Titova E. Methods for building and evaluating aggregated association rules in intelligent databases

This thesis is devoted to solving an actual problem of the effectiveness of searching logical dependencies in databases on the basis of developing methods for building and evaluating aggregated association rules. Forming knowledge patterns in data in the form of association rules is one of the main directions in intelligent data analysis development. A method for association rule generation in the case of non-binary features using a cover tree has been further developed. A comparative analysis of simple and aggregated association characteristics has been carried out. Theorems necessary for calculating aggregated association rule characteristics have been stated and proven. A method for decomposing an aggregated association rule into simple associations has been suggested. A method for evaluating the information value of an association rule has been developed. A formula for calculating the integral characteristic of an association rule, its information value, has  been obtained. A comparative analysisof the integral characteristic and standard characteristics of an association rule has been carried out.