# Nixtla

## Docs

- [Differences](https://nixtla-mintlify-2a3f5c2a.mintlify.site/coreforecast/differences.md): Find the optimal number of differences
- [Expanding](https://nixtla-mintlify-2a3f5c2a.mintlify.site/coreforecast/expanding.md): Compute expanding mean, std, min, max, and quantile
- [Exponentially weighted](https://nixtla-mintlify-2a3f5c2a.mintlify.site/coreforecast/exponentially_weighted.md): Compute exponentially weighted mean
- [Grouped Array](https://nixtla-mintlify-2a3f5c2a.mintlify.site/coreforecast/grouped_array.md): Group arrays by a categorical variable
- [coreforecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/coreforecast/index.md): Fast implementations of common forecasting routines
- [Lag transformations | CoreForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/coreforecast/lag_transforms.md): Compute lag transforms
- [Rolling](https://nixtla-mintlify-2a3f5c2a.mintlify.site/coreforecast/rolling.md): Compute rolling mean, std, min, max, and quantile
- [Scalers](https://nixtla-mintlify-2a3f5c2a.mintlify.site/coreforecast/scalers.md): Scale arrays
- [Seasonal](https://nixtla-mintlify-2a3f5c2a.mintlify.site/coreforecast/seasonal.md): Find the seasonal period
- [Utils](https://nixtla-mintlify-2a3f5c2a.mintlify.site/coreforecast/utils.md)
- [Favorita](https://nixtla-mintlify-2a3f5c2a.mintlify.site/datasetsforecast/favorita.html.md): Favorita dataset
- [Hierarchical](https://nixtla-mintlify-2a3f5c2a.mintlify.site/datasetsforecast/hierarchical.html.md): Hierarchical dataset
- [datasetsforecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/datasetsforecast/index.html.md): Datasets for time series forecasting
- [Long Horizon](https://nixtla-mintlify-2a3f5c2a.mintlify.site/datasetsforecast/long_horizon.html.md): Download and wrangling utility for long-horizon datasets.
- [Long-Horizon Original Datasets](https://nixtla-mintlify-2a3f5c2a.mintlify.site/datasetsforecast/long_horizon2.html.md): Download and wrangling utility for long-horizon datasets. These datasets have been used by `NHITS, AutoFormer, Informer, PatchTST, TiDE` among many other neural forecasting methods. The datasets include the original [ETTh1, ETTh2, ETTm1, ETTm2, Weather, ILI, TrafficL](https://github.com/zhouhaoyi/ET…
- [M3](https://nixtla-mintlify-2a3f5c2a.mintlify.site/datasetsforecast/m3.html.md): M3 dataset
- [M4](https://nixtla-mintlify-2a3f5c2a.mintlify.site/datasetsforecast/m4.html.md): M4 dataset
- [M5](https://nixtla-mintlify-2a3f5c2a.mintlify.site/datasetsforecast/m5.html.md): M5 dataset
- [PHM2008](https://nixtla-mintlify-2a3f5c2a.mintlify.site/datasetsforecast/phm2008.html.md): PHM2008 dataset
- [Utils | DatasetsForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/datasetsforecast/utils.html.md): Utility functions for datasetsforecast
- [Core | HierarchicalForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/core.html.md): Core
- [Hierarchical Evaluation](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/evaluation.html.md)
- [Bootstrap](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/examples/australiandomestictourism-bootstraped-intervals.html.md)
- [Normality](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/examples/australiandomestictourism-intervals.html.md)
- [Multi-model Aggregation](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/examples/australiandomestictourism-multimodel.html.md)
- [PERMBU](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/examples/australiandomestictourism-permbu-intervals.html.md)
- [Geographical Aggregation (Tourism)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/examples/australiandomestictourism.html.md)
- [Geographical and Temporal Aggregation (Tourism)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/examples/australiandomestictourismcrosstemporal.html.md)
- [Temporal Aggregation (Tourism)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/examples/australiandomestictourismtemporal.html.md)
- [Geographical Aggregation (Prison Population)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/examples/australianprisonpopulation.html.md)
- [Exogenous Variables](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/examples/exogenousvariables.html.md)
- [Hierarchical Forecasting at Scale](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/examples/hierarchicalforecastingatscale.html.md)
- [Tutorials](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/examples/index.html.md)
- [Install HierarchicalForecast with pip or conda](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/examples/installation.html.md): Install HierarchicalForecast with pip, conda, or uv, plus source-install instructions for cloning the repo and creating a virtual environment.
