Challenges and Prospects of Poverty Estimation Using Facebook: A Comparative Study Between Japan and the U.S. with Random Forest

Abstract

As inclusive business progresses, estimating the annual income status of users on SNS offers the possibility of appropriate intervention methods for people with low incomes. This paper analyzed people with low incomes in Japan and the United States using Facebook advertising data. It used Random Forest to examine the differences between regions with high and low poverty rates. The results showed that high interest in tobacco in the United States is associated with poverty rates. However, there is room for improvement in the model's accuracy and the difficulty of estimating income status from Facebook advertising data. Future challenges include increasing the types of data and further consideration of analytical methods.