Özet:

This paper reports the construction of an effective mechanism for synthesizing classification trees according to the fixed initial information in the form of a training sample for the task of recognizing the current state, as well as flood phenomena, of river basins. The built algorithmic classification tree could unmistakably categorize the entire training sample underlying the constructed classification scheme. Moreover, it would demonstrate minimal structural complexity by including components such as the algorithms for autonomous classification and recognition to serve the structure’s vertices. The devised method for building the models of algorithms’ trees makes it possible to operate training samples composed of a large amount of diverse information of discrete type. It ensures high model accuracy, the rational utilization of the system’s hardware resources in the process that generates the final classification scheme, thereby making it possible to build models with predetermined accuracy. The proposed approach to synthesizing the new recognition algorithms is based on a library of already known algorithms and methods. Based on the proposed concept of algorithmic classification trees, a set of models was built that ensured effective categorization and prediction of flood-related events across the Tisza river basin. The proposed indicators of data generalization and quality of the classification tree model make it possible to effectively represent the general characteristics of the model allowing their application to select the optimal algorithm tree from a set of random classification tree methods. The classification trees built have ensured the absence of errors on the data of the training and test sample and have confirmed the efficiency of the approach of algorithm trees Author Biography Igor Povkhan, Uzhhorod National University Narodna sq., 3, Uzhhorod, Ukraine, 88000 PhD, Associate Professor Department of Software Systems

Anahtar Kelimeler: