There have been a series of sometimes jaw-dropping developments in data science in the last few years, with large language models by far the most prominent (and with good reason). But another story has been the huge explosion in time series packages.
Were you really a tech firm circa 2020–2023 if you didn’t release your own time series package? Looking at what’s available and from who, maybe not: Facebook/Meta got the ball rolling with Prophet, but since then we’ve seen ones from Uber, LinkedIn, Amazon, Google, and Meta again. And it’s not hard to see why time series forecasting might be so valuable at these digital-first, data-rich firms. Just as with data orchestration tools, everyone else is seeing some benefit from their labours.
In the rest of this post, we’ll look at the new(ish) time series packages that are around, who built them, and what they might be good for.