Tree-like Pairwise Interaction Networks

TL;DR

The paper introduces the Tree-like Pairwise Interaction Network (PIN), a neural network architecture that models pairwise feature interactions explicitly, offering interpretability and improved predictive accuracy in tabular data applications.

Contribution

It presents a novel neural network design that mimics decision tree structures to capture pairwise interactions, enhancing interpretability and efficiency in SHAP computations.

Findings

PIN outperforms traditional and neural network benchmarks in accuracy

PIN provides interpretable insights into feature interactions

Efficiently computes SHAP values for pairwise interactions

Abstract

Modeling feature interactions in tabular data remains a key challenge in predictive modeling, for example, as used for insurance pricing. This paper proposes the Tree-like Pairwise Interaction Network (PIN), a novel neural network architecture that explicitly captures pairwise feature interactions through a shared feed-forward neural network architecture that mimics the structure of decision trees. PIN enables intrinsic interpretability by design, allowing for direct inspection of interaction effects. Moreover, it allows for efficient SHapley's Additive exPlanation (SHAP) computations because it only involves pairwise interactions. We highlight connections between PIN and established models such as GA2Ms, gradient boosting machines, and graph neural networks. Empirical results on the popular French motor insurance dataset show that PIN outperforms both traditional and modern neural networks benchmarks in predictive accuracy, while also providing insight into how features interact with each another and how they contribute to the predictions.