Mining Association Rules from Semantic Web Data without User Intervention

With the introduction and standardization of the semantic web as the third generation of the web, this technology has attracted and received more human attention than ever. Thus, the amount of semantic web data is continuously growing, which makes them a rich source of useful data for data mining techniques. Semantic web data have some complexities, such as the heterogeneous structure of data, the lack of well-defined transactions, and the existence of typed relations between items. In this paper, a new technique named SWApriori is presented, which by using both entities and relations in the extraction of frequent itemsets, generates a new class of association rules (ARs) from semantic web data. The proposed technique by considering the complex heterogeneous nature of semantic web data, without any need to a domain expert, and without any data conversion to transactional data format extracts ARs from semantic web data directly. For evaluation, the proposed technique is applied to Factbook and DBPedia datasets. The experimental results demonstrate the ability of the proposed technique in extracting relational ARs from semantic web data by considering the mentioned challenges. Supplementary experiments show that the proposed technique can extract interesting patterns that are not discoverable by state-of-the-art association rule mining techniques.

Keywords:

Semantic Web , Association Rules , Data Mining , SWApriori