Estimating Village Development Index Based on Satellite Imagery Using Machine Learning Application

Abstract

BPS Statistics-Indonesia measures the development level in rural areas using the village development index obtained from village potential data collection. The problem with the current method carried out by BPS Statistics-Indonesia is that it requires a large number of funds and long interval data collection. Machine learning techniques combined with satellite imagery are expected to overcome the problem of collecting data. This study applies transfer learning techniques by classifying the nighttime light intensity from satellite imagery with the highest accuracy result of 0.6572 and predicting the village development index. This study produces a model with an R2 of 0.5. These results indicate that satellite imagery can be used as a predictor for the value of the village development index in an area.