Sensitivity Analysis of History Size to Forecast Skill with ARIMA in Python
Last Updated on August 28, 2019
How much history is required for a time series forecast model?
This is a problem-specific question that we can investigate by designing an experiment.
In this tutorial, you will discover the effect that history size has on the skill of an ARIMA forecast model in Python.
Specifically, in this tutorial, you will:
- Load a standard dataset and fit an ARIMA model.
- Design and execute a sensitivity analysis of the number of years of historic data to model skill.
- Analyze the results of the sensitivity analysis.
This will provide a template for performing a similar sensitivity analysis of historical data set size on your own time series forecasting problems.
Kick-start your project with my new book Time Series Forecasting With Python, including step-by-step tutorials and the Python source code files for all examples.
Let’s get started.
- Updated Aug/2017: Fixed a bug where the models were constructed on the raw data instead of the seasonally differenced version of the data. Thanks David Ravnsborg!
- Updated Jun/2018: Removed duplicated sentence. Thanks Rahul!
- Updated Apr/2019: Updated the link to dataset.
- Updated Aug/2019: Updated data loading to use new API.