机器学习和数据科学方法用于USA的CDC死亡率数据的趋势和预测因素分析
Machine Learning and Data Science approach towards trend and predictors analysis of CDC Mortality Data for the USA
摘要:
死亡率研究是根据所提供的数据和条件得出结论的任何国家的一个活跃研究领域。为了使用机器学习和数据科学实践基于数据直觉得出结论,领域知识是必不可少的,但不是强制性的技能(尽管仍然需要一些知识)。进行该项目的目的是根据所提供数据集的统计数据得出结论,并使用监督或无监督学习算法预测数据集的标签。该研究得出结论(基于抽样)预期寿命(不分性别)及其主要趋势;人们的婚姻状况也影响他们每个人的死亡频率。该研究还有助于发现,由于更多的分类和数值数据,异常检测或采样不足可能是一个可行的解决方案,因为有可能比其他类别标签更多。这项研究表明,机器学习预测对数据的可行性并不像它可能是显而易见的。
英文摘要:
The research on mortality is an active area of research for any country where the conclusions are driven from the provided data and conditions. The domain knowledge is an essential but not a mandatory skill (though some knowledge is still required) in order to derive conclusions based on data intuition using machine learning and data science practices. The purpose of conducting this project was to derive conclusions based on the statistics from the provided dataset and predict label(s) of the dataset using supervised or unsupervised learning algorithms. The study concluded (based on a sample) life expectancy regardless of gender, and their central tendencies; Marital status of the people also affected how frequent deaths were for each of them. The study also helped in finding out that due to more categorical and numerical data, anomaly detection or under-sampling could be a viable solution since there are possibilities of more class labels than the other(s). The study shows that machine learning predictions aren't as viable for the data as it might be apparent.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。