Predict Diagnosis of ME/CFS and Depression Using K-Nearest Neighbor Classification Method

Abstract

Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS) and depression are two medical conditions that often present with overlapping symptoms, making accurate diagnosis difficult, especially in early stages. This study aims to develop a predictive application using the K-Nearest Neighbor (K-NN) classification algorithm to differentiate between ME/CFS and depression based on clinical and lifestyle-related features. A dataset consisting of 1,000 records and 16 attributes was obtained from a public repository. Preprocessing steps included label encoding for categorical features, mean and mode imputation for missing values, and Min-Max normalization for numerical features. The model was trained using 80% of the data and evaluated on the remaining 20% using Manhattan distance and a k-value of 10. The application was developed with an interactive user interface, enabling predictions based on user input. The model achieved an overall accuracy of 88.5%, with excellent performance in detecting depression and ME/CFS, but moderate performance in identifying comorbid cases. The findings suggest that K-NN can be effectively utilized to support differential diagnosis in mental health, particularly for conditions with overlapping clinical symptoms. Future enhancements may include the incorporation of additional features and algorithmic improvements to address limitations in comorbid case detection.