Scholarlink

Accurate enrollment forecasts are vital for evidence-based educational planning. We train Long Short-Term Memory (LSTM) networks on ten years of Brazilian School Census data to predict public-school enrolment in Maceió, a low-income urban centre in Brazil’s Northeast. To support equitable resource allocation by the National Textbook and Educational Material Programme (PNLD), we couple the model with Explainable-AI techniques—SHAP values and Granger causality—to uncover the drivers of enrolment dynamics at preschool, elementary, and high-school levels. Results show that infrastructure quality (adapted restrooms, multimedia equipment) is the main lever for preschool access, teacher availability and libraries sustain elementary retention, while science labs, sports facilities, and ICT resources shape high-school demand. Intervariable analysis reveals that targeted budget cuts can trigger indirect declines in enrolment, underscoring the need for holistic investment.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Understanding Public School Enrollment Trends in the Northeast of Brazil: Unraveling Features Through AI Related to the Learning Process for Low-Income Students

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Artur Cavalcante de Jesus,
  • Leonardo Marques,
  • Diego Carvalho do Nascimento,
  • Bruno Almeida Pimentel

摘要

Accurate enrollment forecasts are vital for evidence-based educational planning. We train Long Short-Term Memory (LSTM) networks on ten years of Brazilian School Census data to predict public-school enrolment in Maceió, a low-income urban centre in Brazil’s Northeast. To support equitable resource allocation by the National Textbook and Educational Material Programme (PNLD), we couple the model with Explainable-AI techniques—SHAP values and Granger causality—to uncover the drivers of enrolment dynamics at preschool, elementary, and high-school levels. Results show that infrastructure quality (adapted restrooms, multimedia equipment) is the main lever for preschool access, teacher availability and libraries sustain elementary retention, while science labs, sports facilities, and ICT resources shape high-school demand. Intervariable analysis reveals that targeted budget cuts can trigger indirect declines in enrolment, underscoring the need for holistic investment.