Improving the Accuracy-Memory Trade-Off of Random Forests Via Leaf-Refinement

TL;DR

This paper introduces a leaf-refinement algorithm for Random Forests that improves the accuracy-memory trade-off by fine-tuning leaf predictions, outperforming existing pruning methods on multiple datasets.

Contribution

The paper presents a novel leaf-refinement technique using stochastic gradient descent to enhance Random Forests' accuracy-memory trade-off beyond traditional ensemble pruning.

Findings

Our method outperforms 7 state-of-the-art pruning techniques on 11 of 16 datasets.

Leaf-refinement achieves better accuracy-memory trade-off than pruning alone.

The approach is effective in real-world applications, demonstrated by a case study.

Abstract

Random Forests (RF) are among the state-of-the-art in many machine learning applications. With the ongoing integration of ML models into everyday life, the deployment and continuous application of models becomes more and more an important issue. Hence, small models which offer good predictive performance but use small amounts of memory are required. Ensemble pruning is a standard technique to remove unnecessary classifiers from an ensemble to reduce the overall resource consumption and sometimes even improve the performance of the original ensemble. In this paper, we revisit ensemble pruning in the context of `modernly' trained Random Forests where trees are very large. We show that the improvement effects of pruning diminishes for ensembles of large trees but that pruning has an overall better accuracy-memory trade-off than RF. However, pruning does not offer fine-grained control over this trade-off because it removes entire trees from the ensemble. To further improve the accuracy-memory trade-off we present a simple, yet surprisingly effective algorithm that refines the predictions in the leaf nodes in the forest via stochastic gradient descent. We evaluate our method against 7 state-of-the-art pruning methods and show that our method outperforms the other methods on 11 of 16 datasets with a statistically significant better accuracy-memory trade-off compared to most methods. We conclude our experimental evaluation with a case study showing that our method can be applied in a real-world setting.