What do filipino children eat: Applying market basket analysis on the meals of school-aged children in the Philippines

• Main Authors: Castillon, Enzo Gabriel, Flores, Richell Isaiah, Lim, Jacob, Miro, Eden Delight, Go, Clark Kendrick
• Format: Conference Proceeding
• Language: English
• Subjects: Children; Children & youth; Data mining; Diet; Food; Machine learning; Malnutrition; Meals; Natural language processing; Nutrition; Plates; Rural areas; Unsupervised learning; Urban areas
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0213578

Nutrition is vital to the development of a child and is characterized primarily by what they eat. In the Philippines, however, undernutrition is more severe and prevalent in rural than in urban areas. Previous studies have primarily focused on predicting malnutrition among schoolchildren using various features such as household income and other quantitative data, but there is a gap to be filled in analyzing the exact food items that the children eat for specific meal denominations. In this work, we present a summary of the schoolchild’s plate in an urban and rural setting in the Philippines while using unsupervised learning methods to analyze relationships among food items. Through natural language processing (NLP) of a 24-hour food recall survey, common food items were collated and summarized. Market basket analysis (MBA) through association rules was then performed to identify food items that constitute schoolchildren’s meals. Through a frequency analysis of food items, rice was empirically shown to be part of all the meals (breakfast, lunch, snack, dinner) of a schoolchild for both areas. Moreover, the rural area’s diet is revealed to have a mix of fish and vegetable dishes, while the urban area’s diet consists mainly of processed and fried foods. Association rule mining among food itemsets was also used to predict and give insights on the nutritional makeup of the child given the acquisition of prior food itemsets. Similar food combinations were seen from both frequency-based plates and association rules-based plates. Hence, through NLP and MBA techniques, analysis of nutrition could be brought to a meal-level recommendation. Thereby, through this study, a possibility of an individualized, tailor-fitted response to malnutrition is presented, complementing established machine learning prediction methods towards a more involved intervention for a child’s nutritional development.