The Application Of Nonlinear Prediction Method In AIDS Therapic Effect Research

Directed at the clinical data (ACTG320) which come from AIDS therapic effect,the paper adopts quantitive prediction method to process the data, tries to find itsinternal regularity, roughly obtains the influence of different AIDS treatments onpatients' state of illness and finally predicts the effect of follow-up treatment.To begin with, the paper introduces the related background information about thetheme including the conventional AIDS therapies and the corresponding effects,makes it known that CD4 amount and HIV consistency are the two important indexesto measure AIDS patients' state of illness and elicits the main task of the paper, i.e., touse some mathematical prediction methods to simulate the nonlinear relation betweenCD4 and HIV.To fulfill the prediction of AIDS therapic effect, the paper sets up a BP neuralnetwork, using the initial values of CD4 amount and HIV consistency and treatmentlength as input value, the variations of CD4 amount and HIV consistency as outputvalue. And then, through the grouping and pretreatment of the initial data, the trainingof the network and the research on the generalization ability of the network, the paperdraws out a rational network model and applies it to the prediction of the effect ofAIDS long-term treatment. Aiming at different AIDS patients, the model predicts theeffect of the follow-up treatment and offers a time for terminating the treatment aheadof schedule for those who have bad therapic effect.Secondly, the paper establishes a regression backing vector machine model usingthe initial values of CD4 amount and HIV consistency and treatment length as imputvalue and the variations of CD4 amount and HIV consistency as output value. Thepaper applies the trained regression backing vector machine to the prediction of AIDStherapic effect through the study training of the sample.Comparing the applications of BP neural network and regression backing vectormachine in the research of AIDS therapic effect, the paper finds that the predictionmodel of AIDS therapic effect based on backing vector machine has superiority overBP neural network in terms of estimated accuracy, predictive ability and so on.