El mètode de la signatura i aplicacions a dades de tipus financer

Abstract

[en] The signature captures a non-parametric characteristic of a data stream that has been previously transformed into a path through an embedding algorithm. In this work, we will introduce the signature method, which consists of embedding a data stream into a path to hen use its signature as a feature in machine learning problems. First, we will define the signature of a path and present its most important properties. Then, we will explore different ways of embedding data streams into a continuous path and examine the application of this method to financial time series data. Finally, we will see how we can implement a supervised machine learning model based on signatures.