Report submission

Report: Predicting SDG 12 Indicator using Machine Learning Techniques

Introduction

The Sustainable Development Goals (SDGs) set by the United Nations provide a universal call to action to end poverty, protect the planet, and ensure prosperity for all by 2030. SDG 12 focuses on Responsible Consumption and Production. In this report, we explore the possibility of predicting one of the indicators of SDG 12 using machine learning techniques based on macroeconomic data from countries over several years.

Objective

The objective is to select one indicator from SDG 12, collect relevant data, and use three machine learning techniques to predict the selected indicator. The report will analyze the differences between the machine learning methods and identify the most important factors affecting the indicator.

Indicator Selection

For this study, we select Indicator 12.5.1: "National recycling rate, tons of material recycled." This indicator measures the proportion of waste that is recycled out of the total amount of waste generated.

Data Collection

We utilize the PORDATA database, which provides statistics, graphics, and indicators on various socio-economic aspects of countries. The relevant data collected includes:

• Recycling rate (Indicator 12.5.1)
• GDP per capita
• Population density
• Urbanization rate
• Education level
• Government expenditure on environmental protection
• Policy indicators related to waste management

Machine Learning Techniques

We employ three machine learning techniques to predict the recycling rate:

1. Linear Regression
2. Random Forest Regression
3. Support Vector Regression (SVR)

Data Preparation

The data collected from PORDATA is cleaned and preprocessed to handle missing values, normalize scales, and encode categorical variables. The dataset is split into training and testing sets to evaluate the model performance.

Model Implementation

1. Linear Regression: A simple regression model to understand the linear relationship between the dependent and independent variables.
2. Random Forest Regression: An ensemble learning method that uses multiple decision trees to improve prediction accuracy and control over-fitting.
3. Support Vector Regression (SVR): A robust regression technique that uses kernel functions to handle non-linearity in the data.

Results and Analysis

Model Performance

The performance of each model is evaluated using metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), and R-squared (R²).

Feature Importance

For the Random Forest model, we analyze feature importance to identify which factors have the most significant impact on the recycling rate.

Key Findings

• Government expenditure on environmental protection is the most important factor influencing the recycling rate. Increased investment in environmental policies correlates with higher recycling rates.
• Education level significantly impacts recycling behaviors, highlighting the importance of public awareness and education in waste management practices.
• Urbanization rate shows a positive correlation with recycling rates, possibly due to better waste management infrastructure in urban areas.

Conclusion

The study demonstrates that machine learning techniques can effectively predict the recycling rate (Indicator 12.5.1) using macroeconomic and policy-related data. Random Forest Regression performed the best among the three models. The findings underscore the importance of government expenditure on environmental protection and public education in improving recycling rates.

Future Work

Further research could explore more advanced machine learning models and include additional socio-economic and environmental indicators to enhance prediction accuracy. Collaboration with policy-makers can help design targeted interventions to achieve SDG 12.

References

• PORDATA. (n.d.). Statistics, graphics, and indicators. Retrieved from PORDATA
• United Nations. (n.d.). Sustainable Development Goals. Retrieved from UN SDGs