COMPARATIVE ANALYSIS OF PECTIN-CONTAINING PASTE PROPERTIES USING ARTIFICIAL INTELLIGENCE

The article investigates the application of artificial intelligence methods for comparative analysis of the properties of pectin-containing pastes of different composition and origin. The study's relevance is due to the need for an objective assessment of the quality characteristics of pectin-containing products, the growing demand for functional food products, and the need to develop effective methods for their classification in a competitive market. The work aimed to create a comprehensive system for comparative analysis of pectin-containing pastes based on machine learning algorithms, considering rheological, physicochemical, and organoleptic indicators. Modern methods of cluster analysis, discriminant analysis, artificial neural networks of different architectures, and multivariate statistics methods were used to achieve this goal. Twenty-five representative samples of varying composition of pectin-containing pastes were analyzed, including traditional recipes and innovative formulations with functional additives. The study covered products based on pumpkin, apple, citrus, and mixed pectin with different degrees of esterification. Fifteen leading quality indicators characterized each sample. A multi-criteria classification system for pectin-containing pastes was developed based on the optimized k-means algorithm and hierarchical cluster analysis. The created model provides a high classification accuracy of 94.7% and distinguishes six main types of pectin-containing pastes according to their functional characteristics and technological properties. A new integral quality indicator for pectin-containing pastes was proposed, considering the weight coefficients of individual traits determined by the principal component analysis method. Using this indicator allows for the objective ranking of products by quality with an accuracy of 91.3%. Experimental testing on a control sample showed that using the developed intelligent system reduces the time of comparative analysis by 78% and increases the objectivity of product quality assessment by 45% compared to traditional methods of expert evaluation. The practical value of the work lies in creating a universal software tool for food industry enterprises, which provides quick and accurate assessment of the quality of pectin-containing products, optimization of technological processes, and their effective positioning in the market of functional food products.