美国数学建模竞赛特等奖论文——1996 B O Select the Winners Fast【数学建模】.pdfVIP

Select the Winners Fast 317 Select the Winners Fast Haitao Wang Chunfeng Huang Hongling Rao Center for Astrophysics University of Science and Technology of China Hefei, China Advisor: Qingjuan Yu Summary Assuming that judges are ideal, we provide a model to determine the top papers in almost the shortest time. We use a matrix record the orderings that we get from judges, and we reject as many papers as possible after each round. We then consider real-life judges and estimate the probability that the final papers contain a paper not among the best papers. Furthermore, considering the possibility of systematic bias in a scoring scheme, we improve the model by using a Bayesian estimation method, which makes it possible to some extent to compare different judges’ scores. We performed many computer simulations to test the feasibility of our model. We find that our model would be improved by increasing the num- ber of papers selected from the first round. We also made a stability analysis by altering , , and got an empirical formula to predicts the total time of judging. We used data from real life to test our model and got a perfect result: For , , and , we get the first- and third-best papers with our scheme; with , we got the top three papers. We conclude by summarizing a practicable and flexible scheme and offering some suggestions, and estimating the budget with an empirical formula. Assumptions The judges are equal. None is more authoritative than the others. When a judge is evaluating a paper, the judging result is not influenced by adventitious factors, such as taking bribes. The time that a judge takes is proportional to the number of papers to read. 318 The UMAP Journal 17.3 There exists an objective criterion with which we can tell which of two papers is “better.” There