The Application of Machine Learning Algorithms for Classifying Different Artificial Raman Datasets with Different Numbers of Glucose Levels

The application of machine learning technology for invasive diabetes diagnosis has become a research trend in the medical sector in recent years. Recent research has demonstrated a relationship between human glucose levels and Raman spectroscopy. In this research, we utilize the Raman spectroscopy of a glucose fluid sample to detect the glucose level. We create several glucose-liquid samples with different mixed rates between glucose and pure water to simulate the variety of glucose levels in human blood. Then, the Raman spectroscopy of each sample is obtained by using a spectrometer from Vietnam National University. Several simple machine learning models with different configurations are utilized to classify glucose levels from Raman data-samples. We create two datasets, including a spare and a dense dataset, based on the minimal distance of value between glucose levels. Then, we were able to extract an association between the glucose level in the fluid and the Raman pattern. We employed three machine learning models with 2 configured versions each model, consists of Ensemble learning and Support Vector Machine. The outcomes are completely optimistic, with high accuracy for predicting the glucose level of the sample, and they demonstrate the effectiveness of machine learning for classification.