The Rising of Food Insecurity: A Predictive Modeling for Production, Area Harvested and Yield Outcome

1. STEP 01

   Large-scale dataset

   Worked with the Food Bank of the World dataset: 1,895,975 rows, 150+ crops and 247+ localities.

2. STEP 02

   Twelve models

   Trained four algorithms (Random Forest, XGBoost, Linear Regression and an artificial neural network) against each of three targets: Production, Area Harvested and Yield.

3. STEP 03

   Explainability

   Used feature importance as an explainable-AI (XAI) layer, alongside 3D visualisation of the results.

4. STEP 04

   Delivery

   Shipped the models in a Streamlit app, developed in GitHub Codespaces.

• 0.9998

  R², Production

  Random Forest

• 0.9987

  R², Area Harvested

  Random Forest

• 0.9895

  R², Yield

  Random Forest

• 12

  Models trained

  4 algorithms × 3 targets

• Random Forest gave the reported R² for all three targets: 0.9998 for Production, 0.9987 for Area Harvested and 0.9895 for Yield.
• The project includes feature-importance XAI, 3D visualisation and a Streamlit app.

• Python
• scikit-learn
• XGBoost
• TensorFlow/Keras
• Streamlit
• GitHub Codespaces

Group project for DSC761 Advanced Data Science Technology, Master of Data Science, UiTM.

• Muhammad Aiman Akmal bin Pardan
• Syaza Amani binti Yacob
• Nur Amynah binti Abdul 'Adzim Shukri
• Nur Fathihah binti Salmizi
• Nabil TakiyuddinThis site