A seminar titled "Stochastic Methods of Optimization in Data Mining Problems" will be presented by Jamolbek MATTIEV on Monday, 18. December 2017 at 16.00 at the Faculty of Mathematics, Natural Sciences and Information Technologies of the University of Primorska.
PLACE: FAMNIT-1-MP2 at 16:00
LECTURER: Jamolbek MATTIEV, Student of the Computer Science PhD Program UP FAMNIT
TITLE: Stochastic Methods of Optimization in Data Mining Problems
The importance of data mining is growing with the growth of the amounts of data that are automatically gathered from various sources. This fast growth demands the development of new algorithms and methods to cope with this, so called, big data. In this seminar we will discuss numerical algorithms for selecting optimal boundaries - intervals of feature values - for classified objects. We will evaluate our algorithms on real-life data and compare them to existing results. The methods for computing weights of nominal features in neural network models based on the product of inter-class similarities and cross-class differences will be presented. We will also discuss the method of extracting informative feature sets.