Poverty classification: A comparative analysis of classification algorithms on poverty in households in the top three richest and poorest regions in the Philippines using the family income and expenditure survey 2021

• Main Authors: Dy, Jonathan Arthur L., Butardo, Chrisha Mae Tan, Hernandez, Aaron Anthony Munoz
• 格式: text
• 語言: English
• 出版: Animo Repository 2023
• 主題: Poverty > Philippines; Algorithms; Statistics and Probability
• 在線閱讀: https://animorepository.dlsu.edu.ph/etdb_math/28; https://animorepository.dlsu.edu.ph/context/etdb_math/article/1030/viewcontent/2023_Butardo_Dy_Hernandez_Poverty_Classification__A_Comparative_Analysis_Full_text.pdf

Poverty remains one of the most significant issues the Philippines faces today. Despite the country’s poverty rate slowly decreasing over the years, the COVID-19 pandemic caused the situation to worsen once again. This study aimed to propose an alternative classification for poverty by using machine learning and k-fold cross-validation among the decision tree algorithm, logistic regression, and Naïve Bayes classifier to get a better representation of the poverty-stricken households in the Philippines. The criteria used to determine the best classification algorithm will be accuracy, specificity, recall, and F1 score. This study found that the algorithm with the highest sensitivity was the Naïve Bayes classifier, while the algorithm with the highest specificity was the decision tree algorithm. However, the logistic regression algorithm was deemed the “best” among the three since it is able to determine both poverty and non-poverty households due to it having the most balanced results across all four criteria.