Search academic texts

Information × Registration Number 0211U004290, 0109U001144 , R & D reports Title High-performance information technologies of data mining for socio-economic events and processes popup.stage_title Head Rashkevych Yuriy Mykhaylovych, Registration Date 02-02-2011 Organization Lviv Polytechnic National University popup.description2 In the final report on a theme "High-performance information technologies of data mining for socio-economic events and processes" the results of creation and software implementation of neural network software package for Data Mining on the basis of Geometrical Transformation Model (GTM) have been presented. Noniterative neural network algorithms on the basis of GTM for solving of classification tasks have been developed and implemented in software. An information technology for solving of classification tasks on the rule "winner takes all" with "penalties and encouragements" matrix has been created. The experiments on usage of the neural network software package for solving of typical tasks of Data Mining have been carried out. The fundamental advantages of GTM neural network such as high speed of training and using, results repeatability, ability of operating both small and large data samples have been confirmed. An efficiency of using final simulating annealing optimization procedure to increase classification precision has been proved. Developed information technologies of decision making for e-commerce tasks (shop clients classification, e-auction eBay lots classification) allow increase efficiency of solving of Data Mining tasks with high dimensions (hundreds of attributes) and large training samples (tens of thousands vectors) in conditions of compactness hypothesis violation. A neural network method for compression of data from various sources without/or with regulated losses has been developed and implemented in software. A representation of compressed data in the fixed-point format to simplify implementation in hardware of high speed GTM neural networks has been offered. Product Description popup.authors Дорошенко Анастасія Володимирівна Купчак Марія Іванівна Медвідь Андрій Ярославович Мельник Степан Петрович Пелешко Дмитро Дмитрович Поліщук Уляна Володимирівна Рашкевич Юрій Михайлович Ткаченко Роман Олексійович Цимбал Юрій Вікторович popup.nrat_date 2020-04-02 Close