Unsupervised Event Outlier Detection in Continuous Time

TL;DR

This work develops the first unsupervised outlier detection approach to detecting abnormal events, based on ideas from generative adversarial networks (GANs) and reinforcement learning (RL), and employs the generator to detect outliers in an online manner.

Abstract

Event sequence data record the occurrences of events in continuous time. Event sequence forecasting based on temporal point processes (TPPs) has been extensively studied, but outlier or anomaly detection, especially without any supervision from humans, is still underexplored. In this work, we develop, to the best our knowledge, the first unsupervised outlier detection approach to detecting abnormal events. Our novel unsupervised outlier detection framework is based on ideas from generative adversarial networks (GANs) and reinforcement learning (RL). We train a 'generator' that corrects outliers in the data with a 'discriminator' that learns to discriminate the corrected data from the real data, which may contain outliers. A key insight is that if the generator made a mistake in the correction, it would generate anomalies that are different from the anomalies in the real data, so it serves as data augmentation for the discriminator learning. Different from typical GAN-based outlier detection approaches, our method employs the generator to detect outliers in an online manner. The experimental results show that our method can detect event outliers more accurately than the state-of-the-art approaches.

Figures (8)

• Figure 1: GAN + RL framework for unsupervised event outlier detection in continuous time.
• Figure 2: Training performance of our algorithm on a synthetic and real dataset. We include two more in Appendix \ref{['sec:more_results']}. The curves are plotted across 10 independent runs. The shaded regions represent the standard errors.
• Figure 3: Model Architectures and Training Framework
• Figure 4: Training performance of our algorithm on one synthetic and real dataset. The curves are plotted across 10 independent runs. The shaded regions represent the standard errors.
• Figure 5: Sensitivity to corruption in the data.