Enhanced Depression Diagnosis Model Using Bayesian Network Technique

Source abstract

This paper aimed to develop an enhanced depression diagnosis model using Bayesian network techniques, focusing on collecting a diverse dataset, constructing a probabilistic model, integrating multi-modal data, and rigorously validating the system. The dataset was sourced from Kaggle, supplemented by clinical records, self-report questionnaires (PHQ-9 and BDI), semi-structured interviews, wearable device data (heart rate and sleep patterns), and public health databases to ensure robustness. The data underwent preprocessing steps, including cleaning, normalization, feature selection, and integration, to ensure quality and relevance. The Bayesian network model was trained and validated to capture probabilistic relationships between variables, demonstrating strong performance with 80% accuracy and a 79%ROCAUC Score. The results highlight the model's reliability and potential for clinical use, though further refinements, such as reducing false negatives and testing on external datasets, are recommended to enhance its real-world applicability.