- [Introduction](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/examples/introduction.html.md)
- [Local vs Global Temporal Aggregation](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/examples/localglobalaggregation.html.md)
- [Temporal Aggregation with THIEF](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/examples/m3withthief.html.md)
- [Neural/MLForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/examples/mlframeworksexample.html.md)
- [Non-Negative MinTrace](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/examples/nonnegativereconciliation.html.md)
- [Probabilistic Reconciliation Methods Comparison](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/examples/probabilistic-reconciliation-comparison.html.md)
- [Reconciliation Diagnostics](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/examples/reconciliationdiagnostics.html.md)
- [Probabilistic Forecast Evaluation](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/examples/tourismlarge-evaluation.html.md)
- [Quick Start | HierarchicalForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/examples/tourismsmall.html.md)
- [Quick Start (Polars)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/examples/tourismsmallpolars.html.md)
- [Hierarchical Forecast 👑](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/index.html.md): Probabilistic hierarchical forecasting with statistical and econometric methods
- [Reconciliation Methods](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/methods.html.md)
- [Probabilistic Methods](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/probabilistic_methods.html.md)
- [Aggregation/Visualization Utils](https://nixtla-mintlify-2a3f5c2a.mintlify.site/hierarchicalforecast/utils.html.md)
- [Nixtlaverse](https://nixtla-mintlify-2a3f5c2a.mintlify.site/index.md)
- [Auto](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/auto.html.md)
- [Callbacks](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/callbacks.html.md): Utility functions use in the predict step.
- [Conformal Prediction](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/conformal_prediction.html.md): Conformal prediction intervals and transfer conformal methods
- [Core | MLForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/core.html.md)
- [Distributed Forecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/distributed.forecast.html.md): Distributed pipeline encapsulation
- [DaskLGBMForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/distributed.models.dask.lgb.html.md): dask LightGBM forecaster
- [DaskXGBForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/distributed.models.dask.xgb.html.md): dask XGBoost forecaster
- [RayLGBMForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/distributed.models.ray.lgb.html.md): ray LightGBM forecaster
- [RayXGBForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/distributed.models.ray.xgb.html.md): ray XGBoost forecaster
- [SparkLGBMForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/distributed.models.spark.lgb.html.md): spark LightGBM forecaster
- [SparkXGBForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/distributed.models.spark.xgb.html.md): spark XGBoost forecaster
- [End to end walkthrough | MLForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/getting-started/end_to_end_walkthrough.html.md)
- [Install | MLForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/getting-started/install.html.md)
- [Quick start (distributed)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/getting-started/quick_start_distributed.html.md)
- [Quick start (local)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/getting-started/quick_start_local.html.md)
- [Analyzing the trained models](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/how-to-guides/analyzing_models.html.md)
- [Cross validation | MLForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/how-to-guides/cross_validation.html.md)
- [Custom date features](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/how-to-guides/custom_date_features.html.md)
- [Custom training](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/how-to-guides/custom_training.html.md)
- [Exogenous features](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/how-to-guides/exogenous_features.html.md)
- [Hyperparameter optimization | MLForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/how-to-guides/hyperparameter_optimization.html.md)
- [Lag transformations | MLForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/how-to-guides/lag_transforms_guide.html.md)
- [MLflow | MLForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/how-to-guides/mlflow.html.md)
- [One model per step](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/how-to-guides/one_model_per_horizon.html.md)
- [Pooled lag transforms](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/how-to-guides/pooled_lag_transforms.html.md)
- [Predict callbacks](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/how-to-guides/predict_callbacks.html.md)
- [Predicting a subset of ids](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/how-to-guides/predict_subset.html.md)
- [Probabilistic forecasting | MLForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/how-to-guides/prediction_intervals.html.md)
- [Sample weights](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/how-to-guides/sample_weights.html.md)
- [Using scikit-learn pipelines](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/how-to-guides/sklearn_pipelines.html.md)
- [Target transformations](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/how-to-guides/target_transforms_guide.html.md)
- [Training with numpy arrays](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/how-to-guides/training_with_numpy.html.md)
- [Transfer Learning | MLForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/how-to-guides/transfer_learning.html.md)
- [Transforming exogenous features](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/how-to-guides/transforming_exog.html.md)
- [Electricity Load Forecast | MLForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/tutorials/electricity_load_forecasting.html.md)
- [Detect Demand Peaks | MLForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/tutorials/electricity_peak_forecasting.html.md)
- [Incremental Forecast generation](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/tutorials/incremental_forecasting.html.md)
- [M4 Competition](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/tutorials/m4.html.md)
- [Prediction intervals](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/tutorials/prediction_intervals_in_forecasting_models.html.md)
- [Transfer Conformal Prediction Intervals](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/docs/tutorials/transfer_conformal_prediction.html.md)
- [Feature engineering | MLForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/feature_engineering.html.md): Compute transformations on exogenous regressors
- [MLForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/forecast.html.md): Full pipeline encapsulation
- [Grouped Array](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/grouped_array.md): Something abou `Grouped Array`
- [Machine Learning 🤖 Forecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/index.html.md): Scalable machine learning for time series forecasting
- [Lag transforms](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/lag_transforms.html.md): Built-in lag transformations
- [LightGBMCV](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/lgb_cv.html.md): Time series cross validation with LightGBM.
- [Optimization](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/optimization.html.md): Utilities for hyperparameter optimization
- [Target transforms](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/target_transforms.html.md)
- [Utils | MLForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/mlforecast/utils.html.md)
- [Hyperparameter Optimization | NeuralForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/common.base_auto.html.md): BaseAuto class for hyperparameter optimization in NeuralForecast. Integrates Optuna, HyperOpt, Dragonfly through Ray for automated model tuning with cross-validation.
- [NN Modules](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/common.modules.html.md): Neural network building blocks for NeuralForecast: MLP layers, temporal convolutions, Transformer encoders-decoders, attention mechanisms, and embeddings.
