| Abstrakt: |
Tomatoes are a perishable agricultural product due to their high nutritive values and moisture content. However, they are prone to wastage and spoilage during their seasonal period. The quest for healthy dried tomato products by consumers necessitates the need to evaluate and model the quality traits of microwave-dried tomatoes. The soft-computational modelling of quality traits of microwave-dried tomatoes using the Adaptive Neuro-Fuzzy Inference System (ANFIS) and Artificial Neural Network (ANN) was adopted for this study. The Taguchi experimental design was used to interact with the pretreatments (water blanching (WB), ascorbic acid (AA), and sodium metabisulphite (SM)), tomato slice thicknesses (4, 6, and 8 mm), and microwave powers (90, 180, and 360 W). The interaction effects were used to evaluate the quality traits (percentage shrinkage, rehydration ratio, lycopene, β-carotene, and ascorbic acid) of dried tomatoes using standard procedures. The data obtained from the effects of pretreatments, slice thickness, and microwave power on the quality traits of microwave-dried tomatoes were used to train, test, and validate the ANN and ANFIS models. The coefficient of determination (R2), root mean square error (RMSE), and mean absolute error (MAE) were used to determine the model's performance and efficacy. The quality traits vary, with lycopene ranging between 8.86 and 12.48 mg/100 g, β-carotene (4.23 and 8.77 mg/100 g), and ascorbic acid (20.77–26.85 mg/100 g), respectively. Both modelling techniques are applicable, but the ANFIS model, with a greater R2(≥ 0.9998) and lower RMSE (≤ 0.01335) and MAE (≤ 0.01043), presents more reliable and higher accuracy results than ANN. |
Full text from SpringerLink