I want to create a model for air pollution prediction using deep learning. I have 5 kinds of raster data and each one is related to one of my dependent or independent features. These raster data cover the same area. I have 24 raster files for each feature, each for a month over two years. I have 5 independent variables and 1 dependent variable. This means I have 144 raster files each including numerous cells. if I want to use this data to train a deep learning model, what is the best data structure to work with? Is it better to export each raster data as a numpy array and try to merge the data for each month in a notebook and then work with that data? or create a multiband raster data each band representing one variable amount in that cell for special month which means 24 multi-band raster data. Or do you have any other ideas?
