Computational Intelligence and Healthcare Informatics

4A Smart Web Application for Symptom-Based Disease Detection and Prediction Using State-of-the-Art ML and ANN Models

Parvej Reja Saleh* and Eeshankur Saikia

Department of Applied Sciences, Gauhati University, Guwahati, India

Abstract

Detection of diseases using symptoms might seem to be very normal in everyday life, but things get serious when symptoms increase in complexity and/or variety. With this increase in complexity or variety, we as human beings struggle to reliably diagnose any particular illness that may occur as a result of the observed symptoms. Various symptoms typically indicate different disease possibilities, that too with different levels of severity. An automatic generated database including diseases and its symptoms, based on textual discharge summaries of patients is used for the present work. The data includes 149 most common diseases along with symptoms on the basis of strength of association. Using this data, we have aimed at developing a Machine Learning model, based on Classification techniques to detect symptom-based diseases. Industry-standard classification methods, such as Multinomial Naïve Bayes, Logistic Regression, Decision Tree, K-nearest Neighbor, and Support Vector, and Random Forest Classifier are extensively used in this work. For accurate prediction, we have also implemented a Feedforward Neural Network using MLP for training purposes. In this work, we have proposed, for the first time, a smart and simple web-based application, integrated ...