Prediction of Photodetector Response Using Supervised Learning

Responsivity of a photodetector is a significant parameter which describes the response of the detector to a particular wavelength of light. The experimental determination of responsivity of the photodetector requires necessary light source of a particular wavelength. For wavelengths in deep UV region, it is difficult to get a light source, thereby making it difficult to estimate the responsivity of a photodetector experimentally. In this work, Supervised Learning algorithm has been used for predicting the response of a Graphene based photodetector in Field Effect Transistor architecture operating in deep UV range of wavelength. The dataset is collected from an experiment. Exploratory data analysis is done to understand the relationship between the different features of the dataset and the target variable which is the photocurrent flowing through the photodetector. Algorithms like Multiple Linear Regression and Polynomial Regression are used to build predictive models. It was observed that the fit with Polynomial Regression gave much better performance than of that with Multiple Linear Regression. This proves that the dependency of response of the photodetector is not strictly linear. The performance of the model is evaluated in terms of R2 score and mean square error.