Decision tree model training process

A method comprising:

• encoding applications via one-hot encoding;
• assigning the applications to a plurality of nodes in a decision tree model by executing the decision tree model on the encoded applications and corresponding criteria assigned to the plurality of nodes;
• determining purities of the plurality of nodes based on approved applications assigned to the plurality of nodes;
• identifying a single path of sequentially interconnected nodes extending from a root node toward a leaf node among the plurality of nodes in which every node on the single path comprises a purity above a purity threshold;
• pruning by excluding remaining nodes from the decision tree model to generate a reduced executable decision tree model;
• mapping attributes of a new application to the single path;
• and
• loading and executing, via a processing device, the reduced executable decision tree model without executing the pruned remaining nodes to generate a decision on the new application.

A non-transitory computer-readable medium comprising instructions which when executed by a processor cause the processor to perform:

• encoding applications via one-hot encoding;
• assigning the applications to a plurality of nodes in a decision tree model by executing the decision tree model on the encoded applications and corresponding criteria assigned to the plurality of nodes;
• determining purities of the plurality of nodes based on approved applications assigned to the plurality of nodes;
• identifying a single path of sequentially interconnected nodes extending from a root node toward a leaf node among the plurality of nodes in which every node on the single path comprises a purity above a purity threshold;
• pruning by excluding remaining nodes from the decision tree model to generate a reduced executable decision tree model;
• mapping attributes of a new application to the single path;
• and
• loading and executing the reduced executable decision tree model without executing the pruned remaining nodes to generate a decision on the new application.