feature: add time-series models with statsmodels and river library

[jsamantaucd]

Feature request

Feature request I have implemented and tested two examples that trains a time series model such as ARIMA or River ML model and serves it using BentoML's custom runner to make predictions.

This feature was tested on Bentoml v1.4

Motivation

I believe it would be a nice guide to someone exploring time series models with custom runner implementation for models like ARIMA. River offers a diverse selection of ML models for online learning that are much more efficient than the sklearn models, and integrating this MLflow would create a comprehensive ML solution.

Other

This was already requested as a feature in: https://github.com/bentoml/BentoML/pull/4979 https://github.com/bentoml/BentoML/pull/4896

And as suggested I have created two standalone repositories for the above two examples: https://github.com/jsamantaucd/BentoStatsmodel https://github.com/jsamantaucd/BentoRiverModel

Please kindly review and link it in the README.

[frostming]

Thanks for that, Please test if they work on bentoml 1.4

[jsamantaucd]

Hi @frostming, I have tested with the latest bentoml 1.4 version, updated requirements, and this works fine. Regards!

[jsamantaucd]

Hi @frostming, could you please add the two examples as part of Bentoml, this would be nice addition for someone exploring stats and online learning models, I already tested with the latest version, Thanks!