E-commerce Anomaly Detection: A Bayesian Semi-Supervised Tensor Decomposition Approach using Natural Gradients

TL;DR

This paper introduces a Bayesian semi-supervised tensor decomposition method with natural gradients for anomaly detection in e-commerce seller-reviewer data, outperforming existing unsupervised techniques.

Contribution

It presents a novel semi-supervised Bayesian tensor decomposition framework using Polya-Gamma augmentation and natural gradients, improving anomaly detection accuracy.

Findings

Semi-supervised approach outperforms unsupervised baselines.

Natural gradient learning surpasses stochastic gradient and Online-EM.

Polya-Gamma augmentation simplifies Fisher information matrix calculation.

Abstract

Anomaly Detection has several important applications. In this paper, our focus is on detecting anomalies in seller-reviewer data using tensor decomposition. While tensor-decomposition is mostly unsupervised, we formulate Bayesian semi-supervised tensor decomposition to take advantage of sparse labeled data. In addition, we use Polya-Gamma data augmentation for the semi-supervised Bayesian tensor decomposition. Finally, we show that the P\'olya-Gamma formulation simplifies calculation of the Fisher information matrix for partial natural gradient learning. Our experimental results show that our semi-supervised approach outperforms state of the art unsupervised baselines. And that the partial natural gradient learning outperforms stochastic gradient learning and Online-EM with sufficient statistics.