| 摘 要: 针对血糖水平预测及潜在风险预警,提出一种新型机器学习模型,它结合了变模态分解(VMD)、循环神经网络(RNN)和长短期记忆(LSTM)网络。该模型通过分解血糖信号并引入迁移学习策略,有效捕捉局部特征和长期依赖性,进而建立个性化的血糖预测模型,提高正常血糖水平的比例。结果表明,VMD-RNN-LSTM 模型在均方根误差(RMSE)方面较传统模型平均提升62%,在平均绝对误差上提升49%,拟合优度平均提升约30%。研究表明,VMD-RNN-LSTM 模型能有效提升1型糖尿病患者的血糖控制能力。 |
| 关键词: 血糖预测 变模态分解 循环神经网络 长短期记忆 时间序列 |
|
中图分类号: TP319.1
文献标识码: A
|
|
| Improving Blood Glucose Controlin Type1 Diabetes Patients UsingM achine Learning and Continuous Glucose Monitoring Systems |
|
|
(1. School of Health Science and Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China;
2.Shanghai MicroPort Lifesciences Co., Ltd., Shanghai 201203, China)
xq15873661230@163.com; iamyeping@163.com; liuxk1220@126.com; xdliu@microport.com
|
| Abstract: This study presents a novel machine learning model that integrates Variational Mode Decomposition(VMD), Recurrent Neural Network (RNN), and Long Shor-t Term Memory (LSTM) network for blood glucose level prediction and potential risk warning. The model effectively decomposes blood glucose signals and incorporates a transfer learning strategy to capture local features and long-term dependencies, thereby establishing a personalized blood glucose prediction model that increases the proportion of normal blood glucose levels. Results indicate that the VMD-RNN-LSTM model reduces the root mean square error by an average of 62% compared to traditional models, reduces the mean absolute error by 49% , and increases the goodness of fit by approximately 30% . The study demonstrates that the VMD-RNN-LSTM model can effectively improve blood glucose control ability in patients with type 1 diabetes. |
| Keywords: blood glucose prediction VMD RNN LSTM time series |
