Nixtla/neuralforecast
Neuralforecast
Neuralforecast is a neural time series forecasting library designed to predict future values for one or multiple series using deep learning architectures. It functions as a distributed machine learning forecasting framework that enables the training of global models across multiple time series to improve generalization through cross-learning.
The project distinguishes itself as a probabilistic forecasting toolkit that produces uncertainty intervals and probability distributions rather than single point estimates. It also includes a hierarchical forecast reconciler to ensure that predictions across different organizational or geographical levels remain consistent with their aggregate totals.
The library covers a broad range of capabilities, including the integration of exogenous variables, automated hyperparameter optimization, and transfer learning. It provides tools for model validation via cross-validation and sliding window techniques, as well as the ability to distribute training workloads across compute clusters using Spark.
The system supports model persistence via cloud storage integration and provides mechanisms for interpreting forecast components such as trend and seasonality.
Features
- Time Series Forecasting - Provides a comprehensive library for predicting future values of one or multiple time series using deep learning architectures.
- Time Series Forecasting - Predicts future values of one or multiple time series using a variety of deep learning architectures.
- Distributed Training Frameworks - Integrates with Apache Spark to distribute the training of forecasting models across compute clusters for large datasets.
- Probabilistic Loss Functions - Optimizes model parameters to output probability distributions and uncertainty intervals instead of point estimates.
- Time Series Deep Learning Libraries - Offers a specialized deep learning framework for the forecasting, regression, and classification of sequential time series data.
- Global Forecasting Models - Trains single neural networks across multiple time series to improve generalization through cross-learning.
- Probabilistic Forecasting - Generates uncertainty intervals and probability distributions for forecasts instead of single point estimates.
- Exogenous Variable Integration - Integrates external data from the past or future into the model to improve forecast accuracy.
- Global Forecasting Models - Trains shared models across multiple time series to leverage common patterns and improve generalization.
- Global Modeling - Trains global models across multiple time series simultaneously to improve generalization through cross-learning.
- Large-Scale Model Training - Distributes training of large-scale forecasting models across compute clusters using Apache Spark.
- Distributed Training - Scales the training of forecasting models across multiple compute nodes using Apache Spark.
- Forecasting Interpretation Tools - Decomposes predictions to analyze the specific influence of trend, seasonality, and exogenous variables.
- Model Interpretability - Provides tools for explaining model predictions through trend and seasonality visualization and feature attribution.
- Hyperparameter Optimization - Implements automated search and selection of the best configuration parameters for forecasting models.
- Model Performance Evaluators - Quantifies model accuracy using scale-dependent, percentage-based, and scale-independent error metrics.
- Cross-Validation Utilities - Simulates multiple historical forecasts using a validation set for early stopping and hyperparameter selection.
- Predictive Maintenance Models - Implements supervised regression to analyze sensor data and estimate the remaining useful life of systems before failure.
- Long-Horizon Forecasting - Predicts a large number of future timestamps using self-attention to capture long-range dependencies.
- Forecast Evaluation - Measures time series prediction accuracy using error metrics and parametric likelihoods.
- Hierarchical Reconciliation - Adjusts forecasts across nested aggregation levels to ensure consistency between parent and child totals.
- Multivariate Forecasting - Predicts multiple interdependent time series variables simultaneously by capturing multivariate interactions.
- Time Series Backtesting - Provides sliding window and cross-validation techniques to assess forecasting accuracy and tune hyperparameters.
- Remaining Useful Life Estimation - Analyzes sensor data through supervised regression to predict the remaining useful life of a system.
- Transfer Learning - Enables loading pre-trained models for use on new datasets to leverage previously learned temporal patterns.
- Forecasting Validations - Assesses forecasting accuracy by iteratively shifting training and validation sets across historical data.
- Hierarchical Forecast Reconciliation - Adjusts forecasts across different levels of a hierarchy to ensure aggregate totals remain consistent.
- Time Series - Scalable neural network-based time series forecasting.
- Time Series Analysis - Scalable neural network-based forecasting algorithms.
- Time Series Analysis - Neural network-based time series forecasting models.
- Analysis Toolkits - High-performance neural network models for forecasting.
Star history
How this analysis was created: This summary and feature list are AI-generated from collected project material and can contain mistakes. Stars, license and language are imported from GitHub. Inclusion does not mean that we have tested or audited this project. Check the source documentation for any feature you depend on. Learn more on our About page.
