Abstract

The global population of malnourished children has been declining for the past 30 years; however their are growing concerns that underdeveloped countries are being left behind. This research applies machine learning algorithms to a WHO and World Bank child malnutrition data set to predict the four types of malnutrition: stunting, wasting, underweight and overweight. The goal of this report is to compare previous studies to the machine learning algorithms in hopes to find new conclusions. In addition, the report investigates the merits and demerits of machine learning algorithms like classification and regression trees (CART), bagged CART and random forest. The comparative analysis provides insight on how decision trees are more advantageous for prediction compared to traditional methods like multiple linear regression. The family of recursively partitioned decision trees all behave similarly; however this research discovers the nuances between them. The classification and regression tree extension algorithms provide greater accuracy in prediction results, but lack the ability for interpretations. The machine learning algorithms constructed in this research supply quantifiable interpretations and visualizations that drive discussion for limiting the global population of malnourished children.

Advisor

Frazier, Marian

Department

Mathematics

Keywords

child malnutrition, machine learning, classification and regression trees, random forest, bootstrap aggregation, bagging, stunting, wasting, underweight, overweight

Publication Date

2019

Degree Granted

Bachelor of Arts

Document Type

Senior Independent Study Thesis

2019Sansom_Kory.Rmd (16 kB)
The R code associated with the report.

Malnutrition Estimates Historical.xlsx (26788 kB)
The malnutrition data set acquired from WHO and World Bank.

Share

COinS
 

© Copyright 2019 Kory Sansom