High dimensional regression coefficient compression model and its application

Abstract

From stock market risk to genetic data survival analysis to energy consumption impact analysis, statistical modeling of high-dimensional regression plays an important role in different fields. Based on the financial data of China for the past 15 years, we select fifteen predictors related to fiscal revenue, design a ten-fold cross-validation algorithm based on the Ridge Regression and Lasso Regression models. Empirical examples show that Lasso Regression is a great way in big data modeling by comparing the cross-validation mean square error and the equation interpretation ability, which achieves the process of coefficient compression.