I'm using this dataset : https://archive.ics.uci.edu/ml/datasets/optical+recognition+of+handwritten+digits a dataset that consists of 65 columns , the last column is the label for 10 classes i.e 0,1,2,3...9. The other 64 columns are the features of each digit. Reshaping the features to (8,8) matrix and plotting it, results in the image of the digit. For example like the one in the picture. I'm trying to use decision tree regression for training and afterwards I want to give only the first 48 features and let the regression model predict the next 16 pixels. I think I have to create 16 decision tree regression models, using the sklearn library, I'm not sure what I have to start with. I appreciate any help. Thanks in advance.