Властивості асоціативних залежностей у аналізі даних

Abstract

UK: У статті наведено результати дослідження властивостей асоціативних залежностей та можливостей їх ефективного агрегування. Розрозбено метод виявлення асоціативних залежностей широкого класу у великих наборах даних. EN: This paper describes the results of research in the field of associative dependencies properties and effective aggregation possibilities. Also it briefly describes the developed method of special class of associative dependencies detection in large data volumes. The main idea of this research is aggregation of elementary associative dependencies into more complicated once. This approach gives good performance results and allows processing data volumes with millions records. Current paper shows how it is possible to define algebra of associative dependencies with few main operations and rules of inference, taking place in such algebra. The rule set completeness is also proven here to be sure that no rules are lost during inference. The outcome of described theory is highly effective data analysis method, capable to detect wide range of associative dependencies in relational data.