A Comparative Study of Various Data Transformation Techniques in Data Mining

This research paper presents a technique to select an ideal transformation technique of original and transformed features. The paper reviews about a comparative study of various data transformation techniques used in data mining which includes six types of transformation techniques-Wavelets, Genetic Algorithm and Wrappers, Identity transform, Program synthesis, Data refinement transformation, and Feature Selection technique. The feature selection technique is considered best as it utilizes Wavelets and Genetic Algorithm and Wrappers methods that employ classification accuracy as its fitness function. The selection of transformed features provides new insight on the interactions and behaviors of the features. This method is especially effective with temporal data and provides knowledge about the dynamic nature of the process. The comparative study from the feature selection technique demonstrates an improvement in classification accuracy, reduction in the number of rules, and decrease in computational time.