A novel adaptive sigma KNN model for depression and anxiety detection following the COVID 19 pandemic

Mental health disorders, such as depression and anxiety, are increasing, and thus, there is a necessity for accurate and effective detection. K-Nearest Neighbors (KNN) and extensions have been extensively used in disease detection. In this work, Adaptive Sigma KNN (ASKNN) is proposed, an improved version of KNN that adapts neighbor influence dynamically with an adaptive sigma parameter, enhancing predictive accuracy and stability in mental health classification. ASKNN is tested on our dataset and various public mental health and benchmark medical datasets. Performance has been measured through Accuracy, Precision, Recall, and F1-score, with cross-validation ensuring reliability. Friedman and Wilcoxon signed-rank tests were used to evaluate statistical significance, whereas Cohen's d and Cliff's Delta estimated effect size. ASKNN performed with 91.00% accuracy for depression and 84.50% for anxiety, with precision, recall, and F1-score of 0.91, 0.87, and 0.89 for depression, and 0.86, 0.79, and 0.82 for anxiety. AUC-ROC values additionally supported performance, with values of 0.95-0.91 (depression) and 0.93-0.78 (anxiety) between classes. Our statistical testing validates that ASKNN considerably surpasses baseline KNN models for all four measurements. In addition, the research resonates with the United Nations Sustainable Development Goal (SDG) 3, underlining the impact of artificial intelligence-based methodologies in optimizing mental health diagnoses. By increasing depression and anxiety identification accuracy, ASKNN promotes early intervention plans, ultimately benefiting international efforts to lower the global burden of mental illnesses.