Emerging Harris Hawks Optimization for the Environmental Prediction of Mycotoxins in Food-Virtual Water Samples: A Comparative Study of Nature-Inspired Algorithms

Abstract The need for qualitative determination of mycotoxins in food samples is of paramount importance in designing procedures for their prevention, as well as understanding their adverse effects on humans and animals. In this article, a machine learning technique, specifically support vector regression (SVR) hyphenated with two metaheuristic algorithms - Harris-hawks optimization (HHO) and Particle Swarm Optimization (PSO) models (i.e., SVR-HHO and SVR-PSO) was used to forecast the chromatographic behaviour of various classes of mycotoxins in food samples. Three different metrics were employed to anticipate the model's performance: mean square error (MSE), correlation coefficient (CC), and Nash-Sutcliffe efficiency (NSE). The simulation results showed that the M3 input variable combination demonstrated higher performance accuracy than M1 and M2 with both SVR and its hybridized versions (SVR-HHO and SVR-PSO) during both the training and testing stages. In general, the hybridized model, based on the used assessment measures, SVR-HHO performed better in the training and testing phases than the other two data-driven techniques. Overall, the results show that employing chromatographic techniques, machine learning and metaheuristic approaches can both accurately predict the qualitative characteristics of mycotoxins in food samples.