PriscillaKungu/Ngao-Labs-Cohort-3-Capstone-Project-Predicting-Household-Food-Security

Using Wave 4 of the Kenya COVID-19 Rapid Response Phone Survey (World Bank / KNBS / UC Berkeley, ~4,892 households, collected January–March 2021), predict a household's food security status based on household demographics, employment status, income loss during the pandemic, and access to transfers for Kenyan households by macro-averaged F1 scores. # Ngao-Labs-Cohort-3-Capstone-Project-Predicting-Household-Food-Security **Problem Statement** Using Wave 4 of the Kenya COVID-19 Rapid Response Phone Survey (World Bank / KNBS / UC Berkeley, ~4,892 households, collected January–March 2021), predict a household's food security status (food secure / food insecure, or a 3–4 level severity category derived from the survey's food security module) based on household demographics, employment status, income loss during the pandemic, and access to transfers for Kenyan households nationally, both urban and rural measured by macro-averaged F1 score, disaggregated by urban/rural location and by county where sample size allows. **Four Required Elements** What is being predicted: A household-level food security category. The exact number of classes (binary secure/insecure, or a multi-level severity scale) will be finalized once the Wave 4 food security module questions are reviewed in the data dictionary likely built from a small set of experience-based questions (e.g., worry about food, skipped meals, ran out of food). For which population: Kenyan households nationally, drawn from a sample representative of both the 2015/16 KIHBS pilot phone-owning households and Kenya's broader mobile-phone-using population, covering urban and rural areas across counties. Using what data: Kenya COVID-19 Rapid Response Phone Survey, Wave 4, hosted on the FAO Microdata Catalogue (microdata.fao.org), household-level file (hhid as key), approximately 4,892 rows. Target and predictor columns will be confirmed against the accompanying data dictionary after registration and download. Measured how: Macro-averaged F1 score, because the food security classes are likely to be imbalanced (most households not in the most severe category), and macro F1 weights performance on minority (most vulnerable) classes fairly rather than letting the majority class dominate the metric. **Real-World Context** A county government …