naeljb/uthabiti

Storage of Machine Learning applied to a USAID/BHA funded program in Uganda and implemented by Save the Children 1. Project Description: Provide a brief summary of your project and what it does. This will help readers quickly understand what your project is all about. In December 2022, Save the children Uganda conducted a joint beneficiary registration and baseline of a BHA's funded food security project, Uthabiti. Unsupervised Machine Learning algorithms were applied to complement the baseline report for an in-depth understanding of the household selected, particularly the refugees. The primary purpose was to do a segmentation analysis to find similar groups of refugee households based on key food security indicators and demographic characteristics, as well as to identify if there were refugee households significantly different from the rest of the data. Three scenarios of variable combination were constructed for conducting this analysis. Two BHA's food security indicators (reduced Coping Strategy Index and Food Consumption Score) listed in Uthabiti's Indicator Performance tracking table (IPTT) were considered in the first scenario. In the second scenario, Food Expenditure Share (as a proxy of income) was computed and added to the variables considered under the first scenario. The third scenario included the household size and the BHA's gendered household classification in the second scenario. After conducting a series of data cleaning, shaping, and feature engineering tasks, an Exploratory Data Analysis (EDA) was performed. Then, the two most popular clustering algorithms (K-Means and DBSCAN) using the optimum number of clusters/groups detected by each algorithm were run for each scenario. Finally, silhouette and calinski harabasz score were computed and used to compare the quality of the clusters obtained under each scenario and to suggest the model that best fits the primary objective of this analysis. Python programing language was used to conduct the EDA and run clustering algorithms. 2. Installation instructions: Project is done in Python using Jupyter Noteboo …