When less is more: How increasing the complexity of machine learning strategies for geothermal energy assessments may not lead toward better estimates