Reliable Neighborhood-Aware Multi-View Outlier Detection

In recent years, multi-view outlier detection (MVOD) has gained increasing attention, with the primary objective of recovering the underlying structure of normal data from outlier-contaminated multi-view datasets. However, this objective is hindered by two fundamental challenges:(i) outlier propagation, (ii) scale discrepancy. To address these issues, we propose RNAMOD (Reliable Neighborhood-Aware Multi-View Outlier Detection), which introduces the concept of reliability and constructs a reliable neighborhood structure to avoid outlier propagation. We introduce a leave-one-out directional consensus mechanism to align cross-view neighborhood structures while preventing scale discrepancy by aligning geometric directions that remain invariant to scaling. Extensive experiments on six benchmark datasets demonstrate that RNAMOD consistently outperforms state-of-the-art methods.

Many real-world objects are described by information from several sources. For example, a news article may have text and images, or a medical record may contain different types of measurements. Sometimes, a small number of items are unusual: they may contain corrupted information, or the information from different sources may not match. Detecting such unusual items is important, but difficult, because these items can also mislead the learning process itself. This paper proposes a new method for finding unusual items in multi-source data. The key idea is to estimate which items and relationships are trustworthy during learning. Less reliable items are prevented from strongly influencing the model, so normal patterns can be learned more accurately. The method also compares relationships across sources in a way that is less affected by differences in numerical scale. Experiments on multiple benchmark datasets show that the proposed method detects unusual items more effectively than existing approaches.