Algorithms

Table of Contents

  1. Specific Methods introduction
  2. Foundation Methods introduction

Specific methods introduction

TFB evaluated a diverse range of methods, including statistical learning, machine learning, and deep learning methods.

Statistical Learning (SL): ARIMA, ETS, Kalman Filter (Kalman), and VAR

Machine Learning(ML): XGB Model (XGB) , Linear Regression (LR), and Random Forest (RF)

Deep Learning (DL):

  • RNN-based models (RNN)
  • CNN-based models (MICN, TimesNet, and TCN)
  • MLP-based models (NLinear, DLinear, TiDE, N-HiTS, and N-BEATS)
  • Transformer-based models (PatchTST, Crossformer, and FEDformer, Non-stationary Transformer (Stationary), Informer, and Triformer)
  • Model-Agnostic models (FiLM)

Foundation methods introduction

FoundTS envaluated a diverse range of time series foundation models,including time series pre-trained models pretrained with multi-domain time series and LLM-based models pretrained with large-scale text, along with specific models.

Time Series Pre-trained Models (TS Pre-trained Models):

  • Reconstruction methods: MOIRAI, UniTS, Moment
  • Autoregressive methods: TimesFM, Timer
  • Direct prediction methods: TTM
  • Hybrid pre-training methods: ROSE

LLM-based Models:

  • Parameter-efficient fine-tuning methods: GPT4TS, SS2IPLLM
  • Prompting methods: UniTime, Time-LLM
Categorization of comparison methods