The Prediction Model Of Stroke On Climate Factors By Multiple Regression

Stroke has a great threat on human health,and its incidence is gradually increasing trend in recent years.If we can analysis and predict the occurrence of the disease,it has important significance for people to take preventive measures and medical method in advance.At the same time,due to the birth of data mining technology,at home and abroad,a lot of data mining techniques have emerged to analyze the relationship between the occurrence of stroke and the climate environment,these studies show that the important connection with stroke and climate.This thesis is based on those studies,abandoning the existing simply techniques by linear regression or logistic regression to predict the relationship between stroke and climatic factors, making full use of the advantages and disadvantages of data preprocessing technology and a variety of regression analysis in data mining to get a more accurate prediction result.In this thesis,firstly I collect and analysis the data which collect from many hospitals in a city included stroke patients case information and the corresponding period of climate data by data preprocessing.It includes removing dirty data and unifying data format to get unified and valuable data.Multiple regression techniques are used to screen the climatic factors which have important influence on the occurrence of stroke from eight possible climatic factors,finally using nonlinear regression to analysis the screened climate factors and stroke incidence to get the prediction model for stroke and climatic factors.The experimental results show that the mean air pressure,maximum pressure,maximum temperature,average humidity and minimum humidity have important effects on the occurrence of stroke,this is in accordance with the research results in the medical field.At the same time,the prediction model is more accurate than the previous research methods,the decision coefficient factor R~2 which is used to evaluate the prediction equation is 0.973,it is more closed to 1 than the prediction models by the simple linear regression.This indicates that the experimental method is more accurate in this paper,it is very meaningful to establish the prediction model of stroke.