Amping up Time Series Forecasting: Signature Transformation Method in Python. Part 1

Author(s): Pavel Zapolskii

Originally published on Towards AI.

Image generated using Simplified [1]

Friends, I am so excited to share with you a truly amazing invention of Stochastic Process Math— the Signature of Time Series! This is a highly complex issue with a lot of detail to cover, so this article will be a two-part series.

In this first part, I will do my best to convey the definition of a Time Series Signature and the potential of using it for classification and forecast. (By “do my best” I mean I want to protect you as much as possible from the complicated math behind this technology). Without further ado, let me tell you about the 2 main ways Signatures are already being applied with huge success:

Feature Engineering for n-dimensional Time Series instead of common .mean(), .std(), .sum(). [Part 1]Monte Carlo generation of similarly distributed n-dimensional processes for further calculation probabilities, confidence intervals, etc. [Part 2]

The signature approach is a nonparametric robust way to extract characteristic features from data that can later be used for ML models.

Article plan. Image by author.

To begin with, let’s define not the signature but the process of transformation itself — that is, what we do with the data.

Signature Transformation —calculating on internal data of n-dim Time… 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