Abstract: The disparity of regional economic development in China is reflected in the imbalance of tax gap to a certain extent. This paper chooses more than 〖BF〗300,000〖BFQ〗 industrial data from 31 provinces in China, and takes machine learning for large data analysis as the core algorithm to analyze tax differences between regions from the perspective of tax loss. The results show that under the premise of the same tax policy, the loss amount, loss rate, loss of households, loss of household ratio between provinces are significantly different. Therefore, tax authorities should take the opportunity of “Internet +” and big data as a chance to scientifically identify the difference of regional tax revenue loss, in order to promote regional tax collection and management balance, and ensure the sound development of economic tax.

Key words: regional economy, big data, machine learning, tax erosion