LLMAD: Large Language Models can Deliver Accurate and Interpretable Time Series Anomaly Detection.

Description

This repository contains the code for the paper: "Large Language Models can Deliver Accurate and Interpretable Time Series Anomaly Detection". It demonstrates the use of Large Language Models (LLMs) to tackle the task of Time Series Anomaly Detection.

Table of Contents

• Installation
• Usage
  • Configuration
  • Running the Scripts
    • Yahoo Dataset
    • WSD Dataset
    • KPI Dataset
• File Descriptions

Installation

To get started, clone the repository and install the necessary dependencies:

cd LLM_AD
pip install -U openai fastdtw pandas numpy scipy

Usage

Configuration

Before running the scripts, set up your configuration file config.yaml with your OpenAI API details:

openai:
  api_key: "your-api-key"
  base_url: "https://api.openai.com/v1"

Running the Scripts

Below are the commands to run the scripts for different datasets.

Yahoo Dataset

bash script/yahoo.sh

WSD Dataset

bash script/wsd.sh

KPI Dataset

bash script/kpi.sh

File Descriptions

File Name	Description
run.py	Program entry point
Prompt_template.py	Structure of the prompt
Eval/*	Scripts to compute evaluation metrics

If you find this repo helpful, please cite the following papers:

@article{liu2024large,
  title={Large Language Models can Deliver Accurate and Interpretable Time Series Anomaly Detection},
  author={Liu, Jun and Zhang, Chaoyun and Qian, Jiaxu and Ma, Minghua and Qin, Si and Bansal, Chetan and Lin, Qingwei and Rajmohan, Saravan and Zhang, Dongmei},
  journal={arXiv preprint arXiv:2405.15370},
  year={2024}
}