PEDL-XAI: A Hybrid Probabilistic Ensemble Deep Learning Approach for Hair Disorder Diagnosis

Abstract

The growing occurrence of hair-related disorders, influenced by modern lifestyles, environmental conditions, and various health issues, has increased the need for intelligent and data-driven healthcare solutions. Conventional methods of hair loss analysis, which rely largely on clinical observation and generalized medical knowledge, often lack precision and fail to account for the combined effects of multiple contributing factors such as genetics, stress, nutrition, and medical history. These limitations underscore the need for advanced systems capable of analysing complex, multi-dimensional data to deliver accurate and reliable predictions. To address this challenge, this study presents an Explainable Artificial Intelligence (XAI)-based hair health prediction system developed using the Flask web framework. The proposed system incorporates data preprocessing, exploratory data analysis, and multiple machine learning algorithms, including Support Vector Machine (SVM), Logistic Regression (LR), Decision Tree (DT), K-Nearest Neighbors (KNN), Gradient Boosting (GB), and AdaBoost (AB). Furthermore, a novel hybrid Probabilistic Ensemble Deep Learning (PEDL) model is introduced, which integrates a Probabilistic Neural Network (PNN) with a Sparse Representation Classifier (SRC) to enhance predictive accuracy. Experimental results demonstrate that the PEDL model achieves a superior accuracy of 0.9950, outperforming conventional machine learning approaches. A significant contribution of this work lies in the integration of XAI techniques, which provide transparent and interpretable predictions by identifying key contributing factors, assessing risk levels, and generating personalized recommendations. The system supports multi-target prediction, including hair loss evaluation, treatment recommendations, and hormonal impact analysis.