The optimal course bidding strategy under limited resource constraint using genetic algorithm

Abstract

In a course bidding system, there are more students than the number of available seats for a course. To enroll a course, students have to bid using their tokens and the system will fill up the available seats with the top bidders. Since the students have limit tokens, they have to allocate their tokens wisely. In this paper, we apply a genetic algorithm to search for the best way to allocate the tokens such that it maximizes the probability of successful enrollment. To estimate the probability, we train a logistic regression on the course registration data and the model achieves 77.11% accuracy. By using the synthesized dataset, we compare the effectiveness of tokens suggested by the genetic algorithm and other approaches such as heuristics and excel build-in solver. The result shows that 69 out of 78 students have a higher average probability of successful enrollment when using the tokens suggested by the genetic algorithm.