Early Depression Detection Using Machine Learning: A Comparative Study Of Ensemble Models And Feature Selection With RFE

Abstract

Kaggle containing multiple behavioral, psychological and demographic variables. The research concentrates on feature selection technique, RFE and evaluates its performance by using various machines’ learning models like XGBoost, Gradient Boosting, AdaBoost and Logistic Regression. Much preprocessing was performed on the dataset, such as managing missing values and scaling of features. A set top 10 key features were chosen through RFE for lessening the dataset complexity and for maximizing the predicting the accuracy. Although ensemble models such as XGBoost and Gradient boosting model attained the highest accuracy for all the experiments (95%), testifying that RFE was able to reduce the dimensionality of the dataset without the loss of discriminating power. The findings suggest AI-enabled predictions of mental health may be a scalable and unbiased approach to screening for depression. The above avenues can form the future direction using deep learning architectures, multimodal data sources and privacy preservative strategies for real world application. This paradigm serves in the expansion of the burgeoning area of artificial intelligence in mental health and provides evidence on the effectiveness of machine learning in predicting early identification of depression and algorithm driven interventions.