CHINA·77779193永利(集团)有限公司-Official website

题 目 ： Goodness of Fit Assessment for Binary  Classification Learning

报告人：Yuhong Yang  (University of Minnesota)  

Abstract：

Assessing a binary regression model based on ungrouped data is a commonly encountered but very challenging problem. Although tests, such as Hosmer–Lemeshow test, have been devised and widely used in applications, they  often have low power in detecting lack of fit and not much theoretical justification has been made on when they can  work well. In this talk, we will present two new approaches to address the problem. The first approach is based on a  cross-validation (CV) voting system: The model under examination is compared to a nonparametric method and a  lack of fit is declared when the nonparametric method wins the competition. Under some mild conditions, the CV  comparison leads to both type I and II error probabilities converging to zero. The second approach intends to control  the probability of type I error while enhancing the power. It is also applicable to assess a general classification learning  method (e.g., neural network). The new methodology, named binary regression adaptive grouping goodness-of-fit  test (BAGofT), is a two-stage solution where the first stage adaptively selects candidate partitions using "training"  data, and the second stage performs tests based on "test" data. A proper data splitting ensures that the test has  desirable size and power properties. From our experimental results, BAGofT performs much better than the Hosmer-Lemeshow and related tests in many situations.

报告时间：7 月 14 日（周三）下午 4:00-5:00

报告地点：首都师范大学本部教二楼 教室 727

联系人：邹国华

欢迎全体师生积极参加！