Shapley Values of Reconstruction Errors of PCA for Explaining Anomaly Detection
This work addresses the need for interpretability in anomaly detection for users in fields like security or finance, but it is incremental as it builds on existing Shapley value and PCA techniques.
The authors tackled the problem of explaining PCA-based anomaly detection results by developing a method to compute Shapley values for reconstruction errors, using a probabilistic PCA approach to handle correlated features, and demonstrated through numerical examples that this method provides better explanations than raw reconstruction errors.
We present a method to compute the Shapley values of reconstruction errors of principal component analysis (PCA), which is particularly useful in explaining the results of anomaly detection based on PCA. Because features are usually correlated when PCA-based anomaly detection is applied, care must be taken in computing a value function for the Shapley values. We utilize the probabilistic view of PCA, particularly its conditional distribution, to exactly compute a value function for the Shapely values. We also present numerical examples, which imply that the Shapley values are advantageous for explaining detected anomalies than raw reconstruction errors of each feature.