Anxiety disorder detection using machine learning approach

Abstract

The objective of this research is to tackle urgent problems in the early identification of Anxiety Disorder, specifically in Bangladesh. The country is renowned for its high prevalence of this mental disorder, which is often overlooked. The study aims to predict disorders using machine learning techniques, specifically targeting young adults. It’s a mental health condition characterized by nervousness, panic, fear, sweating, and rapid heartbeat. The study uses machine learning models to predict anxiety disorders in individuals of all ages. The study employs various algorithms like CatBoost, Random Forest, SVM, Bagging, XGBoost, Logistic Regression, Decision Tree, and Naive Bayes to analyze data from surveys with a focus on university students. The data was processed using Python and Pandas library, with preprocessing performed, including handling null values and manual review. The survey results show a strong correlation between symptoms and anxiety disorder, indicating that machine learning can effectively identify potential anxiety cases. The anxiety dataset was analyzed using various ML algorithms, with CatBoost achieving the highest accuracy of 92.81%, followed by RF, SVM, Bagging, LR, XGBoost, DT, and GNB. The approach is revolutionary in Bangladesh, where data-driven healthcare methods are rare. The study's findings are relevant in both academic and practical circles since they provide a more cost-effective and efficient technique of mental diagnosis than previous approaches. Completing the dataset, refining the models to accommodate a wider range of demographics, and incorporating these results into public health initiatives are possible future directions.