Time Series Prediction of Medical Records Incorporating Stationary Personal Details

Abstract

Improved blood glucose monitoring techniques have emerged over the last century. Adequate glycemic control and minimal glycemic variability necessitate a perfect, accurate, and dependable glucose monitoring system. [2]. There is still research being done on blood glucose monitoring systems in order to find the best one. According to the research proposal, the goal is to model and predict multiple blood glucose time-series from different users efficiently from limited training data in order to control and model their blood glucose levels. Individuals must anticipate blood glucose levels in order to take preventive measures against health risks in good time. There are high quality highlights and plan expectation models for the past endeavors, which lead to low exactness because of incapable component portrayal and limited preparing data for each individual. According to the findings of this study, the best way to predict blood glucose levels is to use a multi-time-arrangement profound LSTM model (MT-LSTM). It uses an individual learning layer for customized forecast and naturally learns highlight portrayals and transient conditions of blood glucose elements by sharing information among different clients. MT-LSTM outperformed traditional predictive relapse models in assessments of 100 clients.