Higher-order Neural Additive Models: An Interpretable Machine Learning Model with Feature Interactions
This addresses the need for interpretable models in high-stakes applications by enabling analysis of high-order interactions, though it is incremental as it builds directly on existing NAMs.
The paper tackles the limitation of Neural Additive Models (NAMs) in capturing only first-order feature interactions by proposing Higher-order Neural Additive Models (HONAMs), which improve predictive accuracy while maintaining interpretability for real-world datasets.
Neural Additive Models (NAMs) have recently demonstrated promising predictive performance while maintaining interpretability. However, their capacity is limited to capturing only first-order feature interactions, which restricts their effectiveness on real-world datasets. To address this limitation, we propose Higher-order Neural Additive Models (HONAMs), an interpretable machine learning model that effectively and efficiently captures feature interactions of arbitrary orders. HONAMs improve predictive accuracy without compromising interpretability, an essential requirement in high-stakes applications. This advantage of HONAM can help analyze and extract high-order interactions present in datasets. The source code for HONAM is publicly available at https://github.com/gim4855744/HONAM/.