Table 4.
Correlation of non normalized and root mean square error of normalized usage dominance
for highly popular tweets (popularity > 500). Highlighted entries represent the best two values in each column.
Model | Ground truth | High ![]() |
Low ![]() |
||||||
---|---|---|---|---|---|---|---|---|---|
corr | p-val | rmse | corr | p-val | rmse | corr | p-val | rmse | |
M2 | 0.79 | 0.01 | 33.66 | −0.5 | 0.17 | 89.51 | 0.76 | 0.02 | 18.31 |
M3 | 0.79 | 0.01 | 32.03 | −0.49 | 0.18 | 87.92 | 0.73 | 0.02 | 16.68 |
M4 | 0.8 | 0.01 | 31.39 | −0.5 | 0.17 | 87.32 | 0.78 | 0.01 | 16.04 |
M5 | 0.85 | <0.01 | 15.38 | −0.58 | 0.10 | 71.26 | 0.77 | 0.02 | 1.83 |
M6 | 0.72 | 0.03 | 28.7 | −0.35 | 0.36 | 84.73 | 0.65 | 0.06 | 13.54 |
M7 | 0.79 | 0.01 | 19.35 | −0.49 | 0.18 | 75.21 | 0.78 | 0.01 | 6.71 |
M8 | 0.75 | 0.02 | 30.14 | −0.44 | 0.23 | 86.07 | 0.69 | 0.04 | 14.83 |
M9 | 0.78 | 0.01 | 16.38 | −0.44 | 0.23 | 72.16 | 0.72 | 0.03 | 5.89 |