Research Of Disease Prediction Model Based On BP Neural And DS Evidence Theory

With the improvement of living standards, people look forward to get more highquality medical service. At present, computer technology has been widely used in all aspects of the medical field. However, the computer technology is mainly used to display, storage medical information of patient, and so on. The disease diagnoses are made still largely by doctors. Computer aided diagnosis is now in the beginning stage of development.With Computer technology applied to auxiliary diagnosis and forecast, it will produce vital significance for detecting and treating disease early, reducing misdiagnosis and so on, At present, the commonly used methods include time series forecasting method, Markov prediction method, artificial neural network, regression forecast method and so on. In addition, people try to combine multiple mathematical models for disease forecast, in order to improve the prediction accuracy. This paper gives a disease prediction model. It use BP neural network to predict the patient’s condition, and DS evidence theory is introduced to improve the prediction accuracy.Firstly, paper analyzes the characteristics of the BP neural network and DS evidence theory. BP neural network has the characteristics of strong self-organization,fault tolerance and adaptivity. But it is easy to fall into local minimum point and produce poor recognition and low credibility in the process of multi-objective prediction. The DS evidence theory can add the narrow differences between evidence together. When these differences are accumulated to a certain extent, they can be distinguished to improve the accuracy of prediction results, the advantages of the DS evidence theory can compensate for the BP neural network’s shortcoming of low prediction accuracy.In view of the characteristics of BP neural network and DS evidence theory, this paper puts forward a model combining BP neural network and DS evidence theory. It uses BP neural network as the disease forecasting model, forecast the diseaseaccording to the patient’s feature data, the prediction results will be synthetized with the DS evidence theory.To test and verify the correctness of the model, this paper uses the heart disease data to test the model. Firstly, the BP neural network is got from heart disease data.Then the data, divided into multiple sets and normalization treatment, is used as inputs of BP neural network to forecast, and multiple sets of output results are produced. The normalized output is a basic probability vector which DS evidence theory can synthesize. Finally, the basic probability vector is synthetized by the DS evidence theory. If synthetic results are expectant, they will be output, and the result is eventually the prediction of heart disease.Experimental results show that heart disease prediction combining the BP neural network and DS evidence theory greatly improves the prediction accuracy, and robustness of the algorithm is better. Applying this model to other disease forecast will provide broad application space to the wisdom medical treatment in the future.