Finding Optimal STI Strategies for HIV Using Fitted Q Algorithm with XGBoost Regression

We are trying to use the fitted Q algorithm with XGBoost regression to compute the optimal structured treatment interruption strategies for HIV infected patients. By using the combination of these two methods, we showed that this approach is computationally more efficient, at the same time the optimal results were obtained using fewer training data. And the speed of convergence of the optimal strategy is much fast than the Extra-tress regression method used in the fitted Q algorithm.