- [TemporalNorm](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/common.scalers.html.md): TemporalNorm: Temporal normalization techniques for neural forecasting. Scalers include standard, robust, invariant, and RevIN for distribution shift handling.
- [Core | NeuralForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/core.html.md): NeuralForecast core class for high-level time series forecasting. Fits multiple PyTorch models on pandas DataFrames with parallelization and distributed computation.
- [NeuralForecast Map](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/api-reference/neuralforecast_map.html.md)
- [Categorical features in NeuralForecast models](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/capabilities/categorical_features.html.md): Learn how NeuralForecast handles categorical exogenous features, how to encode discrete variables, and which deep learning models support them.
- [Cross-validation | NeuralForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/capabilities/cross_validation.html.md)
- [Exogenous Variables](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/capabilities/exogenous_variables.html.md)
- [Hyperparameter Optimization | NeuralForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/capabilities/hyperparameter_tuning.html.md)
- [Optimization Objectives](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/capabilities/objectives.html.md)
- [Forecasting Models](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/capabilities/overview.html.md)
- [Predict Insample](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/capabilities/predictInsample.html.md)
- [Save and Load Models](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/capabilities/save_load_models.html.md)
- [Time Series Scaling](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/capabilities/time_series_scaling.html.md)
- [Data Requirements](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/getting-started/datarequirements.html.md)
- [Installation](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/getting-started/installation.html.md)
- [About NeuralForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/getting-started/introduction.html.md)
- [Quickstart](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/getting-started/quickstart.html.md)
- [Adding Models to NeuralForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/adding_models.html.md)
- [Statistical, Machine Learning and Neural Forecasting methods| NeuralForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/comparing_methods.html.md)
- [Modify the configure_optimizers() behavior of NeuralForecast models](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/configure_optimizers.html.md)
- [Uncertainty quantification with Conformal Prediction](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/conformal_prediction.html.md)
- [Converting Models to ONNX](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/converting_onnx.html.md)
- [Cross-validation| NeuralForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/cross_validation.html.md)
- [Distributed Training](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/distributed_neuralforecast.html.md)
- [Explainability for Deep Learning Forecasting Models](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/explainability.html.md)
- [Forecasting with TFT: Temporal Fusion Transformer](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/forecasting_tft.html.md)
- [End to End Walkthrough | NeuralForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/getting_started_complete.html.md)
- [Hierarchical Forecast | NeuralForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/hierarchical_forecasting.html.md)
- [Intermittent Data](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/intermittent_data.html.md)
- [Interpretable Decompositions](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/interpretable_decompositions.html.md)
- [Using Large Datasets](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/large_datasets.html.md)
- [Long-Horizon Forecasting with NHITS](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/longhorizon_nhits.html.md)
- [Long-Horizon Probabilistic Forecasting](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/longhorizon_probabilistic.html.md)
- [Long-Horizon Forecasting with Transformer models](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/longhorizon_transformers.html.md)
- [Multivariate Forecasting with TSMixer](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/multivariate_tsmixer.html.md)
- [Robust Forecasting](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/robust_forecasting.html.md)
- [Generate simulation paths for probabilistic forecasts](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/simulation.html.md): Use NeuralForecast to generate correlated sample paths for scenario analysis, compare simulations with prediction intervals, and quantify future uncertainty.
- [Temporal Classification](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/temporal_classification.html.md)
- [Transfer Learning | NeuralForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/transfer_learning.html.md)
- [Probabilistic Forecasting | NeuralForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/uncertainty_quantification.html.md)
- [Using MLflow](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/using_mlflow.html.md)
- [Weighting Timesteps | NeuralForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/tutorials/weighting_timesteps.html.md)
- [Detect Demand Peaks | NeuralForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/use-cases/electricity_peak_forecasting.html.md)
- [Predictive Maintenance](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/docs/use-cases/predictive_maintenance.html.md)
- [NumPy Evaluation](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/losses.numpy.html.md): Comprehensive NumPy evaluation metrics for NeuralForecast including MAE, MSE, MAPE, MASE, and probabilistic losses for time series forecast accuracy.
- [PyTorch Losses](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/losses.pytorch.html.md): PyTorch loss functions for neural forecast training: MAE, MSE, MAPE, quantile losses, distribution losses, and robust losses for model optimization.
- [Autoformer](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.autoformer.html.md): Autoformer: Transformer with auto-correlation mechanism and progressive decomposition for reliable long-horizon time series forecasting with trend-seasonality.
- [BiTCN](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.bitcn.html.md): BiTCN: Bidirectional Temporal Convolutional Network for forecasting. Parameter-efficient architecture with forward-backward encoding for probabilistic predictions.
- [DeepAR probabilistic RNN forecasting model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.deepar.html.md): DeepAR is a probabilistic autoregressive RNN that uses Monte Carlo sampling and distribution outputs to quantify forecast uncertainty for time series.
- [DeepNPTS](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.deepnpts.html.md): DeepNPTS: Deep Non-Parametric Time Series forecaster that samples from empirical distributions. Strong baseline for probabilistic forecasting tasks.
- [DilatedRNN model for long-sequence forecasting](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.dilated_rnn.html.md): DilatedRNN uses dilated skip connections to model long time series sequences, mitigating vanishing gradients and improving computational efficiency.
