Skip to content

Latest commit

 

History

History
30 lines (22 loc) · 1.58 KB

README.md

File metadata and controls

30 lines (22 loc) · 1.58 KB

Dataset + Code on YouTube popularity history

We release the sofware and dataset from the following paper,

Honglin Yu, Lexing Xie and Scott Sanner, The Lifecycle of a Youtube Video: Phases, Content and Popularity, (ICWSM-15) [ paper | slides | poster]

Sofware

  1. Algorithm for power-law phase segmentation
  2. YouTube history crawler

Data

The data is hosted on dropbox.

File Description

All files are in python's pickle format.

videoID_category.pickle

  1. Data type : dictionary
  2. Key : videoIDs
  3. Value : category got from google API

videoID_segInfo.pickle

  1. Data type : dictionary
  2. Key : videoIDs
  3. Value : list of description of segments, following chronological order. For example, the value of video XXTey9OjuGc is [(0, 4, -0.373851, 2.98491, 48.4606, 1), (5, 734, 98.0916, -0.955789, 7.4257, 0)], this means the viewcount contains two segments. To see the parameter of each segment, please see this readme file on github.

videoID_viewcount.pickle

  1. Data type : dictionary
  2. Key : videoIDs
  3. Value : viewcount from first day of uploading. (containing 735 days)