Tuning Hyperparameters of Decision Tree Classifiers Using Computationally Efficient Schemes

摘要

Attack types and patterns are constantly evolving which makes frequent detection system updates an urgent need. In contrast, the computation cost of developing machine learning-based detection models such as decision tree classifiers is expensive which can be an obstacle to frequently updating detection models. Tuning classifiers' hyperparameters is a key factor in selecting the best detection model but it significantly increases the computation overhead of the developing procedure. In this research, we have presented a computationally efficient strategy and an algorithm for tuning decision tree classification algorithms' hyperparameters with less budget and time.