Modeling education impact: a machine learning-based approach for improving the quality of school education

Improving literacy through efficacious school education is an essential objective of a society to reduce poverty and inequality. Thus, a micro-level analysis conducted using Geospatial Artificial Intelligence Modeling (GeoAI) and Machine learning (ML) identified the critical factors affecting student success rates in the West Garo Hills (WGH) district of Meghalaya. A total of 69 data variables comprising school infrastructure, teachers, student enrollment and performance, and relevant teaching-learning information assimilated with regional demographic and assets data surrounding the schools are considered. The methodology involves data aggregation, statistical analysis including dimensionality reduction, feature engineering, and subsequent geospatial AI fusion, spatial autocorrelation, and geospatial buffer analysis for deducing valuable insights. Random Forest model shows that the presence of boundary walls, number of qualified and total teachers, school category, Pupil-Teacher ratio, enrollments, a company of a playground, no. of classrooms, electricity, and drinking water (reflects better infrastructure), and literate/Illiterate ratio are some of the most important factors affecting student performance. A boundary wall in schools can be an important variable as it helps to retain students within school premises resulting in better pedagogical impact. Significant gaps were observed in the presence of schools at all four levels, i.e., primary, upper primary, secondary, and senior secondary, in 7 blocks of WGH. The decision tree regressor model is used for forecasting the pass percentage of students in subsequent years with an accuracy of 93%. The research creates a novel microanalysis school education tool for the stakeholders, and reforms based on these findings can lead to a solemn positive impact in the education sector.