Mask aware state space model improves irregular time series classification
MASCIT: A Mask-Aware State Space Classifier for Naturally Irregular Time Series
Machine LearningArtificial Intelligence
Summary
Classifying time series data that is irregular—meaning it has missing points, uneven intervals, or different lengths—is tricky for computers because these gaps can confuse patterns. The authors created a method called MASCIT that tells the model which parts of the data are missing and carefully skips over invalid steps when analyzing the sequence. This approach helps the model focus on the real information without losing important time-related patterns. When tested on many datasets, MASCIT performed better than other methods, especially among neural network models.
What this means in practice
- •For healthcare data teams: Improve classification of patient monitoring data where measurements occur at irregular times with missing entries.
- •For industrial sensor operators: Enhance detection of machine states from sensor streams that suffer from outages and asynchronous sampling.
Authors
Yoo-Min Jung, Hyeon-Gi Kim, Jonghun Park
Abstract
Naturally irregular time series combine asynchronous observations, missing values, unequal lengths, and nonuniform sampling, while dense adapters can discard temporal structure. We propose a mask-aware state space classifier for irregular time series (MASCIT), which supplies observation masks to the encoder and excludes invalid steps from gated temporal aggregation. Across 34 irregular time series datasets, MASCIT yielded the strongest aggregate point estimate and was the only evaluated neural model with three-seed results on every dataset. MASCIT retained the lowest point rank across six overlapping irregularity indicators, while factorial ablations favored partial over full selectivity. These results support selective state space models as effective, executable backbones for naturally irregular time series classification.