Modelling logic mining: A log-linear approach

Logic mining has been widely used in many fields as an aid to extract logical rule that are significance to the data set. However, a previous study in Discrete Hopfield Neural Network (DHNN) formulated random attributes for the logic implemented in the logic mining. Hence, this study introduced statistical analysis which is log-linear that will be used in finding the best attributes for the logic that gives important effect to the outcome. This study will embed 2 Satisfiability based Reverse Analysis method with an approach of log-linear as the attribute selection method (2SATRA) and simulated by using several benchmark data sets. The capability of the log-linear integrated with the logic mining model can be investigated in the retrieval phase by using various performance metrics. In light of the outcomes, the proposed model able to achieve optimal performance as compared to the existing model. These findings indicate an improvement of logical rule as the symbolic language in DHNN and log-linear integrated with 2SATRA competent in doing logic mining.