Parameter tuning, feature selection and weight assignment of features for case-based reasoning by artificial immune system

Case-based reasoning (CBR), a popular problem solving methodology in data mining, solves new problems by analyzing solutions for similar past problems. The many advantages of CBR include rapid learning, the ability to use numerous unrestricted domains, minimal knowledge requirements, and effective presentation of knowledge. However, a major difficulty when applying CBR algorithms is selection of appropriate parameter values, features and weight assignment of features, to avoid constructing poor models. Unfortunately, key CBR parameters, beneficial features and the weight assignment of features vary across different problems. This study developed an efficient CBR approach based on artificial immune system algorithm (AISCBR) to increase classification accuracy by improving parameter tuning, feature selection and weight assignment of features. The proposed approach was then compared with those of other studies using the same University of California, Irvine (UCI) data sets. The experimental results showed that the AISCBR can provide better performance than other existing methods, because higher classification accurate rates can be obtained.