基于机器学习的“三高”风险评估的研究与实现

发布时间：2018-05-14 12:14

本文选题：风险评估 + 三高　；参考：《北京邮电大学》2017年硕士论文

【摘要】：健康数据具有较高的价值,利用机器学习技术来挖掘健康数据中的信息能够为疾病预测、疾病风险评估、疾病诊断等带来新的思路。在某种意义上,机器学习在健康数据中的应用对于改善人类生活水平具有重要的作用。高血压、高血糖和高血脂及其并发症对人们的身体健康造成了严重的危害,我国“三高”的患病率正在逐年升高,目前已经成为了我国重要的公共卫生问题,有必要通过预防来降低其危害。但考虑到“三高”的发病机制较为复杂,很难通过临床来进行预测,因此有必要借助机器学习技术来挖掘“三高”健康数据中的信息。研究“三高”预测模型,进一步完成“三高”风险评估,实现“三高”的早预防、早治疗。总之,通过机器学习技术为“三高”风险评估提供科学依据,是本论文的研究目的和意义。论文研究工作主要围绕“三高”风险评估模型的设计,算法的选择,模型的改进和性能评估等方面展开。首先,结合哈佛癌症风险指数公式和Logistic回归模型提出了一个“三高”风险评估模型。然后,通过对“三高”肥胖相关数据的分析,研究肥胖相关的身体测量参数和“三高”之间的关系。最后,根据数据分析结果,结合决策树和Logistic回归算法提出了另一个“三高”风险评估模型,弥补了上一个模型的不足之处。通过对两个模型的性能进行比较,验证了后一个模型的可行性和性能的提升。论文最终通过实验验证了两个风险评估模型方案的可行性以及算法性能的提升。后一个模型和前一个模型相比,模型的准确率、特异度和灵敏度都获得了提升,尤其是女性高血糖风险评估模型的提升较为理想。本课题的研究成果对于“三高”的风险评估有一定的指导意义。
[Abstract]:Health data is of high value. Using machine learning technology to mine the information in health data can bring new ideas for disease prediction, disease risk assessment and disease diagnosis. In a sense, the application of machine learning in health data plays an important role in improving human living standards. Hypertension, hyperglycemia, hyperlipidemia and their complications have caused serious harm to people's physical health. The prevalence rate of "three high" in China is increasing year by year and has become an important public health problem in our country. It is necessary to reduce its harm by prevention. However, it is difficult to predict the pathogenesis of the "three high" because of its complexity, so it is necessary to mine the information in the "three high" health data by means of machine learning technology. The prediction model of "three high" is studied, and the risk assessment of "three high" is further completed, and the early prevention and treatment of "three high" are realized. In a word, the purpose and significance of this paper is to provide scientific basis for the risk assessment of "three High" by machine learning technology. This paper mainly focuses on the design of the "three High" risk assessment model, the selection of algorithm, the improvement of the model and the performance evaluation. Firstly, combining the Harvard cancer risk index formula and the Logistic regression model, a "three high" risk assessment model is proposed. Then, through the analysis of the obesity data, the relationship between the obesity related body measurement parameters and the three high levels was studied. Finally, according to the data analysis results, combined with the decision tree and Logistic regression algorithm, another "three high" risk assessment model is proposed, which makes up for the shortcomings of the previous model. By comparing the performance of the two models, the feasibility and performance improvement of the latter model are verified. Finally, the paper verifies the feasibility of the two risk assessment models and the performance improvement of the algorithm. Compared with the previous model, the accuracy, specificity and sensitivity of the latter model were improved, especially in the female hyperglycemia risk assessment model. The research results of this paper have certain guiding significance for the risk assessment of the "three High".
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R318;TP181

【相似文献】