Uncovering urban mobility patterns is crucial for further predicting and controlling spatially embedded events. In this article, we analyze millions of geographical check-ins crawled from a Chinese leading location-based social networking service, Jiepang.com, which contains demographical information and thus allows the group-specific studies. We found distinguishable mobility patterns of natives and non-natives in all five large cities under consideration, and by assigning different algorithms onto natives and non-natives, the accuracy of location prediction can be largely improved compared with pure algorithms. We further propose the so-called indigenization coefficients to quantify to which extent an individual behaves like a native, which depend only on check-in behaviors, instead of any demographical information. To our surprise, a hybrid algorithm weighted by the indigenization coefficients outperforms the mixed algorithm accounting for additional demographical information.