Performance estimation of tubular solar still with a wicked rotating drum using DT, LR, and KNN techniques of machine learning

The decision tree (DT), linear regression (LR), and K-nearest neighbours (KNN) models were employed in this work to estimate the thermal performance of tubular solar still with a wicked rotating drum. These three models were developed using real-world experimental data and calculated values. This study used a dataset containing 95 experimental iterations in total. Five input parameters, including solar intensity, basin water temperature, wind speed, ambient temperature, and glass temperature, were used as the independent variables of the DT, LR, and KNN models, and two dependent variables, thermal efficiency and productivity, were predicted. The DT model was the most significant model due to its lowest error and most incredible R2 value compared to the LR and KNN model performances. The MAE, RMSE, and R2 values for the DT model were 0.566828, 0.85135, and 0.9602, respectively, with the model efficiency of 0.961, which is the most significant value compared to other models. These results suggest that the DT model is a good fit for forecasting the thermal performance of tubular solar stills.