Improvement Of Random Forest Algorithm And Its Application In Early Warning Model Of Chronic Diseases

Medical data show that chronic diseases have become one of the most serious diseases that endanger human health.An important characteristics of chronic diseases is that it is difficult to make accurate diagnosis in advance,but there are certain rules for its occurrence and development.The diagnosis of chronic diseases is essentially a problem of data classification in machine learning.With the help of machine learning technology,various rules and connections for the diagnosis of chronic diseases can be excavated to help doctors establish disease early warning models.In this paper,a random forest algorithm based on extreme learning machine with optimized linear combination kernel function is proposed and used in the classification of chronic diseases to further improve the diagnostic accuracy of chronic diseases and provide a reference for doctors in clinical diagnosis.The classification results of chronic diseases by researchers using different models show that the model based on support vector machine(SVM)and artificial neural network(ANN)have better classification performance;however,the model parameters are difficult to select,the performance of single classifier is limited,the training speed is slow,and it is not applicable to the mass medical data environment.In order to solve above problems,a random forest classification algorithm based on extreme learning machine with optimized linear combination kernel function is proposed.The main research content is as follows:1.To solve the problem of irregular and uneven medical data,this paper uses the linear combination of RBF kernel function and polynomial kernel function as the kernel function of the kernel extreme learning machine.The kernel type of the classification model based on the kernel method has a strong relationship with the training data.The linear combination kernel can not only make full use of the advantages of each kernel function to fully adapt to the training data,but also reduce the influence of the kernel type on the classification performance of the model.To select model parameters automatically,the particle swarm optimization algorithm(PSO)is used to adjust the model parameters adaptively,and the iterated global optimal parameters can greatly improve the model classification performance.2.In order to break the performance bottleneck of the single classifier and reduce the training time of the model,this paper utilizes the kernel extreme learning machine with extremely fast learning speed as the base classifier of the random forest algorithm,and uses the order-adding particle swarm optimization approach to optimize the random forest classification algorithm based on extreme learning machine.This optimized new model further improves the classification performance and reduces the training time.3.A Map-Reduce parallel computing model is used to parallelize the proposed algorithm for the problem that the single machine version of the random forest algorithm can not process mass chronic disease medical data.4.This paper designed and carried out the experiment.The UCI breast cancer chronic disease dataset was used as experimental data,and popular classification algorithms such as optimized SVM,artificial neural network,extreme learning machine,original random forest,and unoptimized decision tree were used as experimental comparison objects.Experimental results show that the proposed method has better classification performance and lower time consumption.5.A chronic disease warning prototype system based on big data platform was implemented.The system includes functions such as data acquisition,preprocessing,modeling,intelligent diagnosis,and risk warning.