Analysis of Child Malnutrition and Possible Related Factors

Project Members: Noah Lin, Annie Liu, Amy Nguyen

Inspiration

In a recent report by UNICEF on child malnutrition, it was found that although malnutrition levels are decreasing in recent years, malnutrition is still responsible for about half of the deaths in young children globally. Our motivation in researching this topic was to gain a better understanding of child malnutrition and its factors as a global issue as well as what countries may potentially be at more risk for it.

What it does

• Explores the possible factors that contribute to stunting.
• Generates statistical models to see whether cost features contribute to child malnutrition.
• Determines the accuracy of ML models in predicting child malnutrition based on year, country, and poverty levels
• Outputs a GIF of poverty and child malnutrition levels (determined by stunting) over time for qualitative analysis of time-based factors contributing to child malnutrition.

How we built it

The program consists of three subroutines:

1. Main Program:
   • Contains functions used to answer the three research questions
   • Loads data using functions from the loading and cleaning program
   • Plots RQ1 visualizations
   • Trains, tests, and measures ML models for RQ2
   • Generates animation for RQ3
2. Loading and Cleaning Program:
   • Contains functions that load and clean specific datasets used for each research question individually
   • Merges some datasets with shape files for mapping purposes using GeoPandas
3. Testing program
   • Tests specific functions used from loading and merging data
   • Tests visualization capabilities of the main functions

Challenges we ran into

• The data cleaning took longer than expected due to the different country name formats across datasets.
• GIFs are difficult to generate in the Ed workspace, had to develop half of RQ3 code locally
• Difficulty downloading geopandas locally
• Implementing a slider for the visualizations

Accomplishments that we're proud of

• Finishing the project on time
• Successfully creating a GIF

What we learned

• There is no significant relationship between the cost features in the affordability of diets dataset and stunting.
• LinearRegression model performed better than the DecisionTreeRegressor
• Over time, increased poverty levels seemed to spread geographically across neighboring countries

What's next for Analysis of Child Malnutrition and Possible Related Factors

• Research Question 1:
  • Further research on other features/factors that contribute to child malnutrition
• Research Question 2:
  • Find more datasets on poverty metrics
  • Evaluate more types of machine learning models
• Research Question 3:
  • Bolster malnutrition data by merging with other datasets
  • Explore quantitative measures on the correlation between poverty and malnutrition

Built With

• python