Research on a Multi-Scale Intelligent Recommendation Method for College Entrance Examination Volunteer Selection Based on Grey Correlation and Information Matching

Abstract

The National College Entrance Examination (Gaokao) is the core mechanism for talent selection in China’s education system and a crucial milestone in students’ academic careers. With the gradual implementation of the new Gaokao reform across the country, the process of filling out college applications has become a focal point of attention for both students and parents. In recent years, along with the rapid advancement of big data and artificial intelligence technologies, various AI-based college recommendation models have emerged. These models can provide school selection suggestions based on students’ scores and provincial rankings. However, most existing models rely heavily on quantitative data such as historical admission scores, resulting in outcomes that remain at the level of statistical matching. Such approaches often fail to comprehensively reflect students’ personalized needs in areas such as academic interests, career orientation, regional preferences, and future development potential. Therefore, current AI-based recommendation methods still exhibit shortcomings in terms of demand coverage, decision adaptability, and scientific guidance.Based on this, this paper analyzes and compares historical admission score data, university information, and candidates’ personal information and preferences. By applying the grey correlation analysis algorithm and relevant machine learning techniques, it aims to provide practical assistance to candidates during the application process, helping them develop application strategies that best meet their personalized needs. Specifically, data on universities’ historical admission scores for each major were collected and integrated from official university websites and enrollment information portals, along with students’ input scores, rankings, and personalized preferences. Evaluation indicators were assessed to determine their weights, forming an evaluation index system. After a preliminary filtering of universities and major groups matching the candidates’ score rankings, the grey correlation algorithm was used to generate an initial recommendation list. Subsequently, to address the issue of matching between university information and candidates’ personalized preferences, both datasets underwent preprocessing and word segmentation. Using the TF-IDF algorithm, key information was extracted from both universities and students’ input preferences. These key elements were then vectorized, and cosine similarity was computed to compare the degree of match between universities and students. Within the preliminary recommendation list derived from the grey correlation analysis, the final list of recommended university-major groups was generated based on the matching results.