Evaluating Simplification Algorithms for Interpretability of Time Series Classification
Brigt Håvardstun, Felix Marti-Perez, Cèsar Ferri, Jan Arne Telle
TL;DR
The paper addresses the interpretability of Time Series Classification (TSC) by introducing Complexity and Loyalty as metrics to evaluate simplifications of time series. It analyzes four segment-based simplification algorithms (RDP, VW, BU, OS) under a normalized parameter to study the trade-off between brevity and fidelity, using Rocket as the primary TSC model across 40 UCR datasets and augmenting with a human-forward-simulation study. The results show that Optimal Simplification (OS) and Ramer-Douglas-Peuker (RDP) generally offer the best interpretability prospects, with OS achieving strong AUC at lower complexity but higher computational cost ($O(n^3)$) compared to RDP's $O(n \log n)$, and a practical framework (flowchart) to decide when simplifications are beneficial. The work provides a principled approach to prototype-based explanations in TSC, enabling end-users and systems to gauge when simplified representations will meaningfully support understanding of model decisions.
Abstract
In this work, we introduce metrics to evaluate the use of simplified time series in the context of interpretability of a TSC -- a Time Series Classifier. Such simplifications are important because time series data, in contrast to text and image data, are not intuitively under- standable to humans. These metrics are related to the complexity of the simplifications -- how many segments they contain -- and to their loyalty -- how likely they are to maintain the classification of the original time series. We focus on simplifications that select a subset of the original data points, and show that these typically have high Shapley value, thereby aiding interpretability. We employ these metrics to experimentally evaluate four distinct simplification algorithms, across several TSC algorithms and across datasets of varying characteristics, from seasonal or stationary to short or long. We subsequently perform a human-grounded evaluation with forward simulation, that confirms also the practical utility of the introduced metrics to evaluate the use of simplifications in the context of interpretability of TSC. Our findings are summarized in a framework for deciding, for a given TSC, if the various simplifications are likely to aid in its interpretability.