- [DLinear](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.dlinear.html.md): DLinear model: Simple, fast linear architecture with trend-seasonality decomposition for accurate long-horizon time series forecasting with minimal complexity.
- [FEDformer](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.fedformer.html.md): FEDformer: Frequency Enhanced Decomposition transformer for long-term forecasting using Fourier transform and sparse attention in frequency domain.
- [GRU gated recurrent unit forecasting model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.gru.html.md): GRU is a Gated Recurrent Unit model for sequential forecasting. It improves on LSTM with a simplified gating mechanism and an MLP decoder for predictions.
- [HINT](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.hint.html.md): HINT: Hierarchical Mixture Networks for coherent probabilistic forecasting. Combines neural architectures with reconciliation for hierarchical time series.
- [Automatic Forecasting](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.html.md): AutoModel classes for NeuralForecast hyperparameter optimization. Automated grid search, Bayesian optimization with Ray Tune for 34 forecasting architectures.
- [Informer](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.informer.html.md): Informer: Efficient Transformer with ProbSparse attention for long-sequence time series forecasting. Reduces O(L^2) complexity for scalable predictions.
- [iTransformer](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.itransformer.html.md): iTransformer: Inverted Transformer architecture for multivariate time series forecasting with attention on time points and feed-forward on series dimensions.
- [KAN](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.kan.html.md): KAN: Kolmogorov-Arnold Networks for time series forecasting. MLP alternative using learnable activation functions for improved non-linear pattern modeling.
- [LSTM recurrent neural network forecasting model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.lstm.html.md): LSTM is a Long Short-Term Memory recurrent network for time series forecasting, with a multilayer encoder-decoder that handles vanishing gradients.
- [MLP](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.mlp.html.md): MLP: Multi-Layer Perceptron for time series forecasting. Simple feedforward neural network with ReLU activations and autoregressive structure for predictions.
- [MLPMultivariate joint multivariate forecasting model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.mlpmultivariate.html.md): MLPMultivariate is a multi-layer perceptron that predicts all time series jointly, using shared feedforward layers for multivariate forecasting.
- [NBEATS](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.nbeats.html.md): NBEATS: Neural Basis Expansion Analysis with interpretable or generic configurations. MLP-based architecture with residual links for M3/M4 competition performance.
- [NBEATSx](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.nbeatsx.html.md): NBEATSx: Neural Basis Expansion Analysis with exogenous variables. MLP-based architecture with interpretable trend-seasonality blocks for forecasting.
- [NHITS](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.nhits.html.md): NHITS: Neural Hierarchical Interpolation for Time Series. MLP architecture with multi-rate processing for long-horizon forecasting, 50x faster than Informer.
- [NLinear](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.nlinear.html.md): NLinear: Normalized linear model for long-horizon forecasting. Handles distribution shifts with simple subtraction-addition normalization for robust predictions.
- [PatchTST](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.patchtst.html.md): PatchTST: Efficient Transformer model for multivariate forecasting using patched time series and channel-independence for scalable long-term predictions.
- [Reversible Mixture of KAN - RMoK](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.rmok.html.md): RMoK: Reversible Mixture of Kolmogorov-Arnold Networks. Combines Taylor, Jacobi, and wavelet functions for expressive time series forecasting with reversibility.
- [RNN Elman recurrent network forecasting model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.rnn.html.md): RNN is the classic Elman recurrent neural network for time series forecasting, with tanh or ReLU activations and an MLP decoder over stacked layers.
- [SOFTS](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.softs.html.md): SOFTS: Spectral Optimal Fourier Transform model for multivariate time series forecasting using frequency-domain analysis and temporal pattern recognition.
- [SOFTSSharp](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.softssharp.html.md): SOFTSSharp: SOFTS extension with stochastic variable-position encoding for multivariate time series forecasting.
- [StemGNN](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.stemgnn.html.md): StemGNN: Spectral Temporal Graph Neural Network for multivariate forecasting. Learns temporal dependencies and inter-series correlations in spectral domain.
- [TCN temporal convolutional forecasting model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.tcn.html.md): TCN is a Temporal Convolutional Network that uses dilated causal convolutions to capture long-range dependencies for efficient time series forecasting.
- [TFT](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.tft.html.md): TFT: Temporal Fusion Transformer with interpretable multi-horizon forecasting. LSTM encoder, multi-head attention, variable selection for complex time series.
- [TiDE](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.tide.html.md): TiDE: Time-series Dense Encoder with MLP-based architecture. Encoder-decoder model for long-term univariate forecasting with exogenous input support.
- [Time-LLM](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.timellm.html.md): Time-LLM: Reprograms large language models for time series forecasting. Transforms forecasting tasks into language tasks using off-the-shelf LLM backbones.
- [TimeMixer](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.timemixer.html.md): TimeMixer: Temporal mixing architecture for multivariate time series forecasting with multi-scale decomposition and frequency-domain feature extraction.
- [TimesNet](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.timesnet.html.md): TimesNet: 2D-variation modeling with Inception blocks for capturing intraperiod and interperiod temporal patterns in univariate time series forecasting.
