基于朴素贝叶斯算法的氡潜势预测方法研究

发布时间：2018-05-31 23:27

本文选题：氡潜势预测 + NB算法　；参考：《中国地质大学(北京)》2017年硕士论文

【摘要】：现如今人们愈来愈注重环境对人类健康的影响,许多科研工作者已经展开了对环境方面的调查和研究。与人类生命相关的空气中,还存在着一些天然放射性的气体,例如氡。氡是铀衰变而来的,普遍存在于岩石,土壤和空气中。氡对健康有危害,有可能导致肺癌,故在高氡水平地区全面研究氡风险是有必要进行的。中国幅员辽阔,人口也比较稠密,若在所有地区进行氡测量将会是一项浩大的工程,这在目前是难以实现的。朴素贝叶斯分类器由于其算法简单易懂,分类效率高效稳定,并且有坚实的理论基础,在研究中将朴素贝叶斯分类器添加到氡潜势预测中将有效减少测氡工作量。本论文以广东中山地区56个有效测点的氡测量数据以及相关参数为研究范围,首先对朴素贝叶斯算法(NB)和基于相关系数的加权朴素贝叶斯算法(WNB-CC)的理论进行了深入研究,利用MATLAB软件编写工作代码并进行了正确性验证;而后对56个有效测点的参数进行了分步骤等级划分,最后确定了岩性、土壤氡析出率、铀含量、土壤氡浓度的等级划分标准,同时对NB算法和WNB-CC算法的预测概率进行了比较,得到WNB-CC算法的预测概率总体上高于NB算法的预测概率;随后对WNB-CC算法的单点预测结果的预测失败点进行了分析。最后论文对未知点预测的可接受性进行了分析,结合NB算法和WNB-CC算法单点预测结果,总结了未知点预测的可接受标准。在此基础上,论文对中山地区的未知点进行了预测,结果可接受。本文研究基于朴素贝叶斯算法的氡潜势预测方法,旨在减少大量的氡实地调查工作,也为后来人提供一些参考。
[Abstract]:Nowadays, people pay more and more attention to the impact of environment on human health. There are also natural radioactive gases, such as radon, in the air associated with human life. Radon is the decay of uranium and is widespread in rocks, soil and air. Radon is harmful to health and may lead to lung cancer, so it is necessary to study radon risk in high radon level areas. China is a vast and densely populated country. Radon measurement in all regions would be a huge project, which is difficult to achieve at present. Since the naive Bayesian classifier is simple and easy to understand, the classification efficiency is efficient and stable, and has a solid theoretical foundation, adding naive Bayesian classifier to radon potential prediction will effectively reduce the radon measurement workload. In this paper, the radon measurement data and related parameters of 56 effective measuring points in Zhongshan area of Guangdong Province are taken as the research scope. Firstly, the theory of naive Bayes algorithm and weighted naive Bayesian algorithm based on correlation coefficient (WNB-CC) are studied in depth. The working code is compiled and verified by MATLAB software, and then the parameters of 56 effective measuring points are classified step by step, and the classification criteria of lithology, soil radon exhalation rate, uranium content and soil radon concentration are determined. At the same time, the prediction probabilities of NB algorithm and WNB-CC algorithm are compared, the prediction probability of WNB-CC algorithm is higher than that of NB algorithm, and the prediction failure point of single point prediction result of WNB-CC algorithm is analyzed. Finally, the acceptability of unknown point prediction is analyzed, and the acceptable criteria of unknown point prediction are summarized by combining the results of NB algorithm and WNB-CC algorithm. On this basis, the unknown points in Zhongshan area are predicted and the results are acceptable. In this paper, the prediction method of radon potential based on naive Bayes algorithm is studied in order to reduce a large number of radon field surveys and to provide some references for future generations.
【学位授予单位】：中国地质大学(北京)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：X837;TP18

【参考文献】