Log-Scale Correlation Classifier for Mushroom Identification in Agricultural Internet of Things Systems

Classifying edible and poisonous mushrooms is crucial to food safety, as misidentification can pose severe toxicological risks. Conventional probabilistic classifiers, such as Naïve Bayes and Logistic Regression, often underperform on categorical datasets with correlated attributes and skewed distributions. This study introduces the Log-Scale Feature Correlation Classifier, a novel probabilistic framework that integrates logarithmic transformation and correlation-weighted probability estimation to address these challenges. Using the UCI Mushroom dataset and a 10-fold cross-validation scheme, LSFCC was benchmarked against standard models. The results demonstrate that LSFCC achieved consistently superior accuracy (0.99), precision, and recall, significantly outperforming both Logistic Regression and Naïve Bayes, as confirmed by statistical tests (p<0.01). Its lightweight design and interpretability make it highly suitable for real-time deployment on resource-constrained IoT devices, particularly within Agricultural IoT systems for autonomous mushroom identification. Future research will explore LSFCC’s adaptability to noisy, multimodal data and hybrid architectures, ensuring broader applicability in real-world bioinformatics and food safety domains.