- [TimeXer transformer for multivariate forecasting](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.timexer.html.md): TimeXer is a cross-series attention transformer for multivariate time series forecasting, using patch-based processing and exogenous variable support.
- [TSMixer](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.tsmixer.html.md): TSMixer: MLP-based multivariate forecasting with time and feature mixing. Stacked mixing layers learn temporal and cross-sectional representations jointly.
- [TSMixerx MLP forecasting with exogenous variables](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.tsmixerx.html.md): TSMixerx extends TSMixer with exogenous variables, combining MLP-based temporal-feature mixing with static and future covariate support for forecasts.
- [Vanilla Transformer](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.vanillatransformer.html.md): Vanilla Transformer: Classic attention-based architecture for time series. Full O(L^2) attention mechanism with encoder-decoder for long-sequence forecasting.
- [XLinear gated MLP model for multivariate forecasts](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.xlinear.html.md): XLinear is an MLP-based multivariate forecasting model that uses temporal and cross-channel gating with a global token to capture patterns across time series.
- [xLSTM extended long short-term memory model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/models.xlstm.html.md): xLSTM extends the LSTM architecture with exponential gating and new sLSTM and mLSTM memory cells for parallelizable sequence modeling and forecasting.
- [PyTorch Dataset/Loader](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/tsdataset.html.md): PyTorch Dataset and DataLoader classes for time series. TimeSeriesDataset and TimeSeriesDataModule for efficient batch processing with Lightning integration.
- [Example Data](https://nixtla-mintlify-2a3f5c2a.mintlify.site/neuralforecast/utils.html.md): NeuralForecast utility functions and datasets. Includes AirPassengers data, time feature generation, prediction intervals, and synthetic panel data generators.
- [Contribute to Nixtla](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/contribute/contribute.html.md)
- [Nixtla Documentation](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/contribute/docs.html.md)
- [Understanding Issue Labels](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/contribute/issue-labels.html.md)
- [Submit an Issue 📢](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/contribute/issues.html.md)
- [Step-by-step Contribution Guide](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/contribute/step-by-step.html.md): This document contains instructions for collaborating on the different libraries of Nixtla.
- [Contributing Code to Nixtla Development](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/contribute/techstack.html.md): A guide on the technical skills and tools needed to contribute code to the Nixtla project.
- [Dask](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/distributed/dask.html.md)
- [Ray](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/distributed/ray.html.md)
- [Spark](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/distributed/spark.html.md)
- [Amazon Forecast vs StatsForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/experiments/amazonstatsforecast.html.md)
- [AutoARIMA Comparison (Prophet and pmdarima)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/experiments/autoarima_vs_prophet.html.md)
- [AutoARIMAProphet Adapter](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/experiments/autoarimaprophet_adapter.html.md)
- [Forecasting at Scale using ETS and ray (M5)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/experiments/ets_ray_m5.html.md)
- [StatsForecast ETS and Facebook Prophet on Spark (M5)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/experiments/prophet_spark_m5.html.md)
- [End to End Walkthrough | StatsForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/getting-started/getting_started_complete.html.md)
- [End to End Walkthrough with Polars](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/getting-started/getting_started_complete_polars.html.md)
- [Quick Start | StatsForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/getting-started/getting_started_short.html.md)
- [Install | StatsForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/getting-started/installation.html.md)
- [Automatic Time Series Forecasting](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/how-to-guides/automatic_forecasting.html.md)
- [Exogenous Regressors](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/how-to-guides/exogenous.html.md)
- [Generating features](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/how-to-guides/generating_features.html.md)
- [Numba caching](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/how-to-guides/numba_cache.html.md)
- [Sklearn models](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/how-to-guides/sklearn_models.html.md)
- [ADIDA Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/adida.html.md)
- [ARCH Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/arch.html.md)
- [ARIMA Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/arima.html.md)
- [AutoARIMA Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/autoarima.html.md)
- [AutoCES Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/autoces.html.md)
- [AutoETS Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/autoets.html.md)
- [AutoRegressive Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/autoregressive.html.md)
- [AutoTheta Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/autotheta.html.md)
- [Conformal Seasonal Pool (CSP)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/conformalseasonalpool.html.md)
- [CrostonClassic Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/crostonclassic.html.md)
- [CrostonOptimized Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/crostonoptimized.html.md)
- [CrostonSBA Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/crostonsba.html.md)
- [Dynamic Optimized Theta Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/dynamicoptimizedtheta.html.md)
- [Dynamic Standard Theta Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/dynamicstandardtheta.html.md)
- [GARCH Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/garch.html.md)
- [Holt Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/holt.html.md)
- [Holt Winters Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/holtwinters.html.md)
- [IMAPA Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/imapa.html.md)
- [MFLES](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/mfles.html.md)
- [Multiple Seasonal Trend (MSTL)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/multipleseasonaltrend.html.md)
- [Optimized Theta Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/optimizedtheta.html.md)
- [Seasonal Exponential Smoothing Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/seasonalexponentialsmoothing.html.md)
- [Seasonal Exponential Smoothing Optimized Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/seasonalexponentialsmoothingoptimized.html.md)
- [Simple Exponential Smoothing Optimized Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/simpleexponentialoptimized.html.md)
- [Simple Exponential Smoothing Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/simpleexponentialsmoothing.html.md)
