# NeuralProphet: Explainable Forecasting at Scale

**Authors:** Oskar Triebe, Hansika Hewamalage, Polina Pilyugina, Nikolay Laptev, Christoph Bergmeir, Ram Rajagopal  
**arXiv:** [2111.15397](https://arxiv.org/abs/2111.15397)  
**Date:** 2021-11-29  
**Summary page:** https://www.summarizepaper.com/en/paper/2111.15397v1/

## Key points

- NeuralProphet is a new forecasting framework based on PyTorch and trained with standard deep learning methods
- It is introduced as a successor to Facebook Prophet, which set an industry standard for explainable, scalable, and user-friendly forecasting frameworks
- The proliferation of time series data has made explainable forecasting a challenging task for business and operational decision making
- Hybrid solutions are needed to bridge the gap between interpretable classical methods and scalable deep learning models
- NeuralProphet introduces local context with auto-regression and covariate modules that can be configured as classical linear regression or as Neural Networks
- While Prophet lacks local context, which is essential for forecasting the near-term future and is challenging to extend due to its Stan backend, NeuralProphet retains the design philosophy of Prophet while providing the same basic model components
- Our results demonstrate that NeuralProphet produces interpretable forecast components of equivalent or superior quality to Prophet on a set of generated time series
- NeuralProphet outperforms Prophet on a diverse collection of real-world datasets by improving forecast accuracy by 55% to 92% for short- to medium-term forecasts
- Hybrid solutions like NeuralProphet are necessary because they bridge the gap between interpretable classical methods and scalable deep learning models.
- Overall, NeuralProphet represents an important step forward in explainable forecasting at scale. Its ease-of-use combined with its ability to produce high-quality forecasts make it an attractive option for businesses looking for reliable forecasting frameworks.

## Summary

NeuralProphet, a new forecasting framework based on PyTorch and trained with standard deep learning methods, has been introduced as a successor to Facebook Prophet. While Prophet set an industry standard for explainable, scalable, and user-friendly forecasting frameworks, the proliferation of time series data has made explainable forecasting a challenging task for business and operational decision making. Hybrid solutions are needed to bridge the gap between interpretable classical methods and scalable deep learning models. NeuralProphet is designed to address this challenge by introducing local context with auto-regression and covariate modules that can be configured as classical linear regression or as Neural Networks. While Prophet lacks local context, which is essential for forecasting the near-term future and is challenging to extend due to its Stan backend, NeuralProphet retains the design philosophy of Prophet while providing the same basic model components. The framework makes it easy for developers to extend it according to their needs. Our results demonstrate that NeuralProphet produces interpretable forecast components of equivalent or superior quality to Prophet on a set of generated time series. Moreover, NeuralProphet outperforms Prophet on a diverse collection of real-world datasets by improving forecast accuracy by 55% to 92% for short- to medium-term forecasts. The authors believe that hybrid solutions like NeuralProphet are necessary because they bridge the gap between interpretable classical methods and scalable deep learning models. They view Prophet as a precursor to such solutions but acknowledge its limitations in terms of lacking local context and being difficult to extend due to its Stan backend. By introducing local context with auto-regression and covariate modules that can be configured as classical linear regression or as Neural Networks, NeuralProphet addresses these limitations while retaining the design philosophy of Prophet. Overall, NeuralProphet represents an important step forward in explainable forecasting at scale. Its ease-of-use combined with its ability to produce high-quality forecasts make it an attractive option for businesses looking for reliable forecasting frameworks.