• The modified k-means algorithm and its application to type-1 diabetes glucose data clustering

      Dai, Jin (University of Bedfordshire, 2010-11)
      Most of previous studies of were concentrated on data mining algorithms for type 2 diabetes patients. This study aims to design and implement a data mining algorithm to assist doctors to diagnose and analyze type 1 diabetes patients' condition. In order to achieve the aim of this study, data of glucose of the diabetes patients have been collected first. Mainstream data mining algorithms have been then studied and compared through literatures review. A K-means algorithm has been initially selected to be applied to deal with diabetes patients' data. However, there are three disadvantages of the K-means algorithm: a) The performance of the K-means algorithm tightly relies on the order of input data. b) Outliers can determine the performance of the algorithm. c) The data samples which fall into the overlap are difficult to deal with. Therefore, fuzzy logic techniques have been introduced to collaboratively work with the K-means algorithm. Experiments are to be carrying out in order to test and verify the proposed algorithm after the implementation of the software. The proposed algorithm and the software are going to be optimized in the nearly future.