Applying Machine Learning to Predict Electron Transfer Kinetics From Voltammetry Experiments

Electrochemical measurements of surface-bound kinetics have matured in recent decades to become a

very precise and sensitive field of analytical techniques. Electronic sensors, biochemical devices, fuel cells,

and many more modern developments rely on the accuracy of many electrochemical techniques.

However, some methodologies of analyzing electrochemical data currently rely on subjective and time-

consuming fitting techniques. Here we demonstrate that various machine learning methodologies can be

leveraged to predict surface-bound kinetic fitting parameters for electron transfer in a complex biological

system with reliable accuracy. Training data from experimental work is used to train various models, which

we optimize and contrast against one another. Models including Gaussian process regression, randomized

forests of decision trees, and ensemble methods are of primary focus. This work establishes a foundation

for the implementation of such machine learning techniques with non-idealized experimental data,

producing models in the 90-99% accuracy range with a relatively small initial dataset