Adapting Time Series Classification Algorithms to Regression

• Grant number: 23/05041-9
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Effective date (Start): May 01, 2023
• Effective date (End): July 31, 2024
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Diego Furtado Silva
• Grantee: Andre Guarnier De Mitri
• Host Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil
• Associated research grant: 22/03176-1 - Machine learning for time series obtained in mHealth applications, AP.PNGP.PI
• Associated scholarship(s): 23/11745-9 - Experimenting deep learning-based strategies to deal with multivariate time series tasks., BE.EP.IC

Abstract

With the growing interest in time series in recent decades, which has been intensifying in recent years, researchers have proposed dozens of specific algorithms for classifying and forecasting time series. However, there is still a great demand for techniques associated with the extrinsic regression of time series. By "extrinsic," it is understood that the target of the prediction is an external value, in opposition to the forecasting definition. For example, when considering time series of physiological signals, extrinsic regression would aim to estimate clinical parameters, such as blood oxygen saturation. Given that extrinsic regression is a critical task to the project that comprises this proposal, and considering the lack of algorithms for this task, this activity will identify the mechanisms used in designing time series classification algorithms and adapt them to the regression task. Due to the recent advancement of deep learning in time series classification, this work will focus on analyzing neural architectures in the context of extrinsic regression. This work will study the advantages and disadvantages of different approaches to define guidelines for using these networks in the researched task.