testing

Abstrakt:

In the introduction, the concept of interactive trees is defined and the purpose of the study is presented. Then, the RGM-2 fuse is described, as are the results of its tests which served as a basis for building specific models. The types of ammunition in which this variation of an artillery fuse is used are listed. A method of building interactive classification trees, allowing the author of the model to interfere with its structure, is described as well. Models of interactive classification trees, such as C&RT, CHAID and XAID have been designed and built. For each model, a tree diagram, a predictor importance sheet, a risk assessment sheet, and a summary of the observed and predicted values are presented. The method of interacting with the constructed classification tree structures, whose task was to improve the designed models, is shown using the examples of two models. The analysis of the models built after the interaction has been performed and, based on the obtained results, the best designed model was selected.
; In the introduction, the concept of interactive trees is defined and the purpose of the study is presented. Then, the RGM-2 fuse is described, as are the results of its tests which served as a basis for building specific models. The types of ammunition in which this variation of an artillery fuse is used are listed. A method of building interactive classification trees, allowing the author of the model to interfere with its structure, is described as well. Models of interactive classification trees, such as C&RT, CHAID and XAID have been designed and built. For each model, a tree diagram, a predictor importance sheet, a risk assessment sheet, and a summary of the observed and predicted values are presented. The method of interacting with the constructed classification tree structures, whose task was to improve the designed models, is shown using the examples of two models. The analysis of the models built after the interaction has been performed and, based on the obtained results, the best designed model was selected.