The invention relates to the technical field of non-invasive monitoring, in particular to a diabetes blood glucose fluctuation
trend prediction method based on foot temperature and
humidity change, which comprises the following steps: continuously collecting temperature,
humidity and motion data of multiple areas of a foot through a foot wearable device, extracting a
feature set related to blood glucose fluctuation from the preprocessed data, and calculating the blood glucose fluctuation trend according to the
feature set; and fusing the
feature set and context event information, inputting the fused feature set and context event information into a multi-task
deep learning prediction model, visually displaying a prediction result in the forms of a trend arrow, a probability interval and a
risk level, and triggering early warning in advance when a hyperglycemia or
hypoglycemia event is predicted to be about to occur. The method is completely noninvasive and high in compliance, prospective
trend prediction can be carried out based on existing
diabetic foot monitoring wearable equipment, individual physiological characteristics of a user can be continuously learned through a built-in personalized self-adaptive mechanism, dynamic calibration is carried out on the model through
incremental learning, and the prediction accuracy is improved. And the prediction precision is continuously improved along with the use time.