Forecasting experience might shift in direction of approaches and algorithms for evaluating accuracy and mannequin equity. We’ve seen this shift within the giant language mannequin (LLM) house, the place organizations trying to incorporate them must validate fashions and guarantee they’re free from systematic bias. We’ll want to do this identical work in forecasting because it turns into simpler to make use of forecasting fashions in additional settings.
As with LLMs, we are going to see a rising household of forecasting fashions with completely different strengths and weaknesses (efficiency, measurement, velocity, specialization).
On the time of writing, the variety of basis time collection fashions continues to be restricted however rising quick. As the sector evolves, we’ll begin to see extra specialised fashions, for particular industries and functions, to deal with particular knowledge varieties or to present customers finer management to steadiness velocity and accuracy, relying on what’s essential in every job.
It gained’t simply be the fashions that may specialize but additionally the choices and UI for interacting with them. The number of choices obtainable for bettering accuracy, the available exogenous variables, and the way customers entry the fashions through code or interface may even be differentiators.
