Big Data Research ( IF 3.3 ) Pub Date : 2020-12-17 , DOI: 10.1016/j.bdr.2020.100176 Kiichi Tago , Shoji Nishimura , Atsushi Ogihara , Qun Jin
With the surge in popularity of wearable devices, collection of personal health data has become quite easy. Many studies have been conducted using health data to estimate the onset and progression of illness. However, life habits may vary among individuals. By analyzing the life cycle from health-related data, conventional studies may be improved. This study proposes a new approach to improving diagnosis estimation by considering the life cycle analyzed from health-related data. The periodic span of the life cycle is estimated via autocorrelation analysis. In the range of the periodic span, dimension reduction for health data is performed by principal component analysis, and health features are extracted and used for diagnosis estimation. In our experiment, we used personal health data and pulse diagnosis data collected by a traditional Chinese medicine doctor. Using six multi-label classification methods, we verified that a combination of pulse and health features could improve the accuracy of diagnosis estimation compared with that using only pulse features.
中文翻译:
考虑基于个人健康数据的生命周期的周期跨度,提高诊断估计率
随着可穿戴设备的普及,个人健康数据的收集变得非常容易。已经使用健康数据进行了许多研究,以估计疾病的发作和进展。但是,生活习惯可能因个人而异。通过从健康相关数据分析生命周期,可以改善常规研究。这项研究提出了一种新方法,通过考虑从健康相关数据中分析的生命周期来改善诊断估计。生命周期的周期性跨度是通过自相关分析估算的。在周期范围内,通过主成分分析对健康数据进行降维,然后提取健康特征并将其用于诊断估计。在我们的实验中 我们使用了中医收集的个人健康数据和脉搏诊断数据。使用六种多标签分类方法,我们验证了与仅使用脉搏特征相比,脉搏和健康特征的组合可以提高诊断估计的准确性。