Online Ridge Regression Method Using Sliding Windows

Author

Arce, Pedro ; Salinas, Luis

Author_Institution

Center for Technol. Innovation in High Performance Comput., UTFSM, Valparaiso, Chile

fYear

2012

fDate

12-16 Nov. 2012

Firstpage

87

Lastpage

90

Abstract

A new regression method based on the aggregating algorithm for regression (AAR) is presented. The proposal shows how ridge regression can be modified in order to reduce the number of operations by avoiding the inverse matrix calculation only considering a sliding window of the last input values. This modification allows algorithm expression in a recursive way and therefore its use in an online context. Ridge regression, AAR and our proposal were compared using the closing stock prices of 45 stocks from the technology market from 2000 to 2012. Empirical results show that our proposal performs better than the other two methods in 28 of 45 stocks analyzed, due to the lower MSE error.

Keywords

financial data processing; learning (artificial intelligence); mathematics computing; matrix inversion; regression analysis; AAR; aggregating algorithm-for-regression; closing stock prices; inverse matrix calculation avoidance; machine learning; online learning; online ridge regression method; sliding windows; Context; Equations; Mathematical model; Prediction algorithms; Predictive models; Proposals; Vectors; Machine Learning; Online Learning; Ridge Regression;

fLanguage

English

Publisher

ieee

Conference_Titel

Chilean Computer Science Society (SCCC), 2012 31st International Conference of the

Conference_Location

Valparaiso

ISSN

1522-4902

Print_ISBN

978-1-4799-2937-5

Type

conf

DOI

10.1109/SCCC.2012.18

Filename

6694077