A Study of Disease Prediction on Weighted Symptom Data Using Deep Learning and Machine Learning Algorithms

Abstract

Artificial intelligence has gained significant power in the health sector with the increase in electronic data obtained from biomedical and health services. This large data repository allows patient data to be processed and meaningful for deep learning and machine learning developments. This study protects people from information pollution on the internet, informs them about their disease with a reliable accuracy score, and prevents terrible scenarios by providing the earliest diagnosis for essential diseases. It also serves many purposes, such as helping doctors make diagnoses about a patient's condition and improving medical students' knowledge by practicing on different types of cases. Our system analyzes the symptom values the user gives and then returns the disease predicted with the highest accuracy using deep learning and machine learning algorithms. The dataset includes 133 symptoms and 42 disease types. There are 306 patient records containing different types of cases. This study uses supervised machine learning techniques, Support Vector Machine, Naive Bayes Classifier, K-Nearest Neighbors, Random Forest Classifier, Decision Tree Classifier, XGBoost, LightGBM, and Multilayer Perceptron Classifier were tried on a dataset available online. As a result of the experiments, it was seen that the highest accuracy score was achieved by using the XGBoost algorithm.