Abstract:
Stunting is a public health concern for children under five years. This issue affects the physical growth and cognitive capacity of children, which affects the child's potential development as growth progresses. Stunting is caused by a nutritional deficiency where, as needed by the health guidelines, the child does not get enough nutrients. Stunting or chronic malnutrition affects children under five years in Tanzania at a rate of 34.7 percent, which is still high.
The goal of this study is to find the most risk factors for stunting in Tanzania, as well as the best classifier for predicting stunting in children under the age of five.
Secondary data from the Tanzania 2015 Demographic and Health Survey, Children file was used in this study. The study included a total of 8289 children under the age of five. Five algorithms, Random forest, Decision tree, K-near-neighbor, Support vector machine and Logistic regression, were used in building the model. To obtain the best classifier evaluation metrics were used to get the performance
of each classifier, the metrics used are precision, recall, F1 score, accuracy and AUC.
The findings reveal that the best classifier which predicts the stunting status of children under five years was Random forest because it has performed better than the other classifier with the precision score of 89%, recall score of 84%, F1 score of 86% and accuracy of 83% and AUC score of 92%. The most risky factors of stunting were children from southern highlands, children born with mothers with primary education, male children, babies from poorest family, children between 36-47 month and children born with mothers between 15-24 years.
The study concludes that advanced technology is playing a major role in contributing to the development in health system. Machine learning as one of the technology tools is being used in health issues for predicting the diseases and identifying hidden pattern which couldn’t been revealed easily by other method. This research managed to improve a model that will be used to predict the stunting status of children under the age of five. This will aid in the reduction of child stunting.