- [Standard Theta Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/standardtheta.html.md)
- [TSB Model](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/models/tsb.html.md)
- [Anomaly Detection](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/tutorials/anomalydetection.html.md)
- [Conformal Prediction](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/tutorials/conformalprediction.html.md)
- [Cross validation | StatsForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/tutorials/crossvalidation.html.md)
- [Electricity Load Forecast | StatsForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/tutorials/electricityloadforecasting.html.md)
- [Detect Demand Peaks | StatsForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/tutorials/electricitypeakforecasting.html.md)
- [Volatility forecasting (GARCH & ARCH)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/tutorials/garch_tutorial.html.md)
- [Intermittent or Sparse Data](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/tutorials/intermittentdata.html.md)
- [MLFlow | StatsForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/tutorials/mlflow.html.md)
- [Multiple seasonalities](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/tutorials/multipleseasonalities.html.md)
- [Trajectory Simulation](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/tutorials/simulation.html.md)
- [Statistical, Machine Learning and Neural Forecasting methods | StatsForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/tutorials/statisticalneuralmethods.html.md)
- [Probabilistic Forecasting | StatsForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/docs/tutorials/uncertaintyintervals.html.md)
- [Statistical ⚡️ Forecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/index.html.md): Lightning fast forecasting with statistical and econometric models
- [Core Methods](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/src/core/core.html.md): Methods for Fit, Predict, Forecast (fast), Cross Validation and plotting
- [Fugue Backend](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/src/core/distributed.fugue.html.md)
- [Models](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/src/core/models.html.md): Models currently supported by StatsForecast
- [StatsForecast's Models](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/src/core/models_intro.html.md)
- [Feature engineering | StatsForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/statsforecast/src/feature_engineering.html.md): Generate features for downstream models
- [Generation and composition](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/composition.html.md): Reference for generate_series, balanced_pool, pretraining_pool, and Multivariatizer helpers that build panels and correlated channels from generators.
- [BaseGenerator](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/core.html.md): BaseGenerator API reference covering shared parameters for length, frequency, seeding, exogenous variables, missing data, anomalies, and changepoints.
- [Dataset composition and augmentation](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/dataset.html.md): SynSet composes multiple generators into one panel; SynAugment analyzes series, fits generators, and produces statistically similar synthetic augmentations.
- [Anomaly injection](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/capabilities/anomalies.html.md): Inject spikes, dips, and level shifts into synthetic time series with known locations to benchmark anomaly detectors and stress-test forecasters.
- [SynAugment: data augmentation](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/capabilities/augmentation.html.md): Expand real time series panels with SynAugment by fitting a generator per series and drawing statistically matched synthetic copies for global forecasters.
- [Data augmentation with real datasets](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/capabilities/augmentation_real_data.html.md): Augment real M4 series with SynAugment and inspect what the fitted generator preserves, what it varies, and how novel each synthetic draw is.
- [Balanced pool](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/capabilities/balanced_pool.html.md): Explore balanced_pool, the 42-generator preset spanning ARMA, seasonality, chaos, counts, and heavy tails for bias-free benchmarking or pretraining.
- [Changepoint injection](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/capabilities/changepoints.html.md): Inject level, trend, variance, or mixed structural breaks at known positions to benchmark changepoint detectors and forecaster regime adaptation.
- [Write your own generator](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/capabilities/custom_generator.html.md): Subclass BaseGenerator to add a custom data-generating process where one method wires up seeding, timestamps, pattern injection, and dataframe engines.
- [Compose datasets with SynSet](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/capabilities/dataset.html.md): Combine several generators into one long-format panel with SynSet to mix behaviors like random walks and seasonal demand in one model-ready frame.
- [Exogenous variables](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/capabilities/exogenous.html.md): Attach datetime calendar features, cyclical encodings, correlated covariates, and pattern-injection ground-truth flags to any synthetic series.
- [Missing data patterns](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/capabilities/missingness.html.md): Inject random, block, or seasonal missing value patterns into synthetic series to test imputation methods and forecaster tolerance for gaps.
- [Multivariate missing data](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/capabilities/multivariate_missingness.html.md): Apply per-channel missingness to VAR and copula generators while preserving cross-series correlation for multivariate imputation benchmarks.
- [Multivariatize a univariate generator](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/capabilities/multivariatize.html.md): Wrap any univariate generator with Multivariatizer to couple channels via contemporaneous mixing or lead-lag lags for realistic cross-dependencies.
- [Robustness testing with known ground truth](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/capabilities/robustness_testing.html.md): Use synthetic ground-truth labels and clean copies to measure detector precision and recall and quantify how contamination hurts forecast accuracy.
- [When does synthetic data help?](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/capabilities/when_synthetic_helps.html.md): Two benchmarks on M3 and M4 that show augmentation is neutral with ample history but synthetic pretraining helps cold-start zero-shot forecasting.
- [Clickstream generator for web sessions and funnels](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/domain/clickstream.html.md): Generate sessionized web clickstream traffic with hour-of-day and day-of-week seasonality, bot traffic, bounces, conversions, and full funnel metrics.
- [Daily active users generator for product engagement metrics](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/domain/daily_active_users.html.md): Generate daily active user metrics with weekly rhythm, growth trend, marketing event boosts, and consumer, business, or gaming app engagement profiles.
