摘要预测精度与可解释性是回归模型评价的两个重要依据。由于变量之间的"共线性"和模型的"过拟合"等问题导致OLS(ordinary least squares)估计量并不总是令人满意。通过在传统的回归模型上引入收缩机制,新一代回归模型能获得更好的预测精度和良好的可解释性。文章比较了几种典型基于收缩机制的回归模型,如岭回归、LASSO,并通过实例分析了不同模型的性能与特点。
7TIBSHIRANI, R. Regression shrinkage and selection via the lasso [ J ]. Jounal of the Royal Statistical Society.1996,58( 1 ) :267 -288.
8VOLKER R. The generalized LASSO [ J ]. IEEE Trans- actions on Neural Networks, 2004, 15( 1 ) : 16-28.
9KOH K, KIM S J, BOYD S. An interior-point method for large-scale Ll-regularized logistic regression [ J ]. Journal of Machine Learning, 2007(8) :1519 -1555.
