Demystifying Time Series Outliers – 4/4

Author(s): Andrea Ianni

Originally published on Towards AI.

Last stop: end of the line

We’ve reached the end of our saga, which began on a cold sunny day with the entry of a blonde-haired kid onto the field, a pandemic ago.

We’ve tracked the growth of the footballer, on the green pitch.

We’ve witnessed the evolution of the public figure, with a surge in fame that… not even Nicolò himself expected!

First Chapter: a storm of tweets

As we visually spotted sudden spikes in tweets at multiple points within the function, we arrived at the reassuring conclusion that this was clearly an error in fetching data from Twitter. End of the article. Time for everyone to head home.

Unfortunately, none of us, including myself, Moro, and Zappa, appeared eager to pay Gatti a visit and have a candid conversation with him. So, hesitating like never before, we delved into a discussion about outlier management. We emphasized the significance of identifying them carefully by thoroughly evaluating the context, selecting the appropriate cleaning approach, and determining whether removal was necessary.

With a certain note of satisfaction, we drew the final lesson:

At this point, however, wrapping it up with “they’re all errors!!!” appeared somewhat hypocritical.

Second Chapter: watch out for the outliers!

As we delved into working with the residuals, we discovered that there were… 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