Dynamic Algorithm Selection for Data Mining Classification

Recommending appropriate classification algorithm for given new dataset is very important and useful task but also is full of challenges. According to NO-FREE-LUNCH theorem, there is no best classifier for different classification problems. It is difficult to predict which learning algorithm will work best for what type of data and domain. In this paper, a method of recommending classification algorithms is proposed. Meta learning tries to address the problem of algorithms selection by recommending promising classifiers based on metafeature. Dynamic Algorithm Selection (DAS) with knowledge base, focus on the problem of algorithm selection, based on data characteristic. Algorithm selection will be better by using DAS in knowledge discovery process. In this paper we discuss the DAS architecture with knowledge base and Recommendation parameter measure. We present the architecture of DAS approach and Analysis of K-similar dataset produced by knowledge base.