Abstract: Geological library has a large number of books and data are huge.It is difficult to solve that data grows too fast and it is difficult to find the reader’s point.To achieve efficient library borrowing data mining analysis and recommendation，is an important means to improve efficiency.For this reason，this paper puts forward a large-scale data mining platform，including clustering analysis，Chinese word segmentation，recommendation system，correlation analysis function，and then through hadoop cluster multi-node recommendation，thus improving the efficiency of the work.