The dataset is provided by Global Technology Services department of Huawei Technologies, Co., Ltd. It represents the relationship between users and behaviors in Telecom network. We believe that this telecom graph is a good supplement to the online graph collections, which can help establish the evaluation standard of modern graph algorithms on Telecom networks. As graphs become more and more popular in Carriers nowadays, the graph benchmark of Telecom field is rather valued by these top companies. --- The telecom graph consists of 4 types of nodes (user, package, app, cell) and 3 types of edges (user-buy-package, user-use-app, user-live-cell). There are 170K nodes and 50M edges, among which 100K are users. All files are plain texts in CSV format, with the file name beginning with 'edge_' or 'node_'. For node files, each row is: nodeID,attr1,attr2,... For edge files, each row is: srcID,dstID,attr1,attr2,... All IDs are 4-byte integers, while attributes are all 4-byte float numbers within [0.f,1.f].