Hybrid Intelligence For Precision Milk Quality Assessment Integrating Naïve Bayes And K-Means Clustering

Abstract

Milk quality is a fundamental factor in the food industry, directly impacting consumer health and product market value. Conventional manual assessment is often inefficient for large-scale operations, requiring rapid, accurate automated solutions. This study proposes a hybrid machine learning approach integrating the Naïve Bayes algorithm for quality classification (Low, Medium, High) and K-Means Clustering for diagnostic analysis based on physicochemical similarities. Using a dataset of 1,059 samples with seven sensory attributes (pH, temperature, taste, odor, fat, turbidity, and color), the data were cleaned, normalized, and modeled. Evaluation results demonstrate that the Naïve Bayes method achieved an accuracy of 85.38%. Notably, the model achieved 100% precision in identifying low-quality milk, making it a highly reliable food safety filter. Concurrently, K-Means with k=5 (selected as the optimal value after testing k=2 - 7) successfully segmented the data into five distinct diagnostic clusters. Centroid analysis revealed that temperature and odor are the primary factors for distinguishing the root causes of quality degradation. This study concludes that combining classification and clustering methods significantly enhances quality control efficiency by providing both quality labels and diagnostic insights into the factors driving milk spoilage.