vincentvdp/repetition-risk

Predictive modeling in Python to predict the repetition and dropout risk of Rwandan primary school pupils. ## Predicting Repetition and Dropout Risk The Analysis conducted here is part of a SYPA -- Second Year Policy Analysis -- in fullfilment of the MPAID prgoram at the Harvard Kennedy School of Goverment. March 2017 ## Authors: Ben Brockman and Vincent Vanderputten ## Analysis summary: In our thesis named "Prediction and Prevention: Reducing Repetition & Dropout of Primary Students in Rwanda", we use machine learning and predictive modeling in Python to predict the repetition and dropout risk for individual pupils in Rwandan primary schools. Distilling a simplified scorecard from this model allows school and community based scorecard creation. Such a scorecard could (1) help teachers identify at risk kids and take discretionary action, or (2) improve targeting of remedial programs. This analysis makes use of publicly available Rwandan Integrated Household Data (statistics.gov.rw). The output is an easy-to-use proof-of-concept scorecard to asses Rwandan primary school pupils' risk of repetition. ## File overview: - SYPA_PolicyBrief_Final.pdf ==> Contains a brief overview of the whole SYPA' report - SYPA_Full_FinalSubmission.pdf ==> This is the full SYPA, which includes a draft scorecard as well as more in depth description of the quantitative analysis. - Data folder ==> Has both raw (survey) and processed data. - Stata ==> Contains the Stata code for cleaning. - Python ==> Contains the actual final analysis (e.g. logistics regression with LASSO) and performance measures.