Skip to main content

Table 4 Contribution of different components of our proposed method

From: Information cascades prediction with attention neural network

 MSE (t = 1*)MSE (t = 2*)MSE (t = 3*)
Twitter
 Proposed (no \(\varvec{gate}\))2.9562.8112.513
 Proposed (no \(\varvec{attention}\))3.2263.1242.825
 Proposed (no\(\varvec{attention+gate}\))3.7263.4193.376
 Proposed2.6092.3492.300
AMiner
 Proposed (no \(\varvec{gate}\))2.2851.7061.592
 Proposed (no \(\varvec{attention}\))2.3391.8161.763
 Proposed (no \(\varvec{attention+gate}\))2.4621.9281.809
 Proposed2.1721.6721.534
  1. p.s. t = 1*, where ‘*’ denotes hour for Twitter (year for AMiner)