CREATE PROXY MEAN TEST BY USING MACHINE LEARNING TECHNIQUE

Abstract

This study aims to develop a model for PMT using Machine Learning techniques. It starts by reviewing the weaknesses of the traditional model for measuring PMT, then develops a model for measuring PMT using Machine Learning techniques to find the most efficient model. The study also analyzes the impact of each variable on the model, which the traditional Machine Learning model cannot explain, using the SHAP value. The results of the study on the efficiency of the model in predicting whether a household is poor or not, show that the model accurately predicts 91% of the households identified as poor, while the model only covers 50% of the total poor households, and the overall model accuracy is 65% (F1-score). These results are significantly more efficient than the traditional methods, and policy recommendations can be made based on the top ten variables with the highest impact on the prediction.