Design of University Student Management Platform Based on Data Mining Technology

This study aims to enhance the informatization efficiency of university student management by focusing on innovative applications of data mining technology in poverty-stricken student subsidy evaluation. Utilizing campus Integrated Circuit (IC) card consumption data from Guilin University of Electronic Technology, a five-dimensional data warehouse model (time, department, specialty, student origin, consumption characteristics) is constructed to systematically elaborate the implementation pathway of data mining technologies. The process includes: (1) employing Extract, Transform, Load (ETL) technology for multi-source data cleansing and integration; (2) applying K-means clustering algorithms to identify student consumption behavior patterns; (3) utilizing Apriori association rules to explore latent relationships between consumption features and family economic conditions; and (4) establishing a random forest classification model for precise identification of impoverished students. Breaking through the limitations of traditional management systems that focus solely on basic data management, this research innovatively integrates Online Analytical Processing (OLAP) multidimensional analysis with data mining to develop an evaluation system containing 12 key indicators across dimensions, including departmental distribution, specialty differences, and geographical origin characteristics. Experimental results demonstrate that the dynamic early-warning model based on consumption behavior mining improves poverty-stricken student identification accuracy by 27.6%, while optimized scholarship allocation solutions achieve 18.9% higher coverage. The proposed technical framework of “data acquisition-feature extraction-model construction-decision support” establishes a replicable paradigm for the intelligent transformation of university management systems. Notably, the dynamic evaluation mechanism accelerates response speed in poverty-stricken student identification by 40%, significantly enhancing the scientific validity of funding decisions.