Abstract: Using interprovincial migration data from the sixth national census in China and other relevant natural and socioeconomic data, this paper first analyzes the spatial distribution characteristics of interprovincial migration in China, then investigates the network autocorrelation among the interprovincial migration flows with global Moran’s I, and establishes an eigenvector spatial filtering model which takes network autocorrelation into account to explore the influencing factors of the interprovincial migration. Specially, the regression results of the eigenvector spatial filtering model are compared with those of conventional gravity model, and thus the driving mechanism of China’s interprovincial migration is effectively revealed. The results show that: (1) Provinces with high immigration and net immigration are mainly concentrated in China’s three Economy Zone of the Pearl River Delta, Yangtze Delta and Beijing-Tianjin-Hebei regions. Provinces with high out-migration are mainly distributed in provinces of Anhui, Henan, Sichuan and Hunan. Provinces with high total migration are mainly located in the Pearl River Delta, Yangtze Delta, Beijing-Tianjin-Hebei and south central China regions. (2) There exists network autocorrelation phenomenon in China’s interprovincial migration behavior, of which immigration flows and out-migration flows are affected by the neighboring immigration or outmigration flows, thus network autocorrelation should be considered when modeling the driving mechanism of migration. Also, regression results indicate that the eigenvector spatial filtering model incorporating network autocorrelation factors has a better model fit than the gravity model, reveals successfully network autocorrelation effect among the interprovincial migration flows of China and reduces the biased estimation of non-network autocorrelation variables. (3) Among the non-network autocorrelation variables, total population, economy and distance are three important factors that influence the migration behavior.