Revolutionizing food quality assessment: Unleashing the potential of artificial intelligence for enhancing honey physicochemical, biochemical, and melissopalynological insights

Abstract

In the pursuit of advancing food quality assessment, this study employs sophisticated data-driven techniques to delve into the complex realm of honey analysis. With the aim of unraveling the multifaceted nature of honey quality, Self-Organizing Maps (SOMs) and Principal Component Analysis (PCA) were employed to scrutinize the interplay of physicochemical, biochemical, and melissopalynological attributes in honey samples collected from the diverse drylands of Algeria. The dataset comprised 62 honey samples and eight crucial parameters. These parameters span climate zones (arid vs. desertic), honeybee breeds (Tellian, Saharan, and hybrid), honey extraction methods (manual pressing vs. electric centrifugation), and beekeeping systems (modern vs. traditional). Using SOMs, honey samples were categorized into distinct clusters that reflect variations across these four honey-related variables. Additionally, SOM heatmaps offer granular insights into individual parameters' distribution across the SOM map. Our analysis revealed nuanced distinctions in honey quality across North African regions, with specific parameters playing a pivotal role in defining honey quality. On average, the honey samples exhibited the following characteristics: a water content of 15.14 %, an electrical conductivity of 0.5 µS/cm, a pH of 4.20, a total sugar content of 83 %, a reducing sugar content of 63.83 %, a proline concentration of 382.7 mg/kg of honey, an hydroxymethylfurfural level of 77.4 mg/kg, and an average pollen grain density of 4.56 × 105 grains per 10 g of honey. Notably, the study identified clear demarcations in honey quality linked to beekeeping systems and revealed characteristics associated with bee breeds and extraction techniques. The results underscored the significance of selected honey parameters as key indicators of quality. This analytical approach not only offered a comprehensive assessment of honey quality but also holds potential for broader applications within the food industry. The findings invite further exploration into the ecological and genetic dimensions of beekeeping practices in North Africa to deepen our understanding of honey's multifaceted attributes. This study showcased the efficacy of SOMs and PCA in unraveling the complex fabric of honey quality assessment. These data-driven techniques, complemented by the structured dataset and analytical approach used, provided valuable insights that contributed to enhancing the scientific understanding of honey quality. By elucidating the complex relationships between physicochemical, biochemical, and melissopalynological parameters and honey quality, this research paves the way for future studies in this field and holds promise for broader applications in food quality assessment and monitoring.