Combining Doubly Robust Methods and Machine Learning for Estimating Average Treatment Effects for Observational Real-world Data