Nowadays the way people seek health-related information that is undergoing significant changes worldwide. One major challenge is the difficulty that individuals face in finding trustworthy online resources or information about illnesses, diagnoses, and treatment options. A solution lies in creating a recommendation system for doctors, patients and medications that utilizes review mining to simplify this process and save time. Four machine learning algorithms are used together to achieve around 96% accuracy. It provides departments and doctors information based on prediction to users. It focuses on securely storing patient history for future references integrating advanced password hashing techniques as well as JWT (Jason Web Token) based authentication. It aims to improve healthcare accessibility to reduce patient confusion for faster as well as informed decision making. However, understanding complex medical language can be overwhelming for non-specialist users. The vast amount of medical data is there across different platforms which further complicates their search. To address these issues, such a system must be designed to cater specifically to the needs of the healthcare sector, ensuring it remains user-friendly and easy to understand for all individuals.
Introduction
With the rise of internet use for health information, many users struggle to understand complex medical data due to scattered sources and difficult terminology. There is currently no comprehensive, user-friendly system that offers personalized doctor and medication recommendations based on symptom inputs. This project aims to develop an intelligent, secure web application that predicts potential diseases from user-reported symptoms and guides users toward appropriate medical specialists.
The system uses a reliable dataset of diseases and symptoms to train multiple machine learning models—Decision Tree, Random Forest, K-Nearest Neighbors (KNN), and Gaussian Naive Bayes—to improve prediction accuracy. It also includes a user management system for storing past predictions and supports easy symptom input with controlled data access.
Related work highlights the strengths of these models in healthcare, with ensemble methods improving accuracy and some deep learning techniques applied in medical imaging. Existing symptom checkers like WebMD lack personalization and doctor recommendations, gaps this system aims to fill.
The methodology involves preprocessing a dataset with 5945 records and 124 relevant symptoms covering 61 diseases. The models achieved around 95% accuracy and high precision, recall, and F1 scores. Using multiple models helps reduce mispredictions by offering users suggestions for multiple possible departments or doctors, thereby speeding up medical intervention.
Conclusion
The implementation of the Symptoms Based Disease Prediction Web Application showcases the potential impacts that machine learning can have in the real world, especially in the healthcare industry. One milestone of the work was constructing a multi-model architecture aimed at predicting diseases with high accuracy based on symptoms provided by users. The performance metrics of the four models published confirm the accuracy and efficiencymachine learning offers for predictive endeavours.In any case, although the technical performance was provided a lot of hope, the greatest takeaway from this project, at least for me, is how it can be applied in the real world. Patients have a hard time of determining which symptoms warrant attention and which doctor or unit they want to see. Our system fills this gap by providing a diagnostic system which will provide department suggestion that stands to significantly reduce the time taken to receive critical care. Some of the salient points learned from the entire endeavours are as follows:
References
[1] Feng J., Yadong J., Hao Z., Yan D., Hui L., Shuang M., Yadong W., “Artificial Intelligence in Healthcare: Past, Present and Future”, Stroke and Vascular Neurology, 2020, 5 (3), 230–243.
[2] Cai W., “Equipment and Machine Learning in Welding Monitoring: A Short Review”, 5th International Conference on Mechatronics and Robotics Engineering (CMRE), Rome, Italy, 2019.
[3] Arjun K., et al., “Web-Based Symptom Checker Using ML”, International Journal of Recent Technology and Engineering (IJRTE), 2019, 8, (no issue number specified).
[4] Xia L., et al., “Deep Learning Systems for Disease Diagnosis”, Nature Medicine, 2019.
[5] Rajendra C.D., “Machine Learning in Medicine”, Circulation, 2015, 132 (20), 1920–1930.
[6] Andre E., Brett K., Roberto A.N., Justin K., Susan M.S., Helen M.B., Sebastian T., “Dermatologist-Level Classification of Skin Cancer with Deep Neural Networks”, Nature, 2017, 542 (7639), 115–118.