Table 2 Monthly mean of macro-averaged F1 scores for detection of Russian troll tweets, with varying predictor sets.

Monthly mean of macro-averaged F1 scores for detection of Russian troll tweets, with varying predictor sets.. User-timing features were removed for task 3. Task 5 is excluded because it is based on a reduced set of features and therefore not comparable to other tasks.

Only content(1) + meta-content(2) + content timing(3) + user timing(4) + network features
Model number(1)(2)(3)(4)(5)
Experiments
  Within-month train/
test (task 1)
0.760.810.820.850.84
  Train on t − 1 test on t
(task 2)
0.740.820.82NA0.85
  Train on t − 1 test on
new users in t (task 3)
0.660.750.750.810.82
  Within-month
cross-release (task 4)
0.660.700.700.740.75