Towards a Machine Learning-Based Approach to Predict Space Object Density Distributions