Skip to main content

Table 8 Performance for training sets going back in time

From: Large-scale online semantic indexing of biomedical articles via an ensemble of multi-label classification models

Size Date Micro-F Macro-F
100,000 December 2012- July 2013 0.5591 0.3616
250,000 January 2012- July 2013 0.5827 0.4567
500,000 August 2010- July 2013 0.5941 0.5130
750,000 January 2009- July 2013 0.5977 0.5358
1,000,000 August 2007- July 2013 0.5993 0.5480
1,500,000 July 2004- July 2013 0.5995 0.5637
2,000,000 August 2001- July 2013 0.5963 0.5652
4,300,000 December 1946 - July 2013 0.58646 0.56014
  1. A fixed test set of 50k abstracts is employed for the experiment, from July 2013 to January 2014