Models

Time Series Forecasting Models @2024

We will be examining some of the Time Series Forecasting Models that published in 2024.

@Fevzi KILAS(aka niexche)

# Currently there are 168 models. Implementation will start from best models.
# Explanations will take time. Thank you for patient.

Time Series Forecasting Models

Stochastic Models - Moving Average (MA) Model - Autoregressive (AR) Model - Autoregressive Moving Average (ARMA) Model - Autoregressive Integrated Moving Average (ARIMA) Model - Seasonal ARIMA (SARIMA) Model - Exponential Smoothing Models - Vector Autoregression (VAR) Model: - Vector Autoregression Moving-Average (VARMA) Model - Vector Autoregression Moving-Average with Exogenous Regressors (VARMAX) Model - Seasonal Decomposition of Time Series (STL) Model - Seasonal-Trend decomposition using LOESS (STL) Model - Seasonal-Trend decomposition using LOESS with ARIMA errors (STL-ARIMA) Model - Seasonal-Trend decomposition using LOESS with Exogenous Regressors (STL-X) Model - Seasonal-Trend decomposition using LOESS with ARIMA errors and Exogenous Regressors (STL-ARIMA-X) Model

ML Models - RNN and LSTM Networks - CNN Networks - Encoder-Decoder Networks - Attention Networks - Transformer Networks - WaveNet Networks - TCN Networks - DeepAR Networks - DeepState Networks - DeepFactor Networks

Evaluation Metrics - Mean Absolute Error (MAE) - Mean Squared Error (MSE) - Root Mean Squared Error (RMSE) - Mean Absolute Percentage Error (MAPE) - Symmetric Mean Absolute Percentage Error (SMAPE) - Mean Absolute Scaled Error (MASE) - Theil’s U Statistic - R2 Score - Adjusted R2 Score - Akaike Information Criterion (AIC) - Bayesian Information Criterion (BIC) - Mean Forecast Error (MFE) - Mean Absolute Deviation (MAD) - Tracking Signal (TS) - Forecast Bias - Forecast Accuracy - Forecast Precision - Forecast Recall - Forecast F1 Score - Forecast ROC AUC Score - Forecast PR AUC Score - Forecast Confusion Matrix - Forecast Classification Report - Forecast ROC Curve - Forecast PR Curve - Forecast Calibration Curve - Forecast Learning Curve - Forecast Residuals Plot - Forecast Residuals Density Plot - Forecast Residuals Distribution Plot - Forecast Residuals Q-Q Plot - Forecast Residuals Autocorrelation Plot - Forecast Residuals Partial Autocorrelation Plot - Forecast Residuals Lag Plot - Forecast Residuals Scatter Plot - Forecast Residuals vs. Time Plot - Forecast Residuals vs. Predicted Plot - Forecast Residuals vs. Actual Plot - Forecast Residuals vs. Forecast Plot - Forecast Residuals vs. Error Plot - Forecast Residuals vs. Squared Error Plot - Forecast Residuals vs. Absolute Error Plot - Forecast Residuals vs. Absolute Percentage Error Plot

PatchTST: A Time Series is Worth 64 Words: Long-term Forecasting with Transformers

17 Apr 2024

Paper Link

Implementation Link

TimesFM: A decoder-only foundation model for time-series forecasting

17 Apr 2024

Paper Link

Implementation Link

Chronos: Learning the Language of Time Series

12 Mar 2024

Paper Link

Implementation Link

iTransformer: Inverted Transformers Are Effective for Time Series Forecasting

14 Mar 2024

Paper Link

Implementation Link

TimeMixer: Decomposable Multiscale Mixing for Time Series Forecasting

23 May 2024

Paper Link

Implementation Link

Lag-Llama: Towards Foundation Models for Probabilistic Time Series Forecasting

08 Feb 2024

Paper Link

Implementation Link

Unified Training of Universal Time Series Forecasting Transformers

04 Feb 2024

Paper Link

Implementation Link

A Survey on Diffusion Models for Time Series and Spatio-Temporal Data

11 Jun 2024

Paper Link

Implementation Link

UniTS: Building a Unified Time Series Model

29 Feb 2024

Paper Link

Implementation Link

Tiny Time Mixers (TTMs): Fast Pre-trained Models for Enhanced Zero/Few-Shot Forecasting of Multivariate Time Series

09 Apr 2024

Paper Link

Implementation Link

#### Buraya kadar incele sonrasi date order

Disentangled Interpretable Representation for Efficient Long-term Time Series Forecasting

Long Term TSF

DiPE-Linear incorporates three temporal components: - Static Frequential Attention (SFA): Extracts and enhances prominent frequencies in the frequency domain. - Static Temporal Attention (STA): Selects input time points in the time domain that are strongly correlated with the predicted sequence. - Independent Frequential Mapping (IFM): Assumes independence across different frequencies and directly maps the input series to the output series at each frequency component in frequency domain.

These components alternate between learning in the frequency and time domains to achieve disentangled inter-pretability.

26 Nov 2024

Paper Link

Implementation Link