Abstract:
Association rule mining as one of the descriptive tasks of data mining aims to extract interesting correlations, frequent patterns or associations among sets of items. Association rules mining became the primary tool for improving decisions in all aspects of life. This dissertation proposes an Intelligent Association Rule Mining Framework (IARMF) that guides inexperienced users through the process of selecting the best technique for their needs, which produces interesting rules.
Four mining tools namely Weka, Orange, Tangara and Knime were explored, regarding their association rule mining algorithms used and the employed interestingness measures. The researcher performed a new method to rank the suitability of measures to the type of the dataset. Experiments were carried out on three data sets from different domains. The experiments implemented the Apriori algorithm on Weka and Orange. Setting different values for the tool’s parameters, the researcher got different results. The method then selects the measure that gives the most consistent rankings than the previous ranking.
Users with different knowledge and expertise tend to extract frequent patterns for their own uses. However users need to acquire the knowledge of using the available tools and mining algorithms. IARMF is a menu-driven, user-friendly framework that can be used by inexperienced users and researchers who wish to fine-tune parameters of their choice. The Sudanese Transplanted Kidney dataset was constructed from the records at the Sudanese Kidney Transplantation Society and Ahmed Gasim Hospital. Experimental evaluation of the proposed framework on the constructed dataset reveals performance that is significantly better than the traditional approaches. Preliminary results from a prototype for the proposed framework show quite useful outcomes and opened up a wide range of interesting future research opportunities.
