Application of Machine Learning Models for Predicting Glucose-Level in the Pure Fluid with Algorithm for Reducing Data Dimension Based on Data Series Extraction

The phenomenon that glucose level of pure liquid is able to define patterns of Raman spectroscopy was demonstrated in several studies. Nevertheless, it is difficult to predict glucose level accurately by manual methods so machine learning techniques are proposed to support it. In the range of the report, we employ three simple machine learning models including Extra Trees, Random Forest, and SVM to predict glucose level from Raman spectroscopy of pure water-mixed fluid which we collected by infrastructures of Vietnam National University. In addition, the Raman data was simplified by dimension reduction algorithms based on handling data series. The results show the effectiveness of the machine learning models for predicting glucose levels as well as the reduction dimension algorithms for enhancing the performance of machine learning techniques.