Analysis of Patient Tuberculosis Tenet Death Reason and Prediction in Bangladesh Using Machine Learning

Abstract

AbstractTuberculosis disease is one of the world's top infectious diseases that leads to a huge number of patient death worldwide. Tuberculosis gradually attacks the lungs of the patients, where Mycobacterium tuberculosis is one of the main reasons for tuberculosis attacks which are caused by the bacteria. As innovation of new techniques helps generation minimize the tasks or observation period, Machine learning is one of the popular techniques by which a person or an organization can easily build a model to evaluate data, generate ideas, or prediction of values. Machine learning algorithms are used enormously in different sectors, thus using Machine learning models in the health sector increasing rapidly. Health professionals can easily predict or observe a patient's disease using the previous history of the same patient or different similar patient's history. In the paper, tuberculosis patient's death rationale is harmonized from the World Health Organization dataset of tuberculosis disease’s class called causes and deaths, where the country Bangladesh's dataset has been used. Feature of the dataset is one of the main concerns of the patient's death, which is identified using the Machine learning regression and classification algorithm. Linear Regression, Logistic Regression, Decision tree, Random forest, KNN, XGB, Adaboost and algorithms are used in the process to create a model which can identify the best features and it is figured out that Random forest provides the best results. The prediction model for finding the number of death of patients build using the machine learning regression algorithms, where linear regression prediction accuracy is 0.99943, however, the linear model's features selection for the process are not the best noticeable. The random forest algorithm's prediction accuracy was found 0.97820, which is nearest to the linear regression accuracy. In one sentence, it is figured out that Random forest is the best-observed algorithm in both prediction accuracy and feature importance detection.KeywordsTuberculosisMachine learningPrediction of death per yearFeature importance