What influence farmers' relative poverty in China: A global analysis based on statistical and interpretable machine...

Area of Science:

• Socioeconomic studies
• Development economics
• Agricultural economics

Background:

• Poverty eradication is a global challenge, with relative poverty governance becoming critical for China post-2020.
• Addressing rural farmer poverty is crucial for China's moderate prosperity goals.
• Existing research often relies on a priori assumptions; innovative methods are needed.

Purpose of the Study:

• To identify and analyze factors influencing relative poverty in rural China.
• To develop and apply a novel machine learning framework for poverty governance.
• To provide a transparent and interpretable model for identifying at-risk populations.

Main Methods:

• Constructed a relative poverty index system tailored to China's context.
• Selected variables from individual characteristics, psychological endowment, and geographical environment.
• Applied machine learning algorithms, including XGBoost, for data analysis and factor identification.

Main Results:

• Machine learning, particularly XGBoost, demonstrated high efficacy in relative poverty research (81.9% accuracy, 0.819 ROC_AUC).
• Identified 25 key factors influencing relative poverty among farmers.
• Interpretable tools (PDP, SHAP) revealed the transparency and non-linear handling capabilities of machine learning models.

Conclusions:

• Machine learning offers a powerful and accurate approach to relative poverty analysis and governance.
• The study provides a robust framework and practical insights for targeted poverty alleviation strategies.
• Interpretable machine learning enhances understanding of complex socioeconomic relationships in poverty research.