Machine Learning Models for Depression Prediction: A Comprehensive Analysis

Millions of people across the globe suffer from depression, a common mental illness that needs to be diagnosed and treated early to avoid this from turning into something much serious. This paper aims at designing a system that predicts the likelihood of depression based on the structured questionnaire responses with the use of machine learning-based approach. The dataset is balanced with respect to the target classes "Depressed" and "Not Depressed" and comprises answers to 14 questions. Accuracy, precision, recall, F1-score, and confusion matrices for five machine learning models were compared: Logistic Regression, Random Forest, XGBoost, Support Vector Machine (SVM), and an Artificial Neural Network (ANN). SVM performed the best in the contest of prediction accuracy and consistency by having an accuracy of 99% on testing.The proposed approach will prove to be a valuable asset for mental health professionals in allowing rapid and accurate depression screening. Future research will attempt to include larger datasets and explore more features in improving the robustness of prediction.