Research On Application Of Support Vector Machine To Chinese Medical Diagnosis

Although Chinese medicine has been considered valuable in applications, its theories and methods have not been accepted widely by modern medicine as it has little precise data derived from scientific experiments to prove its validity and safety. It will be beneficial for the development and extension of Chinese medicine if we can discover some relationship between clinical symptoms and diagnosis. It is very difficult to achieve this goal manually because of complicated knowledge of Chinese medicine and volumes of data.Being a class of automatic and intelligent data analysis techniques, data mining, also called KDD, which aims at extracting novel and useful knowledge from large volumes of data, has emerged rapidly in recent ten years. Support Vector Machine has become one of rising data mining techniques because of its excellent theory.In this thesis we firstly overview data mining techniques, making a brief description about the concept, basic model, typical structure, and some popular techniques of data mining. Then we focus on support vector machine and discuss its theory foundation, basic concepts, and crucial techniques of support vector machine. With these background, we further study several generally algorithms about support vector machine, especially for Plait's SMO (Sequential Minimal Optimization) algorithm. Points out an important source of inefficiency in Platt's Sequential Minimal Optimization (SMO) algorithm that is caused by the use of a single value, then present an improved SMO algorithm employed two threshold parameters. Experiments demonstrate the improved SMO algorithm performs faster than original SMO algorithm on UCI datasets and our nephropathy datasets.And a new preprocessing method using rough set theory is presented in this thesis, aiming at the characteristic of traditional Chinese medicine. The result of experiments is satisfied with us.hi the last part of this thesis, we propose a prototype data mining system based on support vector machine. The case study on a nephropathy dataset illustrate the promise of support vector machine based data mining techniques in Chinese medicine.