Symptom-Based Disease Recognition and Test Recommendation Using XGBoost

Timely diagnosis of a disease is essential to improve the condition of the patient; however, due to the limited access to medical professionals, it often delays the early diagnosis of a health condition. This paper presents a machine learning based Disease Recognition and Test Recommendation System to help the user in predicting any possible disease based on the symptoms reported and perform any appropriate diagnostic test to confirm the diagnosis. The proposed system uses supervised learning algorithms such as Random Forest and XGBoost that are trained using public symptom and disease data. User given symptoms are converted into binary feature vectors and fed to classification model to detect probable diseases with high level of accuracy. A rule-based mapping module is further integrated to recommend the relevant medical tests corresponding to the predicted conditions and fill the gap between the preliminary self- assessment and professional healthcare consultation. The system is tested with standard performance parameters such as accuracy and precision and shows reliable and interpretable results. This work adds a scalable and easy to use decision support system that helps with the early detection of disease and encourages preventive healthcare.