Using School-Level Demographic Composition and Socioeconomic Factors to Predict Dropout Rates in Louisiana (Poster 7)

• In Event: AERA Poster Session Saturday 3:20 p.m.
  In Poster Session: SIG-Educational Statistician Poster Session

Abstract

Louisiana's high school dropout rate has a significant impact locally and nationally. Educational outcomes are influenced by social and economic factors. Using data spanning three academic years, including variables like gender composition, race demographics, academic performance, and teacher and administrative salaries, we employed tree-based machine learning techniques, specifically decision trees, bagging, and random forest to predict dropout rates. Our approach allowed exploring variable relationships and assessing their importance across multiple academic cycles. Notably, our study innovatively incorporates individual school’s demographic characteristics as predictive variables.
Based on Root Mean Square Error (RMSE) and Mean Absolute Error (MAE), we validated the efficacy of our predictive models. Findings illustrate the utility of statistical learning for guiding targeted interventions aimed at reducing dropout rates.

Authors

• Paul H. Atandoh, Mercer University

• Amira Lashawn Abdul-Rahim, Ball State University