Enhancing Depression Detection Accuracy in Northwest Nigerian Adults Using Ensemble Learning Technique
Shuaibu Samaila Mohammed Ali Kawo
Abstract — Millions of individuals all over the world suffer from depression, a serious and common mental illness. In Northwest Nigeria, depression is still not well recognized because of stigma, the use of multiple languages, including English, Hausa, and Fulfulde, and a lack of mental health resources for thorough detection. In order to increase accuracy and equality across language groups, this study suggests an ensemble-based detection architecture that incorporates complementary classifiers trained on clinical screening datasets. Researchers have used artificial intelligence (AI) to automatically detect depression symptoms. Majority of healthcare researchers and management use series of machine learning techniques to improve disease detection, diagnosis, and prediction in order to aid in the decision-making. This study investigates how well machine learning ensemble techniques can improve the precision of depression detection in northwest Nigeria. Four individual base machine learning algorithms such as Randon Forest, Decision Tree, Support Vector Machine and K-Nearest Neighbors will be used as a pipeline combination of the classifiers to form Stack Ensemble technique that will be trained on clinical screened dataset from Primary Health Care Centers to enhance depression detection. The entire model will be evaluated using the appropriate machine learning metrics such as accuracy, precision, F1-score, recall and ROC-AUC respectively. Keywords — Stack Ensemble, Machine learning, Depression, Support Vector Machine, Naïve Bayes.