SAN

Abstract

1. Introduction

Time series data suffer from a non-stationary issue where the statistical properties or the distributions of the data vary rapidly over time. We further argue that the distribution is inconsistent across compact time slices and such inconsistency is not just on a per-instance basis. To alleviate the impact of such property, we propose a model-agnostic normalization framework named SAN. SAN models the non-stationarity in the fine-grained temporal slices and explicitly learn to estimate future distributions, simplifying the the non-stationary forecasting task through divide and conquer.

We conduct comparison experiments on 9 widely used datasets with mainstream forecasting backbones, we also compare the performance of SAN and other plug-and-play non-stationary methods.

This repo is the official Pytorch implementation of our NeurIPS 2023 paper: Adaptive Normalization for Non-stationary Time Series Forecasting: A Temporal Slice Perspective.

2. Objective

Pytorch implementation of NIPS'23 paper: Adaptive Normalization for Non-stationary Time Series Forecasting: A Temporal Slice Perspective

This project demonstrates how Python can be applied to a real-world Data Science problem.

4. Technology Stack

5. System Requirements

General requirements for this technology stack — check the README for exact versions.

• Python 3.8 or later
• pip / virtualenv for dependencies
• VS Code, PyCharm or Jupyter Notebook
• Git (to clone the repository)

6. Installation & Setup

git clone https://github.com/ustc-time-series/SAN.git
cd SAN

pip install -r requirements.txt
mkdir datasets

Full setup instructions are in the project README.

7. Future Enhancements

Suggested extensions you can add to make this your own project.

• Turn the analysis into an interactive dashboard
• Automate data refresh with a scheduled job
• Add a predictive model on top of the analysis

8. Viva / Review Questions

Common questions examiners ask for projects in this domain.

1. What is the source of the dataset and how was missing data handled?
2. Which exploratory analysis steps revealed the most useful insight?
3. Why were these particular charts chosen to present the data?
4. Which statistical or ML technique supports the conclusions?
5. How could the analysis be automated or refreshed with new data?

9. Source Code & License

This project is developed by ustc-time-series and published on GitHub under the Apache License 2.0. Please follow the license terms and credit the original author when you use or modify this code.