📄 中文摘要
随着代谢健康管理的进步，传统的反应式指尖采血已逐渐被预测性维护所取代。连续血糖监测（CGM）图表常常呈现出波动不定的状态，预测这些波动的趋势是一个经典的时间序列预测挑战，远非简单的移动平均所能解决。本文将深入探讨预测健康维护，利用混合的 LSTM-Transformer 架构处理噪声较大的 CGM 数据，识别不规则模式，并预测未来 30 分钟内的血糖波动。通过结合 LSTM 的局部特征提取能力与 Transformer 的全局上下文理解能力，我们能够实现更高精度的预测。
预测健康：构建 LSTM-Transformer 混合模型以预测 CGM 数据中的血糖波动
📄 English Summary
Predictive Health: Building an LSTM-Transformer Hybrid to Forecast Glucose Spikes with CGM Data
Managing metabolic health has evolved from reactive finger-pricks to predictive maintenance. Continuous Glucose Monitor (CGM) graphs often resemble a chaotic rollercoaster, making it a classic time-series forecasting challenge that requires more than just simple moving averages. This article delves into Predictive Health Maintenance, leveraging a hybrid LSTM-Transformer architecture to process noisy CGM data, identify irregular patterns, and forecast glucose fluctuations 30 minutes into the future. By combining the local feature extraction capabilities of LSTMs with the global contextual understanding of Transformers, we can achieve higher accuracy in predictions.