diviner 0.1.1

Is this right for my project?

Diviner is meant to help with large-scale forecasting. Instead of describing each individual use case where it may be applicable, here is a non-exhaustive list of projects that it would fit well as a solution for:

• Forecasting regional sales within each country that a company does business in per day

• Predicting inventory demand at regional warehouses for thousands of products

• Forecasting traveler counts at each airport within a country daily

• Predicting electrical demand per neighborhood (or household) in a multi-state region

Each of these examples has a common theme:

• The data is temporally homogenous (all of the data is collected daily, hourly, weekly, etc.).

• There is a large number of individual models that need to be built due to the cardinality of the data.

• There is no guarantee of seasonal, trend, or residual homogeneity in each series.

• Varying levels of aggregation may be called for to solve different use cases.

The primary problem that Diviner solves is managing the execution of many discrete time-series modeling tasks. Diviner provides a high-level API and metadata management approach that relieves the operational burden of managing hundreds or thousands of individual models.

Grouped Modeling Wrappers

Currently, Diviner supports the following open source libraries for forecasting at scale:

• prophet

• pmdarima

Installing

Install Diviner from PyPI via:

pip install diviner

Documentation

Documentation, Examples, and Tutorials for Diviner can be found here.

Community & Contributing

For assistance with Diviner, see the docs.

Contributions to Diviner are welcome. To file a bug, request a new feature, or to contribute a feature request, please open a GitHub issue. The team will work with you to ensure that your contributions are evaluated and appropriate feedback is provided. See the contributing guidelines for submission guidance.