Table 2 The effects of the discriminator and random noise \(\eta\) on the proposed method and \({E}^{2}\) GAN
From: TLGRU: time and location gated recurrent unit for multivariate time series imputation
Method/Missing | 10% | 20% | 30% | 40% | 50% | 60% | 70% | 80% |
|---|---|---|---|---|---|---|---|---|
Ours | 0.336 | 0.510 | 0.578 | 0.637 | 0.697 | 0.702 | 0.741 | 0.747 |
\({E}^{2}\) GAN | 0.334 | 0.517 | 0.584 | 0.641 | 0.694 | 0.705 | 0.745 | 0.756 |
Ours-no-D | 0.354 | 0.514 | 0.606 | 0.665 | 0.727 | 0.729 | 0.783 | 0.789 |
\({E}^{2}\) GAN-no-D | 0.357 | 0.519 | 0.608 | 0.671 | 0.735 | 0.742 | 0.784 | 0.788 |
\(\mathrm{Ours}\)-no-noise | 0.347 | 0.517 | 0.611 | 0.655 | 0.723 | 0.730 | 0.768 | 0.782 |
\({E}^{2}\) GAN-no-noise | 0.353 | 0.518 | 0.614 | 0.667 | 0.736 | 0.729 | 0.779 | 0.791 |