IPTV dataset

https://www.scholat.com/teamworkdownloadscholar.html?id=5885&teamId=612 

Note that, you must register on https://www.scholat.com, and join our team "new_media", finally download the dataset. Pretecting the user pravicy.  

The dataset is from a major IPTV service provider in southern China. Similar to the dataset used for research [1][2], not containing the detailed time when a viewer switched into a channel. This is the first to publish a complete dataset with the time granularity of second. There are totally 221k devices and 164 channels in the original dataset. Each entry consists of four fields. "DeviceID" refers to a subscriber, i.e. a user or a family;"ChannelNumber" represents the number of watching channel; "StartTime"is when the user starts to watch the channel; and "Duration" is how long the user stays in the channel. For instance, device {AC3B81E6-E531-4B03-8044-7649817F8836} switched into channel 19 at 2014-08-31 23:59:32.603000 and stayed for nearly 47120ms. 

[1]“Can Yang, Sixuan Ren, Yong Liu, Houwei Cao, Qihu Yuan, and Guoqiang Han. 2018. Personalized channel recommendation deep learning from a switch sequence. IEEE Access 6 (2018), 50824– 50838.”

[2]“Chenguang Yu, Hao Ding, Houwei Cao, Yong Liu, and Can Yang. 2017. Follow me: Personalized IPTV channel switching guide. In Proceedings of the 8th ACM on Multimedia Systems Conference. ACM, 147–157”