Home Knowledge Base Time Series Forecasting with Deep Learning

Time Series Forecasting with Deep Learning is the application of neural network architectures to predict future values of temporal sequences — leveraging patterns in historical data including trends, seasonality, and complex nonlinear dependencies, where modern transformer and SSM-based forecasters now compete with and often surpass traditional statistical methods (ARIMA, ETS) on diverse benchmarks from energy demand to financial markets to weather prediction.

Deep Learning Architecture Timeline for Time Series

EraArchitectureKey Advantage
2015-2017LSTM/GRUCaptures sequential dependencies
2017-2019WaveNet/TCN (Temporal CNN)Parallelizable, dilated convolutions
2019-2021Informer/Autoformer (Transformer)Long-range attention, multi-horizon
2022+PatchTST, TimesNetChannel-independent patching
2023+TimesFM, Chronos (Foundation)Pre-trained on many datasets
2024+Mamba/SSM variantsLinear complexity, long sequences

Forecasting Paradigms

ParadigmMethodBest For
Point forecastPredict single future value at each stepSimple predictions
Probabilistic forecastPredict distribution (quantiles, parameters)Risk-aware decisions
Multi-horizonPredict multiple future steps simultaneouslyPlanning applications
MultivariatePredict multiple correlated series jointlyInterconnected systems

PatchTST (2023)

Foundation Models for Time Series

ModelDeveloperApproach
TimesFMGooglePre-trained decoder-only on 100B+ timepoints
ChronosAmazonT5-style tokenization of time series values
Lag-LlamaSalesforceLLaMA-based probabilistic forecaster
MOIRAISalesforceUniversal forecaster, any-variate

Input Representation

Evaluation Metrics

MetricFormulaWhat It Measures
MAEMean Absolute ErrorAverage absolute deviation
MSE/RMSE(Root) Mean Squared ErrorPenalizes large errors
MAPEMean Absolute Percentage ErrorScale-independent accuracy
CRPSContinuous Ranked Probability ScoreProbabilistic forecast quality
WQLWeighted Quantile LossQuantile prediction accuracy

Time series forecasting with deep learning is entering a foundation model era — pre-trained temporal models that generalize across domains are beginning to match or exceed specialized models, promising to make high-quality forecasting accessible without domain expertise, much as language models democratized NLP.

time series forecastingtemporal predictiontime series deep learningforecasting modeltemporal model

Explore 500+ Semiconductor & AI Topics

From EUV lithography to CUDA optimization — search the full knowledge base or chat with our AI assistant.