ST-Tree with Interpretability for Multivariate Time Series Classification

TL;DR

The ST-Tree model is proposed, which combines ST as the backbone network with an additional neural tree model to improve accuracy in multivariate time series classification tasks and provides interpretability through visualizing the decision-making process across different datasets.

Abstract

Multivariate time series classification is of great importance in practical applications and is a challenging task. However, deep neural network models such as Transformers exhibit high accuracy in multivariate time series classification but lack interpretability and fail to provide insights into the decision-making process. On the other hand, traditional approaches based on decision tree classifiers offer clear decision processes but relatively lower accuracy. Swin Transformer (ST) addresses these issues by leveraging self-attention mechanisms to capture both fine-grained local patterns and global patterns. It can also model multi-scale feature representation learning, thereby providing a more comprehensive representation of time series features. To tackle the aforementioned challenges, we propose ST-Tree with interpretability for multivariate time series classification. Specifically, the ST-Tree model combines ST as the backbone network with an additional neural tree model. This integration allows us to fully leverage the advantages of ST in learning time series context while providing interpretable decision processes through the neural tree. This enables researchers to gain clear insights into the model's decision-making process and extract meaningful interpretations. Through experimental evaluations on 10 UEA datasets, we demonstrate that the ST-Tree model improves accuracy in multivariate time series classification tasks and provides interpretability through visualizing the decision-making process across different datasets.

Figures (10)

• Figure 1: ST-Tree's transparent decision-making process, with the original Epilepsy time series on the left, the tree structure in the middle, and the positions of the prototypes corresponding to the original time series is on the right.
• Figure 2: Schematic of ST-Tree classification via tree structure, with the original Epilepsy time series on the left, the tree structure in the middle. The nodes in the tree compute routing scores on the basis of the prototypes. In the figure, $z_{0} - z_{6}$ represent the prototype of each node. $R_{0, 1} - R_{2, 6}$ represent the probability of the left or right child node.
• Figure 3: Schematic of ST module.
• Figure 4: CD diagram of 9 implementations on 10 UEA datasets.
• Figure 5: Comparison of the accuracy of ST-Tree and ST-Tree (without tree), ST-Tree (without attention), the closer to the upper left corner ST-Tree works better.