The Preliminary Research Of Naive Bayesian Classifier On Radon Prediction

Radon, colorless and tasteless, is the only natural radioactive inert gas,produced by α decay of natural radioactive elements radium. After intercepted bythe human bodies when inhaled by the respiratory system, radon and its decayproduct deposited in lung，the α radiation emitted in decay process could damagelung. Epidemiological studies indicate that radon and its decay product is secondfactor of inducing lung cancer. In order to reduce and control the harm of Radon tohuman health, and given that the domestic radon potential mapping still in its initialstage, predicting radon level is of crucial importance.Using Bayes method to predict the radon concentration, it is only needed toestablish the relationships between the radon concentration and the maininfluencing factors, and to ensure that all factors are independent of each other. Dueto its high computational efficiency and accuracy, the Bayesian classifier isgenerally used in various fields. In this paper, data mining tools, NaiveBayesclassifier of WEKA platform, is used to analyze the already classified influencefactors for radon concentration, and to forecast the radon risk level of domesticregion.The main contents of this article including:(1) To understand, learn the latest progress in radon potential mapping, allkinds of influencing factors of the radon concentration, as well as the basicprinciple and method about Bayesian classifier and its application domain with theintention of applying them to radon prediction.(2) To collect and organize data ofthe coastal areas, including radon-related data, geological data and relatedgeographic information.(3) Based on the related experimental data in some area inChina, to use data mining tools containing simple Bayesian classifier for classifiedradon prediction.After detailed analysis of the results of prediction experiments, the followingconclusions are drawn:(1) The selection of condition attributes is important. Lacking of some critical data leads to not ideal classification results. Inconsistency between geographicinformation and collected data affects the accuracy of the test results.(2) Manydifferent kinds of Bayesian classifier exist. To choose what kind of classifier is alsovery important. If cannot get a lot of promotion in the improved classifier, evenreduce the model accuracy, we can choose other Bayesian classifier or otherclassification algorithm to predict, and enhance the accuracy of classification.(3)After establishing the model, the WEKA software can quickly predict the results, itis a kind of high efficient and convenient data mining software. The software canalso edit your own classification algorithm, trying to establish a more effectivemodel.