Autoregressive integrated moving average

The traditional defense against that chaos was time series forecasting.

Planners would use models like ARIMA.

ARIMA, right.

Autoregressive integrated moving average.

It sounds incredibly fancy.

Um, Fink et al.

used a seasonal ARIMA model, which is a great model.

Um, I actually love different types of case series analysis, so I've, I've, I've done it both by negative binomial regression as well as by the seasonal ARIMA.

And, um, I, I, you know, I, I like to do multiple sensitivity analysis to make sure that it isn't about the model.

Um, I'm not sure that Fink did that in his article.

Um, but there were five thousand five hundred and forty-seven observed hangings, um, between August and December, um, in the United States, and the expected was four thousand four hundred.

That's eleven hundred and fifty extra deaths by hanging, uh, plus twenty-six percent.

And both my models agree within five percent, so when I do the ARIMA and the binomial regression.

This is where the tech gets genuinely sci-fi.

Treasury platforms are now deploying machine learning, specifically LSTM models, which stands for long short-term memory, to replace traditional forecasting algorithms like ARIMA.

The shift is huge.

The research is definitive.

LSTM models are yielding thirty percent less error in cash predictions.

I understand that it works, but you need to explain to me how an LSTM model actually achieves that level of accuracy over a traditional statistical model.

To understand why LSTM is superior, you have to understand the limitation of traditional models like ARIMA.

Okay.