Granular Elastic Network Regression with Stochastic Gradient Descent

Linear regression is the use of linear functions to model relationship between a dependent variable and one or more independent variables. models have been widely used in various fields such as finance, industry, medicine. To address problem that traditional difficult handle uncertain data, we propose granule-based elastic network model. First construct granules granular vectors by granulation methods. Then, define multiple operation rules so can effectively data. Further, norm vector are defined design loss function After that, conduct derivative gradient descent optimization algorithm. Finally, performed experiments on UCI datasets verify validity elasticity network. We found has advantage good fit compared with