- [Energy load generator for multi-seasonal grid demand](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/domain/energy_load.html.md): Generate electricity load with layered daily and weekly seasonality, temperature sensitivity, holiday effects, and residential or industrial demand shapes.
- [Intermittent demand generator for sparse spare-parts series](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/domain/intermittent_demand.html.md): Generate sparse spare-parts demand series with zero-inflated Poisson, negative binomial, lognormal, or gamma order sizes for Croston-style forecasters.
- [IoT sensor generator with drift and failure artifacts](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/domain/iot_sensor.html.md): Generate IoT sensor readings for temperature, humidity, pressure, and light with calibration drift, battery degradation, and intermittent device failures.
- [State-space generator for linear latent dynamics](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/domain/state_space.html.md): Generate linear state-space model observations with custom transition and observation matrices, process noise, and hidden latent states for Kalman testing.
- [Vital signs generator for clinical monitoring data](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/domain/vital_signs.html.md): Generate clinical vital signs like heart rate, blood pressure, SpO2, and respiration with circadian rhythms, HRV, and healthy or septic patient profiles.
- [Copula generator for cross-channel dependency structure](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/multivariate/copula.html.md): Generate multivariate series with a Gaussian or t-copula that couples channels through a target correlation matrix while keeping arbitrary marginals.
- [Gaussian process generator with kernel-defined covariance](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/multivariate/gaussian_process.html.md): Generate Gaussian process samples with RBF, Matern, or periodic kernels to control smoothness, length scale, and cyclic structure of the covariance.
- [VAR generator for vector autoregression dynamics](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/multivariate/var.html.md): Generate coupled vector autoregression channels with configurable lag order, cross-effect coefficient matrices, and correlated innovation covariance.
- [KernelSynth generator for GP-kernel pretraining series](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/pretraining/kernel_synth.html.md): Generate Chronos-style pretraining series from random compositions of Gaussian process kernels combined with additive and multiplicative operators.
- [Temporal causal model generator for lagged SCM dynamics](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/pretraining/tcm.html.md): Generate temporal structural causal model series from randomly sampled lagged edges with nonlinear edge functions and stochastic node innovations.
- [TSI generator for trend-seasonal-irregular composition](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/pretraining/tsi.html.md): Generate diverse pretraining pools by randomly composing trend, seasonal harmonic, and irregular components in additive or multiplicative form.
- [ETS (exponential smoothing)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/statistical/ets.html.md): Simulate ETS state-space series with level, trend, and seasonal components. Covers 30 additive and multiplicative Holt-Winters variants with damping.
- [INAR (integer counts)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/statistical/inar.html.md): Generate integer-valued autoregressive count series via binomial thinning, with Poisson or negative binomial innovations for demand and arrival data.
- [Random walk](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/statistical/random_walk.html.md): Simulate a random walk process with configurable drift and volatility. The canonical non-stationary baseline for naive forecasts and unit-root tests.
- [SARIMA](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/statistical/sarima.html.md): Generate SARIMA series from known (p,d,q)(P,D,Q) orders with autocorrelation, differencing, and seasonality for model recovery and forecasting benchmarks.
- [Seasonal](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/statistical/seasonal.html.md): Simulate seasonal time series combining periodic waves, linear trend, and additive noise for testing forecast models on daily, weekly, or yearly cycles.
- [Bounded process (proportions and rates)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/stochastic/bounded_process.html.md): Simulate bounded time series in (0, 1) using Beta-AR or logit-normal models for proportions, utilization rates, and market shares with mean reversion.
- [Chaotic systems](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/stochastic/chaotic_system.html.md): Simulate deterministic chaotic dynamics with Lorenz attractor, logistic map, and Mackey-Glass equations to test nonlinear time-series models.
- [Cyclic (irregular cycles)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/stochastic/cyclic.html.md): Generate irregular cyclic time series with drifting periods and amplitudes for business cycles, ecological booms, and quasi-periodic aperiodic signals.
- [Fractional Brownian motion (long memory)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/stochastic/fractional_brownian_motion.html.md): Simulate fractional Brownian motion with tunable Hurst exponent for long-range dependence, persistent trends, and anti-persistent mean-reverting increments.
- [GARCH (volatility clustering)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/stochastic/garch.html.md): Generate GARCH(p,q) returns with autocorrelated conditional variance to reproduce volatility clustering and heteroscedastic bursts in financial data.
- [Geometric Brownian motion](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/stochastic/geometric_brownian_motion.html.md): Simulate geometric Brownian motion for strictly positive, log-normal series with exponential drift and multiplicative volatility, as in Black-Scholes pricing.
- [Hawkes process (self-excitation)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/stochastic/hawkes_process.html.md): Generate self-exciting Hawkes point process series with clustered event bursts for trades, earthquakes, and social-media cascade arrival modeling.
- [Jump diffusion](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/stochastic/jump_diffusion.html.md): Simulate Merton jump-diffusion prices combining continuous drift, Gaussian volatility, and Poisson jumps for shock-prone assets and news-driven gaps.
- [Levy process (heavy tails)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/stochastic/levy_process.html.md): Simulate alpha-stable Levy processes with heavy-tailed increments and large jumps, generalizing Brownian motion for extreme moves and Cauchy-like tails.
