Data Mining Application in Diabetes Diagnosis using Biomedical Records of Pathological Attribute

Data mining aims at discovering knowledge out of data and presenting it in a form that is easily compressible to humans. It is a process that is developed to examine large amounts of data routinely collected. Artificial intelligence technique like fuzzy, ANN etc are currently used for solving a wide range of problems in different application domain for decision based model designing. These systems allows us to introduce the learning and adaptation capabilities hence such type of framework has been used in several different process of diagnosis of disease. It helps in creating computational paradigm that provides a mathematical tool for dealing with the uncertainty and the imprecision typical of human reasoning. Relational between symptoms and risks factors for Diabetic based on the experts medical knowledge is taken and also related complications or due to some common metabolic disorder it may lead to vision loss, heart failure, stroke, foot ulcer, nerves. In this work review is provided on various methods which are is considered in analysis where symptoms observed in the patient and relation representing the medical knowledge that relates the symptoms in set S to the diseases in set D to diagnose the set B of the possible diseases of the patients can be inferred by means of the compositional rule of inference. It has been observed that Neural Networks are efficiently used for learning membership functions, fuzzy inference rules and other context dependent patterns, fuzzification of neural networks extends their capabilities in applicability [17]. The classifier is the heart of the automatic diagnosis system. The reliable classifier should diagnose the disease at as high accuracy as possible even though there are many uncontrolled variations. In literature, different classifiers have been proposed for automatic diagnosis of PD. The NNs and adaptive neuro fuzzy classifier with linguistic hedges (ANFIS-LH) are investigated for automatic diagnosis of PD. The performance of probabilistic neural network (PNN) for automatic diagnosis of PD is evaluated. SVM classifier is also investigated for the same goal. NNs have some drawbacks such as need of long training time and uncertainties in activation function to be used in hidden layer, number of cells in hidden layer, and the number of hidden layer. In case of SVM, type of kernel function and penalty constant and so forth affects the classification performance. If these parameters are not appropriately selected, the classification performance of SVM degrades. Similarly, the performance of ANFIS depends on type and parameters of membership function and output linear parameters.