Effective Predictions of Gaokao Admission Scores for College Applications in Mainland China

TL;DR

This paper introduces mathematical models to accurately predict university admission scores in China's Gaokao exam, outperforming traditional methods and aiding students in making informed application choices.

Contribution

The paper presents novel, data-driven mathematical models for predicting Gaokao university admission scores, filling a gap left by empirical early prediction methods.

Findings

Prediction accuracy of 91% within 7 points on a 750-point scale.

Models outperform traditional teacher and expert methods.

Significant improvement in early score prediction reliability.

Abstract

Gaokao is the annual academic qualification examination for college admissions in mainland China. Organized by each provincial-level administrative region (PAR), Gaokao takes place at the same time nationwide in early June. To enroll in a university in September, students must take Gaokao and submit common applications for admission to their home PAR Gaokao office in July, listing a small and fixed number of universities and majors they intend to attend and study. About 9.5 million high- school seniors participate in Gaokao every year, and the Gaokao scores are good for just one year. A student has a strong chance to be accepted if their Gaokao score is better than the admission scores of the universities they selected in their applications. However, the admission scores of universities are unknown at the time when filling out applications, which to be determined dynamically during the admission process and may fluctuate from year to year. To increase their chances of acceptance to a best- suited university, students need to predict admission score of each university they are interested in. Early prediction methods are empirical without the backing of in-depth data studies. We fill this void by presenting well-tested mathematical models based on the ranking of Gaokao scores in a PAR. We show that our methods significantly outperform the methods commonly used by teachers and experts, and can predict admission scores with an accuracy of 91% within a 7-point margin in an exam of a 750-point grading scale.