Effects of chromatographic conditions on retention behaviour of different psychoactive agents in high-performance liquid chromatography: A machine-learning-based approach

Background and Aims: High-pressure liquid chromatography (HPLC) data on the effects of various chromatographic conditions on the retention behaviour of three different psychotropic drugs; clonazepam, diazepam, and oxazepam) were considered for simulation using a machine learning approach. Methods: For the simulation of selected psychoactive compounds using HPLC, different machine learning techniques were used in this study: adaptive neuro-fuzzy inference system, multilayer perceptron, Hammerstein-Weiner model, and a traditional linear model in the form of stepwise linear regression. Four evaluation criteria were used to assess the effectiveness of the models: coefficient of determination, root mean squared error, mean squared error, and correlation coefficient. Results: The results show that machine learning approaches, especially multilayer perceptions, are more reliable than classical linear models with an average coefficient of determination value of 0.98 in both calibration and validation phases. Conclusion: The performance results also demonstrate that these models can be improved using additional approaches, such as hybrid models, ensemble machine learning, evolving algorithms, and optimisation techniques.