Leveraging Deep Learning with Case-Based Reasoning for Interpretable Bid-Rigging Detection

Detecting bid-rigging in public procurement is crucial for identifying collusive behaviors in bidding activities. However, the well-accepted cartel-like detection methods based on machine learning are not efficient to identify complex relationships between features associated with collusive behaviors. Further, the detection methods based on deep learning are not competent at interpretability and transparency, which are critical to support decision-making to clarify why the transaction entities are announced to be bid-rigging. Addressing this gap, we propose a novel detection approach that enhances interpretability by integrating deep learning with case-based reasoning. Our deep learning model learns complex patterns and nonlinear relationships in the data, while case-based reasoning provides interpretability through analogies to previous cases. Evaluation on a Japanese public procurement dataset demonstrates that our method not only achieves superior performance but also ensures interpretability. This study is a new and practical attempt to apply deep learning to detect bid-rigging activities with explainable decisions.