数学建模竞赛1996 B J Judge’s Commentary The Outstanding Contest Judging Papers A正文.pdfVIP

Practitioner’s Commentary 337 Judge’s Commentary: The Outstanding Contest Judging Papers Veena B. Mendiratta Bell Labs Lucent Technologies 2000 N. Naperville Road Naperville, IL 60566 The Contest Judging Problem provided the contestants with a challenging real-world problem that lent itself to a range of analysis and modeling methods. In coming up with a “best” selection scheme, the contestants used methods such as rank-ordering, numerical scoring, and bias estimation. What made the problem interesting and challenging to model, and also to judge, were the less well-defined aspects of the problem. For example, how was bias of judges with respect to ranking and scoring papers handled? Also, how was the issue of ensuring that the best papers were not screened out in the early rounds of reading addressed? Consequently, we could not select the winning papers for MCM 1996 based solely on the criterion of how many paper-readings a team’s algorithm required; we also considered how the issues were addressed. Almost all of the successful papers were able to develop a basic model for the ideal case to select the top papers for the specific parameter values specified in the problem statement. A key assumption for the ideal case is that every judge rank-orders or scores all the the papers as the absolute rank-ordering, that is, there is no judge bias. The so-called ideal model, though unrealistic, sets a lower bound for the total number of reads. The stronger papers went significantly beyond the ideal case. Characteristics of the best papers included the following: An explicit modeling of judges’ bias, addressing the issues of systematic bias and the variance of accidental errors in scoring. Estimating statistical bounds on the probability of failure to pick the best papers out of the top papers. In addition, some of the papers