A weighted method based on Lars algorithm

2017 3rd International Conference on Applied Materials and Manufacturing Technology

Chen, Lin^1 ; Chen, Shanxiong^1 ; Chen, Chunrong^1 ; Zhu, Yuchen^1

College of Computer and Information Science, Southwest University, Chongqing, China^1

关键词: Coefficient of variation;    Dependent variables;    Entropy weight method;    Least absolute shrinkage and selection operators;    Least angle regressions;    Linear regression models;    Linear relationships;    Regression coefficient;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/242/1/012110/pdf
DOI  :  10.1088/1757-899X/242/1/012110

来源: IOP

【 摘 要 】

LASSO (Least Absolute Shrinkage and Selection Operator) is mainly used to realize variable selection, simultaneously its algorithm and some improved algorithm have gotten wide attention in many fields. To improve the accuracy of LASSO problem in calculating regression coefficients, this paper proposes a new algorithm based on LASR (Least Angle Regression) algorithm to change its approximation direction, which uses two weighted method (coefficient of variation method or entropy weight method) to calculate the weight of linear relationship between the independent and the dependent variables, so we can calculate a regression coefficients set of linear regression model. Compared with LARS algorithm, it can be proved that the improved algorithm mentioned in this paper has a more precise ability for prediction.