Multi Disease Classification System Based on Symptoms using The Blended Approach

Abstract

In today’s world, everyone is preoccupied with work and other activities, leaving little time to visit doctors about illnesses that may appear to be minor at first but develop into life-threatening conditions as time passes. As a result, the proposed model accesses a public repository that maintains numerous symptoms and their possible diseases as a matrix for early disease prediction and prevention. Symptoms are received from the user and fed into the embedded blending algorithm to estimate the type of disease. The patient’s records are collected from several hospitals, resulting in a massive volume of data, which leads to an inefficient prediction model using machine learning approaches. Since the proposed model is a combined approach of a training mechanism, it can reduce the number of accessing records in every step. Traditional approaches like bagging and boosting construct a larger number of decision trees because of the vast amount of data. This results in the utilization of more resources and sometimes CPU enters into a saturation state. The proposed system solves this problem by using optimized parameters for tree construction and reducing memory and resource utilizations.