Efficient data preparation techniques for diabetes detection

The development of new medical devices and database management systems have created a huge number of databases in biomedical science. The success of data mining on medical databases in the medical data is often affected by incomplete, sparse and inexact parameters. If data is incorrect, irrelevant and noisy, then the final outcome will not be reliable. In particular, the conventional approaches for diabetes detection suffer from lack of attention to data preparation and ignoring the appropriate features. In this paper, some efficient data preparation techniques are study by which effective features are selected for reducing the cost and enhancing the accuracy of diabetes detection. The techniques are conducted by the popular diabetes data set PID.