Evaluation of regression methods for predicting molecule concentrations from voltammetric data

The application of machine learning regression technology to the field of electro- chemical bio-sensing is investigated in this thesis. Experimental data is collected from voltammetry measurements of dopamine and ascorbic acid using tetrahedral amorphous carbon (ta-C) electrodes. Fixed concentrations of dopamine and ascorbic acid in a phosphate buffer solution are measured. And the data is used to train and test in regression algorithms. Feature extraction and dimension reduction is applied to the experimental data as well. The agreement is shown rationally between the actual concentration and the concentration predicted by gradient boosting and support vector machine with polynomial kernel from the experimental results. Cross validation is applied for each models. For the test dataset, the support vector regression algorithm with polynomial kernel achieves R2 score above 0.84. It shows us the potential ability to use machine learning algorithms in the detection and identification of different molecules in the future study