以機器學習法探討心血管疾病之危險因子

心血管疾病是全球人類首要死因，及早發現是很重要的。因此，本研究主要在探討以機器學習法建立心血管疾病預測模式時，危險因子的重要性，以及其對心血管疾病的解釋能力，希望未來能對提升預測模式的預測效能有所幫助。本研究以Kaggle提供的心血管疾病資料集，探討心血管疾病預測模式採用之危險因子。研究結果，被採用之危險因子按其影響大小排序，依次為：胸痛型態－無症狀（相較於非心絞痛）、空腹血糖－高、運動引發的心絞痛－是、運動引起的心電圖ST段壓低、性別－女，以及心電圖ST段的斜度－上升（相較於平坦）；其中，前四項因子的影響是正向的，表示罹患心血管疾病的可能性增加；末二項因子的影響是負向的，即罹患心血管疾病的可能性降低。預測模式對新樣本患心血管疾病與否的解釋能力（R^2）＝0.464。因此，建議未來進一步研究時，應納入其他可能的重要危險因子，以助於提升模式之預測效能。

關鍵字

心血管疾病；危險因子；預測；機器學習；羅吉斯迴歸

並列摘要

Cardiovascular disease is the leading cause of death worldwide, so early detection is important. Therefore, this study mainly explores the importance of risk factors and their explanatory power for cardiovascular disease when using machine learning methods to establish a cardiovascular disease prediction model. It is hoped that the performance of prediction models for cardiovascular disease can be improved in the future through this study. We use the heart disease dataset obtained from Kaggle to explore risk factors adopted by the cardiovascular disease prediction model. According to the result obtained, the risk factors used are, ranked in order of their impact, chest pain type-asymptomatic (compared to the type of non-anginal pain), fasting blood sugar-high, exercise-induced angina-y, exercise-induced ST-segment depression, sex-female, and ST-segment slope-up (compared to the type of flat). Among them, the influence of the first four factors is positive, indicating that the possibility of suffering from cardiovascular disease increases; the impact of the last two factors is negative, and the possibility of suffering from cardiovascular disease decreases. The prediction model's explanatory power (R^2) for new samples is 0.464. It is suggested that investigating other potentially important risk factors to help improve the predictive power of models is needed.

並列關鍵字

Cardiovascular disease ； Risk factor ； Prediction ； Machine learning ； Logistic regression

主題瀏覽