Prediction of Tuberculosis in Sistan Region Using Random Forest Method

Tuberculosis is one of the most important infectious diseases of the current century, which has the ability to involve all organs of the body, but the lungs are more prone to tuberculosis. In the Sistan region, malnutrition and proximity to high-risk countries are two major factors in the high incidence of tuberculosis. Therefore, early prediction and diagnosis of tuberculosis is of particular importance to help the patient and prevent its transmission to others.Therefore, this study has used random forest method to facilitate disease prediction.

The data used in this study were collected from the data of the disease form of suspicious cases of health and medical services centers in Zabol, Nimrooz and Hamoon counties during the years 2015-2018. Then, in order to predict tuberculosis, a random forest method was used in R software.

Preliminary data showed that out of a total of 116 identified forms of disease related to identified tuberculosis patients, 58 had positive smear pulmonary tuberculosis, 36 had negative smear pulmonary tuberculosis and 22 had extra-pulmonary tuberculosis. The results of the analysis showed that the software accuracy for predicting tuberculosis in these three groups was 57%, which increased to 68% with the removal of extra-pulmonary tuberculosis cases.

This study showed that it is difficult to predict cases of extra-pulmonary tuberculosis using data from the disease form. Also, in order to increase the accuracy of predicting cases of positive and negative smear pulmonary tuberculosis by accidental forest method, the risk factors of the disease form should be reviewed.