A human-machine collaborative approach measures economic development using satellite imagery.

Area of Science:

• Remote Sensing
• Socioeconomic Analysis
• Machine Learning

Background:

• Satellite imagery offers accessible socioeconomic inference without physical site visits.
• Many machine learning algorithms require ground-truth data, which is often scarce or absent in many countries.
• Developing nations frequently lack comprehensive socioeconomic data, hindering development initiatives.

Purpose of the Study:

• To develop a human-machine collaborative model for predicting grid-level economic development.
• To overcome the limitations of ground-truth data scarcity in socioeconomic analysis.
• To provide fine-grained economic development predictions for data-poor regions.

Main Methods:

• Utilized publicly available satellite imagery.
• Employed lightweight subjective ranking annotation.
• Developed a human-machine collaborative model to predict economic development at a grid level.
• Applied the model to North Korea and five least developed Asian countries.

Main Results:

• Generated fine-grained economic development predictions for North Korea, a region with limited data.
• Identified substantial development in Pyongyang and areas with state-led projects.
• Demonstrated broad applicability across 400,000 grids in five Asian countries.
• Achieved high-resolution economic information in hard-to-visit and low-resource regions.

Conclusions:

• The human-machine model effectively predicts economic development without ground data.
• The approach yields granular economic insights crucial for understanding development in data-scarce regions.
• This method can significantly guide sustainable development programs in underserved areas.