An Integrated Approach towards the Prediction of Likelihood of Diabetes

With the growth of Information and communication technologies, the health care industry is also producing extensively large data. For managing such large amount of data, an efficient knowledge discovery process is required. This field is developing fast and there is a big scope of early planning towards the treatment of large number of diseases. The planning can be done by developing some strategic solutions based on Data Mining for the treatment of the disease. Classification based on supervised learning is a technique of Data Mining which helps in predicting the label of unknown samples as Class. This is extremely popular technique of Data Mining by which the treatment of a disease could be planned at an early stage. Diabetes is one of the chronic diseases produces metabolism disorder in human bodies. Metabolism refers a chemical process in human body responsible for energy conversion and utilization. The diabetes with type 1 and type 2 indicates excess glucose level in the blood could be cured if regular precautions have been taken persistently under certain clinical guidelines. This paper performs classification on diabetes dataset taken from SGPGI, Lucknow (A super specialty hospital in Lucknow, Uttar Pradesh, India). It predicts an unknown class label for given set of data and helpful to find out whether the class label for the dataset under consideration would be of low risk, medium risk or high risk. The classifier is further trained on the basis of weights assigned to different attributes which are generated by means of expert guidelines. The accuracy of classifier is verified by kappa statistics and accuracy, evolution criteria for classifiers.