Study On The Prediction Classification Algorithm Of Radon Potential In Soil

Radon is a natural radioactive gas,which is the decay product of radioactive element radium.The harm of radon to human body cannot be underestimated.Radon is extremely unevenly distributed in nature.Every day,indoors or outdoors,it is surrounded by human beings.Therefore,it is very important to evaluate the concentration level of radon,so that some effective measures can be taken to reduce the harm of radon to humans as much as possible.If choosing measurements in all regions and then find out the soil radon concentration grades,so the cost is very large,so to save costs and to adopt a relatively scientific radon concentration prediction method is necessary.The purpose of this paper is to find more scientific radon potential prediction algorithm.Naive Bayes method is a multivariate statistical method,is to use some samples of known to predict some samples of the unknown attribute,here for radon concentration level is unknown attribute,other factors influencing the radon concentration of known attribute.In this paper,the data of Zhongshan in Guangdong province were selected for the research of radon concentration prediction.Research before,the correlation between attributes is not considered,here consider the correlation between attributes,and improve the selection of attributes,compared the three prediction algorithm run results,the characteristics of different algorithms were analyzed.The results show that the accuracy of naive bayes algorithm is the same as that of the weighted bayes algorithm with correlation coefficient.Use other data may be there is no such high accuracy.But the trend is consistent.Each of the three algorithms has advantages and disadvantages.Naive Bayes is simple and stable,but the reality seldom meets its conditions and leads to poor accuracy of prediction;Correlation coefficient weighted algorithm takes into account the other attribute information and the correlation of radon concentration,so relatively more scientific,but it is still based on other attributes are independent of each other between the hypothesis,so rises to the accuracy of prediction is not obvious;Naive Bayes based on mutual information,there are more factors that can be used to the information that cannot be seen directly,but the structure is complex and the information is not intuitive.