Table 3 AUC results of ActAttn with respect to different window size k and lead time τ
From: Activism via attention: interpretable spatiotemporal learning to forecast protest activities
|  | Charlottesville | Ferguson I | Ferguson II | ||||||
|---|---|---|---|---|---|---|---|---|---|
k = 1 | k = 2 | k = 3 | k = 1 | k = 2 | k = 3 | k = 1 | k = 2 | k = 3 | |
τ = 1 | 0.842 | 0.843 | 0.823 | 0.807 | 0.815 | 0.822 | 0.853 | 0.832 | 0.800 |
τ = 2 | 0.839 | 0.836 | 0.823 | 0.807 | 0.820 | 0.820 | 0.831 | 0.836 | 0.832 |
τ = 3 | 0.830 | 0.830 | 0.819 | 0.791 | 0.808 | 0.821 | 0.818 | 0.820 | 0.811 |