基于IoT的生活方式疾病分类的优化混合解决方案
An optimized hybrid solution for IoT based lifestyle disease classification using stress data
摘要:
压力、焦虑和紧张都是日常生活中的高危健康状态。以前,压力水平是通过与人们交谈并了解他们最近或过去的经历来确定的。通常,压力是由很久以前发生的事件引起的,但有时它是由未知因素引发的。这是一个具有挑战性和复杂的任务,但最近的研究进展提供了许多机会,自动化它。大多数这些技术的基本特征是皮肤电活动(EDA)和心率值(HRV)。我们利用加速度计来测量身体运动,以解决这一挑战。所提出的新方法采用测量受试者的心电图(ECG)、皮肤电值(GSV)、HRV值和身体运动的测试,以便提供用于使用网络物理系统检测现代压力生活方式疾病的低成本且节省时间的解决方案。这项研究为生活方式疾病分类提供了一种新的混合模型,可以减少执行时间,同时选择最佳的特征集合并提高分类准确性。所开发的方法是能够通过使用WESAD(可穿戴压力和影响数据集)数据集来处理类不平衡问题。新模型使用网格搜索(GS)方法来选择一组优化的超参数,并使用基于相关系数的递归特征消除(CoC-RFE)方法进行最佳特征选择和梯度提升作为估计器来分类数据集,从而实现高准确性,并有助于提供智能,准确和高质量的医疗保健系统。为了证明所提出的方法的有效性和实用性,其性能相比,其他成熟的机器学习模型。
英文摘要:
Stress, anxiety, and nervousness are all high-risk health states in everyday life. Previously, stress levels were determined by speaking with people and gaining insight into what they had experienced recently or in the past. Typically, stress is caused by an incidence that occurred a long time ago, but sometimes it is triggered by unknown factors. This is a challenging and complex task, but recent research advances have provided numerous opportunities to automate it. The fundamental features of most of these techniques are electro dermal activity (EDA) and heart rate values (HRV). We utilized an accelerometer to measure body motions to solve this challenge. The proposed novel method employs a test that measures a subject's electrocardiogram (ECG), galvanic skin values (GSV), HRV values, and body movements in order to provide a low-cost and time-saving solution for detecting stress lifestyle disease in modern times using cyber physical systems. This study provides a new hybrid model for lifestyle disease classification that decreases execution time while picking the best collection of characteristics and increases classification accuracy. The developed approach is capable of dealing with the class imbalance problem by using WESAD (wearable stress and affect dataset) dataset. The new model uses the Grid search (GS) method to select an optimized set of hyper parameters, and it uses a combination of the Correlation coefficient based Recursive feature elimination (CoC-RFE) method for optimal feature selection and gradient boosting as an estimator to classify the dataset, which achieves high accuracy and helps to provide smart, accurate, and high-quality healthcare systems. To demonstrate the validity and utility of the proposed methodology, its performance is compared to those of other well-established machine learning models.
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。