Class-wise dimension selection improves time series classification accuracy
Improving Multivariate Time Series Classification with Class-Wise Training and Model Aggregation
Machine Learning
Summary
Classifying time series data with many variables is hard because some variables may not help and can confuse the system. The authors proposed a new way to pick the useful variables separately for each class, making the system focus better on important information. They train models for each class using these selected variables and then combine the results for the final decision. This method helps improve accuracy, especially when there are many variables, and makes it easier to understand which variables matter for each class.
What this means in practice
- •For industrial process engineers: Enhance fault detection in complex machinery by improving classification accuracy through class-specific variable selection.
- •For medical device developers: Improve disease diagnosis from multi-sensor time series data by focusing on the most relevant signals for each condition.
Authors
Mouhamadou Mansour Lo, Gildas Morvan, Mathieu Rossi, Fabrice Morganti, David Mercier
Abstract
In this paper, we propose a class-wise dimension (channel) selection framework for Multivariate Time Series Classification (MTSC). Rather than applying a single global dimension selection process, the proposed approach independently identifies informative dimensions for each class. A dedicated learning process is subsequently performed for each class, followed by a fusion stage for final prediction. The objective is to improve the generation of discriminative feature representations while reducing the influence of noisy or non-informative dimensions. The proposed framework is evaluated using MiniRocket, a random kernel-based baseline method. Experimental results indicate that class-wise dimension selection improves the quality of extracted representations and can enhance classification performance, particularly in high-dimensional settings. These findings suggest that incorporating class-specific information into the training process represents a promising direction for MTSC, improving robustness through consistent gains across heterogeneous datasets, and interpretability through the explicit identification of class-relevant dimensions.