Deep Time Series Models: A Comprehensive Survey and Benchmark

2024-07-18

arXiv.org

AI Summary

Deep learning for time series lacks standardized evaluation benchmarks and systematic architectural analysis. Method: We introduce TSLib, the first standardized deep time-series benchmark library, encompassing 24 state-of-the-art models, 30 cross-domain datasets, and five core tasks. It decouples modeling paradigms along two dimensions—fundamental modules and holistic architectures—to establish a modular, reproducible, and fair evaluation framework. Contribution/Results: Through extensive experiments, we uncover, for the first time, strong structural-task alignment patterns—i.e., specific architectural characteristics (e.g., attention mechanisms, convolutional depth, or recurrence design) exhibit consistent performance advantages on particular task types (e.g., forecasting, classification, anomaly detection). This finding provides empirical guidance for model selection and architecture design. Comprehensive evaluation across 12 advanced models validates the robustness of this insight. All code, data, and evaluation pipelines are publicly released and have garnered significant attention from the research community.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Computer Vision: Large Vision Models
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Time series, characterized by a sequence of data points arranged in a discrete-time order, are ubiquitous in real-world applications. Different from other modalities, time series present unique challenges due to their complex and dynamic nature, including the entanglement of nonlinear patterns and time-variant trends. Analyzing time series data is of great significance in real-world scenarios and has been widely studied over centuries. Recent years have witnessed remarkable breakthroughs in the time series community, with techniques shifting from traditional statistical methods to advanced deep learning models. In this paper, we delve into the design of deep time series models across various analysis tasks and review the existing literature from two perspectives: basic modules and model architectures. Further, we develop and release Time Series Library (TSLib) as a fair benchmark of deep time series models for diverse analysis tasks, which implements 24 mainstream models, covers 30 datasets from different domains, and supports five prevalent analysis tasks. Based on TSLib, we thoroughly evaluate 12 advanced deep time series models on different tasks. Empirical results indicate that models with specific structures are well-suited for distinct analytical tasks, which offers insights for research and adoption of deep time series models. Code is available at https://github.com/thuml/Time-Series-Library.

Problem

• Surveying deep learning approaches for time series analysis
• Benchmarking model performance across diverse analytical tasks
• Developing comprehensive library for time series model evaluation

Innovation

• Developed comprehensive benchmark for deep time series models
• Implemented 30 models across 30 multi-domain datasets
• Evaluated model-task suitability through empirical analysis