Özet:

Nowadays, there has been a rapid increase in data size due to developments in information and measurement technologies. Due to the high dimensionality, many analyses, such as classification and clustering, require the use of data reduction methods before analysis. Variable selection plays an important role in interpreting the classification and clustering results of the data set and obtaining successful results with fewer variables. In this study, the variable selection problem to be used in classification is considered a multi-criteria decision-making (MCDM) problem. The decision matrix was created by taking variables as alternatives and filtering-based variable selection algorithms as criteria. This decision matrix was analyzed using the Grey Relational Analysis (GRA) method, which is one of the MCDM methods. Thus, a new approach for ranking and selecting variables has been proposed. In addition, entropy, which is an objective method, is used for weighting the criteria. The proposed method is applied to the Wine and Ionosphere datasets, which are frequently used in the classification literature, and compared with the classification success of filtering-based variable selection algorithms. As a result, successful results were obtained with fewer variables using the proposed approach.

Anahtar Kelimeler: