ProtoTS: Learning Hierarchical Prototypes for Explainable Time Series Forecasting

2025-09-27

AI Summary

Existing interpretable time-series models primarily focus on local, static variable attribution, failing to capture dynamic, heterogeneous inter-variable synergies and their joint influence on global temporal patterns (e.g., trend-cycle coupling). To address this, we propose ProtoTS—a hierarchical prototype-based framework featuring three levels: (i) a denoised instance representation layer, (ii) a local pattern prototype layer for interpretable subsequence modeling, and (iii) a global trend prototype layer for macro-level structural abstraction. Prediction and explanation are unified via multi-granularity similarity matching between instances and prototypes. ProtoTS supports expert-guided interactive interpretation while maintaining high forecasting accuracy and decision transparency. Evaluated on multiple real-world benchmark datasets, ProtoTS consistently outperforms state-of-the-art models in both prediction performance and interpretability. It delivers controllable, verifiable, multi-level causal explanations—enhancing model trustworthiness and practical deployability.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Computer Vision: Interpretability, Explainability, and Transparency
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics

Abstract

While deep learning has achieved impressive performance in time series forecasting, it becomes increasingly crucial to understand its decision-making process for building trust in high-stakes scenarios. Existing interpretable models often provide only local and partial explanations, lacking the capability to reveal how heterogeneous and interacting input variables jointly shape the overall temporal patterns in the forecast curve. We propose ProtoTS, a novel interpretable forecasting framework that achieves both high accuracy and transparent decision-making through modeling prototypical temporal patterns. ProtoTS computes instance-prototype similarity based on a denoised representation that preserves abundant heterogeneous information. The prototypes are organized hierarchically to capture global temporal patterns with coarse prototypes while capturing finer-grained local variations with detailed prototypes, enabling expert steering and multi-level interpretability. Experiments on multiple realistic benchmarks, including a newly released LOF dataset, show that ProtoTS not only exceeds existing methods in forecast accuracy but also delivers expert-steerable interpretations for better model understanding and decision support.

Problem

• Understanding deep learning decisions in time series forecasting scenarios
• Revealing how input variables jointly shape temporal forecast patterns
• Providing hierarchical interpretability beyond local partial explanations

Innovation

• Learning hierarchical prototypes for explainable forecasting
• Computing similarity via denoised heterogeneous representations
• Organizing prototypes for multi-level interpretability and steering