Machine learning algorithms to predict khat chewing practice and its predictors among men aged 15 to 59 in Ethiopia: Further analysis of Ethiopian Demographic and Health Survey 2011 and 2016

Mequannent Sharew ^* Lamrot Yohannes Eliyas Addisu Nebebe Baykemagn Demise

• Institute of Public Health, University of Gondar, Gondar, Ethiopia

Introduction: Khat chewing is a significant public health issue in Ethiopia, and various demographic factors influence its utilization. Understanding the magnitude and factors associated with khat chewing practice can inform targeted interventions. Hence, this study aimed to Predict khat chewing practice and its determinant factors among men aged 15 to 59 years in Ethiopia: Using machine learning algorithm Methods: This study used data from Demographics and Health Surveys conducted in Ethiopia from 2011 to 2016. A weighted sample of 26,798 men aged 15 to 59 years was included in the study. The data were cleaned and weighted by STATA version 17 software. Descriptive statistics was executed using STATA version 17 software. Python 3.12 software was used for machine learning prediction of tobacco use among men. Furthermore, Decision tree, Logistic Regression, Random Forest, KNN, Support Vector Machine, eXtreme gradient boosting, (XGBoost), and AdaBoost classifiers was employed to identify the most critical predictors of khat chewing practice among men. In addition, accuracy, and Area under the curve was used to evaluate the performance of the Predictive models. Result: From the total of 26798 (weighted) sample 8,786(32.79) men aged from 15 to 59 years chewing khat. XGBoost achieved an accuracy of 87 % and an area under the ROC curve of 0.94. The Beeswarm plot of the SHAP analysis based on the XGBoost classifier model showed that age, religion, region, wealth index, age at first sex, frequency of watching television, frequency of listening radio and number of sexual partners as the top ranked variables for predicting khat chewing practice among men. Conclusion: Three out of ten men participate for khat chewing practice in Ethiopia. The S=XGBoost model provides better predictive power to predict determinants of khat chewing practice in Ethiopia. The XGBoost model identified age, religion, region, wealth index, age at first sex, media exposure, and number of sexual partners as the strongest predictors of khat chewing among Ethiopian men. Effective khat prevention requires preserving rural norms discouraging its use and extending them to urban areas.