Dynamic Time Series Model Updating

Author(s): Shenggang Li

Originally published on Towards AI.

Enhancing Forecast Accuracy with Incremental Adjustments Without Rebuilding Models

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

Imagine you run an online store with a recommendation system. Each time a customer clicks or buys something, the system instantly updates its suggestions.

If you had to rebuild the whole model whenever new data came in, it would be slow and expensive. Instead, we use incremental adjustments, allowing the predictive model to update its recommendations quickly after each interaction.

This method is known as incremental learning or online learning. It allows the system to continuously improve its predictions without refitting the model.

Building and applying business predictive models is a key part of my job. I enjoy working with time series analysis because it’s practical and interesting.

But I’ve learned that this work is quite challenging:

First, time changes everything. A time series forecasting model that works today might be outdated tomorrow. This makes it hard to keep up.

Second, verifying models is difficult because you can only test them with future data, meaning you won’t know if they truly work until later.

Third, training data is often valid only within a certain time range, which is frequently unknown, making it difficult to maintain reliable forecasts over time.

Lastly, many software… 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