Hierarchical Probabilistic Learning for Tri-Class Hair Health Analytics and Hormonal Status Identification

Abstract

Hair-related disorders have become increasingly prevalent due to factors such as unhealthy lifestyles, environmental pollution, genetic predisposition, nutritional deficiencies, and hormonal imbalances. Conventional diagnostic approaches primarily depend on clinical examination and expert interpretation, making early diagnosis challenging and limiting the ability to accurately evaluate the combined influence of multiple health-related factors. To overcome these limitations, this study proposes an Explainable Artificial Intelligence (XAI)-driven framework for intelligent hair health prediction using machine learning and deep learning techniques. The proposed system is implemented using the Flask web framework and incorporates comprehensive data preprocessing, exploratory data analysis, and comparative evaluation of several machine learning models, including Support Vector Machine (SVM), Logistic Regression (LR), Decision Tree (DT), KNearest Neighbors (KNN), Gradient Boosting (GB), and AdaBoost (AB). In addition, a novel hybrid model combining a Probabilistic Neural Network (PNN) and a Sparse Representation Classifier (SRC) is developed to improve predictive performance. Experimental evaluation demonstrates that the proposed hybrid model achieves an accuracy of 0.9950, outperforming the baseline classifiers. To enhance model transparency and clinical interpretability, XAI techniques are integrated to identify the most influential prediction factors, estimate individual risk levels, and generate personalized treatment recommendations. Furthermore, the proposed framework supports multiple prediction tasks, including hair loss assessment, treatment recommendation, and hormonal impact analysis, providing an interpretable and reliable decision-support system for intelligent hair healthcare.