Study On Heart Disease Decision Based On Rough Set And RBF Neural Network

The massive medical data in modern intelligent medical system requires medical staff to make quick judgments based on patients’ medical records.At present,heart disease is a kind of complex disease,the symptoms of patients are many different,most of the causes of the origin of the disease can not be verified.Traditional medical decision-making method is difficult to give an accurate diagnosis in patients with symptoms much different as a result,effectively extracted from a lot of heart health data collection of attributes,with the aid of RBF neural network to complete the disease intelligent diagnosis,Neighborhood rough set attribute forward search reduction algorithm is used to deal with medical decision instance,and the minimum reduction set is obtained,and the need to keep the classification ability are not affected.The extracted attribute reduction set was used as the input samples of the two heuristic algorithms to optimize the radial basis neural network,and the accurate calculation of patients was made and the decision analysis results were given.Main research contents of this thesis are as follows:(1)A fast algorithm of neighborhood rough set attribute selection is used to deal with uncertainty knowledge by introducing characteristic parameters and threshold value of importance lower limit.For any subset neighborhood radius is calculated after the standard deviation,sample neighborhood information table,the results show that the decision attribute any attribute subset of upper and lower approximation and judgment for sample concentration of each sample neighborhood decision attribute values are the same,find the most-positive value and the corresponding attributes,repeating attribute reduction results are obtained.Neighborhood rough sets do not need discrete preprocessing,retain the integrity of original data,and have a new measurement method for the processing of uncertain knowledge.(2)Using ant colony algorithm for RBF neural network parameter optimization,ant colony size is determined,each ant fitness value calculation,and gives the initial pheromone concentration,based on the node transition probability choose path forward,to prevent the accumulation of pheromone concentration on road,set the pheromones are volatile,and decreased with the increase of time,in the process of iteration by positive feedback mechanism to find out the best route.The original data and attribute selection fast algorithm reduction data in the previous work were used as the training samples of ant colony algorithm optimization RBF neural network.Through comparative analysis of experiments,the combined algorithm improved the efficiency of medical decision-making for heart disease data.(3)Particle swarm optimization RBF neural network algorithm is proposed,the algorithm in the search space,the initialization of particle swarm,the mapping way optimization algorithm and the neural network model is established,the calculation of individual particle fitness value and the initial position in the current position,through the particle’s position and speed adjust behavior,reach the limit numerical iteration times and conditions,iteration to get the optimal particle global extremum,complete the global optimization.The experimental results show that,compared with the single algorithm model,the accuracy 、 running time and other evaluation indexes of the neighborhood rough set are improved by the attribute reduction set and particle swarm optimization RBF neural network.