Using microwave-assisted extraction with advanced artificial intelligence models for predicting tannins in black pepper (Piper nigrum L.)

Abstract

Black pepper (Piper nigrum L.) is a widely used spice that provides great potential for research in the field of natural products. In this work, the recovery of tannins from black pepper was conducted using microwave-assisted extraction (MAE). The study involves four independent variables: power (from 100 to 300 W), extraction time (from 10 to 40 minutes), temperature (from 35 to 50 °C), and the ratio of food to solvent (from 0.25 to 0.5 g/10 mL). The response variable was the extraction yield, which is the total tannin content. A total of 30 different experimental runs were completed in the MAE system. An evaluation and comparison of two non-verbal modeling approaches and artificial intelligence-based models was conducted. In order to predict design performance and results, the three SVR-RSM, M5Tree, and RM5Tree models were compared to a proposed nonlinear regression model. Evaluations were conducted using health criteria such as RMSE and NSE. With an RMSE of 0.035 and an NSE of 0.91, the SVR-RSM algorithm showed the highest level of accuracy. A RMSE of 0.048 and an NSE of 0.83 is obtained from the RM5tree model, while a RMSE of 0.055 and an NSE of 0.78 is obtained from the M5Tree model. Also, an NSE of 0.65 and a RMSE of 0.068 were obtained for the proposed nonlinear model. The SVR-RSM algorithm had maximum accuracy, but tree models for systems requiring a quick response are the right options. Using the proposed non-error model, complex relationships between variables could also be modeled.