Forecasting: Stories of Time Series, LLMs, Causality, and Cats

Author(s): Dr. Alessandro Crimi

Originally published on Towards AI.

Can a Foundation Model Revolutionize Time Series Forecasting, or are we stuck with Granger causality?

This member-only story is on us. Upgrade to access all of Medium.

What is the effect and the cause? (royalty-free picture from www.pexels.com)

For causality, we define the influence by which one event, process, state, or contributes to the production of another event, process, state, where the cause is at least partly responsible for the effect. A critical element is given by the time component. For example, a grumpy cat is grumpy (effect) because before someone washed it with shampoo against its will (cause). The washing phase has occurred temporally before the being grumpy.

In general terms, causal inference provides a framework that integrates statistical and machine learning methods to answer causal questions from time series data, from the stock market to a brain region activity causing the activity in another brain region. Even if we relax the idea of looking for causality, forecasting time series has enormous importance across all industries; they are indeed used for weather and stock market predictions. Large language models (LLMs) are also tools for prediction. Indeed, a word forecast the following word. If I have two words, “a cat…”, I know the next word will be a verb afterward related to the cat. We can use this… Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI