Learning Hyperplane Tree: A Piecewise Linear and Fully Interpretable Decision-making Framework

TL;DR

The paper presents Learning Hyperplane Tree (LHT), a transparent, interpretable, and efficient tree-based classification model that uses hyperplanes and piecewise linear functions to outperform existing models on public datasets.

Contribution

Introduces LHT, a novel hyperplane-based tree model that enhances interpretability and classification performance compared to prior tree methods.

Findings

LHT outperforms state-of-the-art tree models on multiple datasets.

LHT provides high transparency with feature contributions clearly observable.

LHT achieves competitive or superior accuracy while maintaining interpretability.

Abstract

This paper introduces a novel tree-based model, Learning Hyperplane Tree (LHT), which outperforms state-of-the-art (SOTA) tree models for classification tasks on several public datasets. The structure of LHT is simple and efficient: it partitions the data using several hyperplanes to progressively distinguish between target and non-target class samples. Although the separation is not perfect at each stage, LHT effectively improves the distinction through successive partitions. During testing, a sample is classified by evaluating the hyperplanes defined in the branching blocks and traversing down the tree until it reaches the corresponding leaf block. The class of the test sample is then determined using the piecewise linear membership function defined in the leaf blocks, which is derived through least-squares fitting and fuzzy logic. LHT is highly transparent and interpretable--at each branching block, the contribution of each feature to the classification can be clearly observed.