Choosing the best knowledge acquisition for dengue dataset based on rough set approach

Knowledge based system need quality knowledge to provide an accurate decision. Due to irrelevant and incorrect information of expert during knowledge acquisition, the information extraction from data mining approach can reduce the biases and misconceptions of multiple experts. Moreover, the real world data which is huge and incomplete has made it complicated to extract only the quality knowledge. Hence, this paper is to investigate the capability of rough set (RS) in generating quality knowledge for dengue dataset problem. Four RS reduct algorithms i.e Genetic Algorithm (GA), Johnson Reducer (JR), Exhaustive Calculation (EC) and Dynamic Reduct (DR) are compared. The comparisons are made based on the accuracy, rules quantity, and rules length. The dengue dataset contains 182 is obtained from Sungai Petani Hospital, Kedah, Malaysia. The result shows that the performances of all reduct algorithms have comparable accuracy and in several algorithms, they obviously generates more and longer rule. GA is the most accurate classifier with shorter rule. Similarly, JA produces comparable accuracy but distinctly generate lower number and shorter rule. EC and DR are capable to produce good accuracy but use more and longer rules to obtain the good accuracy. In conclusion, JA is chosen as the best method for knowledge acquisition for dengue dataset.