1 of 11

Leveraging Artificial Intelligence for Climate Change Integration in Evaluation Frameworks��Anowai Chineme

2 of 11

Outline

Introduction
Climate and Made in African Evaluation
Case study: AI in climate-centered evaluation
Afridat Algorithm 1.0
Conclusion

3 of 11

Climate Change and the impact in Africa

At COP25, it was highlighted that Africa is the continent most vulnerable to climate change impacts, particularly under scenarios where global temperatures rise above 1.5 degrees Celsius. Despite its minimal contributions to global warming and low emissions.

Rouelle Umali/Xinhua via Getty Images

Despite having contributed the least to global warming and having the lowest emissions, Africa faces exponential collateral damage, posing systemic risks to its economies, infrastructure investments, water and food systems, public health, agriculture, and livelihoods, threatening to undo its modest development gains and slip into higher levels of extreme poverty. The following factors contribute to Africa’s vulnerability: “WFP declared the floods in Mozambique a Level-3 emergency, on a par with Yemen and Syria” Ninety percent of the infrastructure was destroyed in Beira, Mozambique, the epicenter of the 2019 Cyclone Idai. Beira is Mozambique’s fourth-largest city, with important energy infrastructure investments and natural gas deposits off the coastline. • Sub-Saharan Africa has 95% of rain-fed agriculture globally. • A large share of agriculture in GDP and employment adds to vulnerability, as do other weather-sensitive activities, such as herding and fishing, leading to income losses and increased food insecurity.

4 of 11

Case study; Machine Learning in Evaluation- Education

5 of 11

Machine learning in Evaluation for Education

Supervised learning
Exploratory Data Analysis and Model
EDA Findings:

Correlations between climate variables and education outcomes.
Visualization of key relationships (e.g., scatter plots, heatmaps).

Machine Learning Model:

Model Used: Random Forest Regressor.
Training Process: Splitting data into training and test sets, training the model on training data.

6 of 11

Data Collection and Preprocessing

Data Sources:

Education: Literacy rates, school enrollment, educational attainment (e.g., World Bank, UNESCO but preferably local sources).

Climate: Temperature, precipitation, climate impact indicators (e.g., NOAA, CCD).

Preprocessing Steps:

Data Cleaning: Handling missing values.

Normalization/Scaling: Standardizing features.

Merging Datasets: Aligning data by region and time period.(real time)

data = pd.merge(education_data, climate_data, on=['country/state', 'year/month'])

7 of 11

Algorithm and Model Training

Model: Random Forest Regressor.

Feature Selection: Choosing relevant variables from education and climate data.
Model Training:

Data Split: Training (80%) and testing (20%).
Training Process: Fitting the model to training data.

Algorithm Workflow:

Input: Preprocessed data.
Output: Predicted impact of climate variables on education outcomes. Mean Absolute Error (MAE): {2.5} var(1-100) R-squared: {0.85}

8 of 11

Afridat 1.0

9 of 11

Findings

Schools can be designed or retrofitted to withstand extreme weather, with features like elevated structures in flood-prone areas and improved ventilation for heat.
Higher temperatures associated with lower school enrollment rates
Forecasting attendance drops during adverse weather to inform evaluation
AI offers valuable tools and capabilities to support the balance between innovative education development and environmental sustainability.

10 of 11

Machine learning for Climate-centered evaluation- Afridat 1.0

Automated Analysis:

Utilizes machine learning algorithms to replicate generic data analysis.

Automates evaluation processes, enhancing accuracy and efficiency and real time insights

Dynamic Evaluation:

Integrates into project evaluation(baseline, midline, endline and impact measurement.

Provides real-time insights and feedback on interventions.

Increased Accuracy:

Ensures precise and reliable evaluation outcomes.

Supports informed decision-making throughout the project lifecycle, across all sectors.

Localized insights (clime specific)

11 of 11

Conclusion

Getting Africa there, How fast can we harness the data we need?