Evaluating solar drying effects and machine learning models for nutritional quality of jerusalem artichoke

Abstract

This study compares the effects of different drying methods (Natural, Open-sun, Single, Double, and Triple effect solar) on the nutritional components of Jerusalem artichokes. It evaluates the prediction of biochemical properties using machine learning algorithms. The total protein, mineral content, and vitamins were analyzed. The Triple effect solar method best preserved nutrients, maintaining the highest protein content (69379 mg/kg) and beta-carotene (0.68 mg/kg), while the Natural method caused the most significant losses. Ascorbic acid (AA) was also better retained under the Triple effect solar method (94.82 mg/kg vs. 57.41 mg/kg in Natural drying). Machine learning algorithms accurately predicted biochemical properties (R2 > 90.00 %), especially random forest and k-nearest neighbor. Strong correlations were observed between total protein and AA, niacin, and phosphorus. These results show that the Triple effect solar method is optimal for nutrient preservation, and machine learning offers a promising tool for quality prediction and product development in the food industry.