- [Ornstein-Uhlenbeck (mean reversion)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/stochastic/ornstein_uhlenbeck.html.md): Simulate Ornstein-Uhlenbeck mean-reverting series pulled toward an equilibrium level, modeling interest-rate spreads, temperatures, and pairs trading data.
- [Poisson process (event arrivals)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/stochastic/poisson_process.html.md): Generate homogeneous Poisson event counts with a constant arrival rate as the memoryless baseline for queues, click streams, and failure processes.
- [Regime switching (Markov switching)](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/stochastic/regime_switching.html.md): Simulate Markov-switching time series with hidden regimes, distinct means and volatilities, and sticky transitions for bull-bear markets and cycle breaks.
- [Stochastic volatility](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/generators/stochastic/stochastic_volatility.html.md): Generate Heston and SABR stochastic-volatility paths with latent mean-reverting variance and leverage correlation for derivatives pricing and returns.
- [Install SynForecast for synthetic time series](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/getting-started/installation.html.md): Install SynForecast with pip or uv, verify the compiled Rust extension, and add optional dataframe engines like PyArrow, Modin, or cuDF.
- [SynForecast quickstart for synthetic time series](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/getting-started/quickstart.html.md): Generate synthetic time-series panels with SynForecast: one-line generation, explicit generators, injected anomalies, and mixed SynSet datasets.
- [MLForecast with synthetic data](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/integrations/mlforecast.html.md): Train MLForecast global models with SynAugment augmentation and zero-shot pretraining on synthetic panels, benchmarked on the airline passenger series.
- [NeuralForecast with synthetic data](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/integrations/neuralforecast.html.md): Pretrain NeuralForecast NHITS on synthetic monthly series, fine-tune on M4 holdouts, and compare augmentation against observed-only training baselines.
- [Works with the Nixtlaverse](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/integrations/nixtlaverse.html.md): Pass SynForecast panels directly to StatsForecast, MLForecast, and NeuralForecast in Nixtla long format, with augmentation and pretraining workflows.
- [StatsForecast with synthetic data](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/docs/integrations/statsforecast.html.md): Use synthetic panels with StatsForecast local models like AutoETS: augmentation is inert, so validate pipelines against known ETS processes instead.
- [Exogenous variables](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/exogenous.html.md): ExogenousConfig and CorrelatedExogConfig control calendar features, cyclical encodings, anomaly and changepoint flags, and correlated exogenous columns.
- [Domain-specific generators](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/generators_domain.html.md): Domain generators for intermittent demand, IoT sensors, energy load, state-space models, daily active users, vital signs, and web clickstream traffic.
- [Multivariate generators](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/generators_multivariate.html.md): CopulaGenerator, VARGenerator, and GaussianProcessGenerator produce correlated multivariate series with copula, vector autoregression, or GP kernels.
- [Pretraining generators](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/generators_pretraining.html.md): TSIGenerator, TCMGenerator, and KernelSynthGenerator compose randomized trend-seasonal, causal-graph, and GP-kernel processes for foundation-model pretraining.
- [Statistical generators](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/generators_statistical.html.md): Statistical generators covering random walks, seasonal signals, SARIMA, exponential smoothing (ETS), and integer-valued autoregressive INAR counts.
- [Stochastic generators](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/generators_stochastic.html.md): Stochastic generators including GARCH, Ornstein-Uhlenbeck, GBM, jump diffusion, Poisson, cyclic, fBm, Hawkes, regime switching, chaotic, and Lévy processes.
- [Synthetic 🧬 Forecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/synforecast/index.html.md): SynForecast generates synthetic time-series panels with 31 statistical, stochastic, and multivariate generators, plus SynAugment for augmenting observed data.
- [Data](https://nixtla-mintlify-2a3f5c2a.mintlify.site/utilsforecast/data.html.md): Utilies for generating time series datasets
- [Multi-Objective Model Selection with Pareto Frontier](https://nixtla-mintlify-2a3f5c2a.mintlify.site/utilsforecast/docs/tutorials/multi_objective_model_selection.html.md)
- [Rectify Strategy for Multi-Step Forecasts](https://nixtla-mintlify-2a3f5c2a.mintlify.site/utilsforecast/docs/tutorials/rectify_strategy.html.md)
- [Evaluation](https://nixtla-mintlify-2a3f5c2a.mintlify.site/utilsforecast/evaluation.html.md): Model performance evaluation
- [Feature Engineering | UtilsForecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/utilsforecast/feature_engineering.html.md): Create exogenous regressors for your models
- [utilsforecast](https://nixtla-mintlify-2a3f5c2a.mintlify.site/utilsforecast/index.html.md): Forecasting utilities
- [Losses](https://nixtla-mintlify-2a3f5c2a.mintlify.site/utilsforecast/losses.html.md): Loss functions for model evaluation.
- [Plotting](https://nixtla-mintlify-2a3f5c2a.mintlify.site/utilsforecast/plotting.html.md): Time series visualizations
- [Preprocessing](https://nixtla-mintlify-2a3f5c2a.mintlify.site/utilsforecast/preprocessing.html.md): Utilities for processing data before training/analysis

## Optional

- [TimeGPT](https://nixtla.io/docs)