Identifying the suitable reduction technique for mining medical data

Abstract

In real world, organizations often have large amount of data that are stored in databases. The large size of data makes data analysis difficult as data are more complex in terms of number of attributes and number of objects. The use of a sufficient number of attributes and objects are one way to overcome the problem. There are many techniques that can be used to minimize the number of attributes in data mining. In this paper, three reduction techniques namely Genetic Algorithm (GA), Principal Component Analysis (PCA), and Johnson have been tested on medical domain using 5 datasets which obtained from UCI machine learning archive. The study examines which reduction technique is most suitable for medical datasets. In addition, the study also identifies the ranking of the three techniques based on percentage accuracy and number of selected attributes.