A Multi-Source Integrated Benchmark for Student Academic Risk Prediction (Education)

Abstract

The ARPS Integrated Dataset serves as a high-fidelity, multidimensional benchmark for Educational Data Mining (EDM), consolidating 11,523 records across 40 features. Engineered through a SQL Server 2022 ETL pipeline, it synthesizes six heterogeneous sources—including LMS behavioral logs (xAPI-Edu-Data), secondary-school performance records (UCI), and multi-factor performance datasets (Kaggle) into a unified, cohesive analytical repository. Architected under a Star Schema, the dataset ensures a 0% null rate and high structural integrity. It offers a holistic student profile across seven thematic domains: Identifiers, Demographics, Academic Performance, Behavioral Engagement, Family/Social Background, Study Characteristics, and Environmental Factors. This integration enables deep correlation analysis between digital behaviors, socioeconomic status, and academic outcomes. By providing a unified view of these disparate factors, the repository supports the development of predictive frameworks to categorize students into tiered risk levels (Low, Medium, and High). The dataset includes mixed data types: numerical (grades, GPA, absences), categorical (gender, parental education), and identifier fields. It is distributed in multiple formats, including SQL Server tables and UTF-8 encoded CSV files, ensuring compatibility with Python (pandas), R, and standard statistical tools.