Unleashing the power of AI in predicting the technological and phenolic maturity of pomegranates cultivated in Lebanon

The harvesting time of pomegranates is crucial for maximizing their health benefits and market value. However, traditional methods often fail to consider the intricate interactions between environmental conditions and fruit maturity. This study is the first of its kind in Lebanon to address this limitation by applying advanced machine learning techniques to predict key food quality indicators, which can aid in forecasting or determining the optimal harvesting date. The focus is on technological and phenolic maturity. Over three months, 548 pomegranates were meticulously harvested from three distinct geographic regions in Lebanon: Hasbaya, El Jahliye, and Rachiine. By integrating environmental, physical, and geographical data, we developed predictive models, including Linear Regression (LR) and Multi-Layer Perceptron (MLP) Regressor, to estimate key food quality indicators such as Total Soluble Solids (TSS), Titratable Acidity (TA), Maturity Index (MI), phenolic content, and Color Intensity (CI). Our results demonstrated that the MLP regressor achieved high predictive accuracy, with an R-squared value of 0.84 for TA, making it a reliable tool for predicting acidity levels. The model also showed strong performance in predicting phenolic content and color intensity, with R-squared values of 0.70 and 0.65 respectively, and an average classification accuracy of 71% for categorizing polyphenol levels. Principal Component Analysis (PCA) revealed significant geographic variation in phenolic content. In El Jahliye, phenolic levels ranged from low (<185 mg Gallic Acid Equivalent (GAE) per yield of juice) to moderate (185-400 mg GAE/yield of juice). In Rachiine, levels ranged from moderate to high (>400 mg GAE/yield of juice), while Hasbaya displayed all three phenolic content levels. These findings underscore the importance of region-specific harvesting strategies. As the first study in Lebanon to utilize machine learning for predicting food quality indicators in pomegranates, it provides a novel, data-driven approach to linking these indicators with optimal harvest timing. By accurately forecasting maturity-related metrics using simple physical, geographical, and environmental features, this study offers significant implications for refining agricultural practices in Lebanon and other similar agro-ecological regions, enhancing product quality and market value.