Unlocking the Potential of Machine Learning and Deep Learning for Screening of Geriatric Depression

Geriatric depression is one of the important public health issues that need proper attention. It often remains undiagnosed and untreated. Early detection of high-risk elderlies for depression is the key to its successful management. Conventional time-consuming questionnaire-based geriatric depression assessment tools can be replaced by state-of-the-art machine learning-based automated screening tools. Previous epidemiological research established that different socio-demographic variables and comorbid conditions are important predictors of depression. Due to computational advancement, experiments with different machine learning algorithms are going on in the domain of automated mental health diagnostics to develop a robust model with better performance. In this article, primary data were collected by expert healthcare professionals from 105 elderlies living in Kolkata. Comparative evaluation of the Logistic Regression, Linear Discriminant Analysis, Gaussian Naïve Bayes, Decision Tree Classifier (CART), K Nearest Neighbour Classifiers, Support Vector Machine, Extreme Gradient Boosting, and Multi-Layer Perceptron algorithms is made with precision, recall, and accuracy matrices. As there is a complex interaction between the different predictor variables with depression, it was proposed to explore and evaluate the deep learning model for this purpose. It is identified that the proposed deep learning model outperforms all other conventional machine learning models for screening geriatric depression.