Ant colony optimization of rough set for HV bushings fault detection

In this paper we propose the optimization of Rough Set method using ant colony for oil-impregnated paper bushings. Ant colony is used to discretize the training data set. The ant colony optimized rough set is compare to a rough set who's data is discretized using equal frequency bin (EFB). Ant colony optimized (ACO) rough set results show an improvement compared to the EFB. The ACO rough set has an accuracy 4% high than that of EFB rough set. Rules generated are only a third for ACO compared to EFB. Although ACO takes longer to train, it proves to outperform EFB in all other respects.