Prediction of Uptake of Carbamazepine and Diclofenac in Reclaimed Water-Irrigated Lettuces by Machine Learning Techniques

Abstract

Currently, due to the global shortage of water, the use of reclaimed water from the Wastewater Treatment Plants (WWTPs) for the irrigation of crops is an alternative in areas with water scarcity. However, the use of this reclaimed water for vegetable irrigation is a potential entry of pharmaceutical products into the food chain due to the absorption and accumulation of these contaminants in different parts of the plants. In this work we carried out an analysis of five machine learning techniques (Random Forest, support vector machine, M5 Rules, Gaussian Process and artificial neural network) to predict the uptake of carbamazepine and diclofenac in reclaimed water-irrigated lettuces with the consequent saving of environmental and economic costs. For the different combinations of input and output, the prediction results using the of machine learning techniques proposed on the pharmaceutical components in reclaimed water-irrigated lettuces are satisfactory, being the best technique the Random Forest that obtains a model fit value (R-2) higher than 96.5% using a single input in the model and higher than 97% using two inputs in the model.