QBUS3820 Lecture Notes - Lecture 7: Dependent And Independent Variables, False Positives And False Negatives, Conditional Independence

• School: University of Sydney
• Department: Business Analytics
• Course: QBUS3820
• Professor: Peter

Document Summary

Discipline of business analytics, the university of sydney business school. Module 10: classi cation i: brief review of lasso and ridge, classi cation, review of the bayes" rule, introduction to decision theory for classi cation, k-nearest neighbours classi er, na ve bayes classi er, model evaluation for binary classi cation. The lasso (least absolute shrinkage and selection operator) method solves the following problem nxi=1(cid:16)yi 0 b lasso = argmin where is a tuning parameter. pxj=1. The ridge regression method solves the following problem: nxi=1(cid:16)yi 0 b ridge = argmin pxj=1. 2 j where is a tuning parameter. Best subset, ridge, and lasso when predictors are orthonormal. I(cid:16)|b ols b ols b ols j )(|b ols sign(b ols b lasso j =b ols b lasso j =b ols b lasso j = 0 m |(cid:17) | is one of the k largest |b ols. 2 and if j .