Wearable blood glucose detection method, system and storage medium based on data enhancement

Abstract

The invention discloses a wearable blood sugar detection method, system and storage medium based on data enhancement, including: building a random forest model by collecting the real blood sugar value and corresponding feature matrix for a fixed period of time as a training set; The newly added feature matrix is ​​input into the random forest model to obtain the corresponding N predicted values ​​of blood sugar, and the N predicted values ​​of blood sugar are taken as real values ​​and combined with the original collected feature matrix as new training data to supplement the original data In order to continuously expand the number of training samples; build a blood sugar prediction model through the RNN model based on new training samples; input the features obtained by each signal collection when the user actually uses wearable devices for non-invasive testing into the blood sugar prediction model to obtain blood sugar predictions As a result, the invention improves the robustness of the model and effectively improves the prediction accuracy through data augmentation.

Description

technical field [0001] The present invention relates to the technical field of blood sugar detection, and more specifically, to a data enhancement-based wearable blood sugar detection method, system and storage medium. Background technique [0002] As one of the four major chronic diseases, diabetes has already become a major disease that threatens people's physical and mental health. According to the current level of medical technology, diabetes cannot be completely cured. Moreover, diabetes will not only bring complications and pain, but also affect people's diet and living habits for a long time, bring psychological distress, and constantly endanger people's physical and mental health. Therefore, people want to form the concept of "early prevention", "early detection" and "early treatment" of diabetes, and measuring blood sugar is an essential step. In the traditional method of detecting blood sugar, the blood sugar value is detected by using a biochemical blood glucose ...

AI Technical Summary

Problems solved by technology

The realization of wearable non-invasive blood glucose detection requires the collection of the user’s actual blood glucose data in the early stage, and then predicts the blood glucose value through the algorithm model based on the collected user characteristics. However, the original intention of non-invasive detection is to reduce the pain of patients, so it can only be collected in the early stage. Invasive detection data, and the key problem is - less data will lead to low accuracy of prediction results

Images