Artificial neural network and TOPSIS guided ultrasound extraction of Cassia absus L. seed phytochemicals using a natural deep eutectic solvent

Ultrasound-based techniques and biodegradable natural deep eutectic solvents (NaDES) have garnered increasing attention as sustainable alternatives to conventional extraction processes, due to rising environmental concerns. In this context, the present study aimed to develop and optimize an ultrasound–NaDES integrated extraction method for efficient recovery of bioactive phytoconstituents from Cassia absus L. seeds (CAs). Process optimization was done using response surface methodology (RSM) and artificial neural networking (ANN). The response factors included total phenolic content (TPC), total flavonoid content (TFC), DPPH radical scavenging activity (DPPH-RSA), and iron chelating activity (ICA). The NaDES used as extractant was choline chloride–glycerol (CCG), and the independent factors included ultrasound amplitude (%), sonication time (min), and solvent-to-feed ratio (SFR, mL/g). The Box-Behnken design-based experimental data were optimized as per RSM and ANN, with further validation through multi-criteria decision-making (MCDM) using Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) method. The predicted optimal conditions (60% amplitude, 6 min time, and 20 mL/g SFR) yielded 43.88 mg GAE/g TPC, 17.60 mg QE/g TFC, 43.70% DPPH-RSA, and 82.33% ICA. ANN outperformed RSM in terms of predictive accuracy, as indicated by lower AAD, RMSE, MAPE, and PPE values, and a higher R². Thus, combining RSM-ANN-MCDM with NaDES is an efficient strategy for optimizing phytochemical extraction.