Track: Data Analytics and Big Data

Abstract

Multicollinearity is the condition that there is a correlation between independent variables which is a problem. This event often occurs in regression analysis. LASSO (Least Absolute Shrinkage and Selection Operator) method regression can reduce multicollinearity and increase the accuracy of linear regression models. The lasso parameter estimator can be solved by the LARS (Least Angle Regression and Shrinkage) algorithm which calculates the correlation vector, the largest absolute correlation value, equiangular vector, inner product vector, and determines the LARS algorithm limiter for LASSO. LASSO method regression with a more detailed procedure and selecting the best model using the  Mallows statistics is discussed in this paper. LASSO method will be applied to Indonesia's foreign exchange deposit data.