Research Article
Full Data-Processing Power Load Forecasting Based on Vertical Federated Learning
Table 2
Comparison of
and runtime between centralized learning and three-party FL on dataset-3.
| | | 1 | 2 | 3 | 4 | 5 | 6 | Average runtime |
| | Centralized | 0.987 | 0.947 | 0.886 | 0.803 | 0.665 | 0.540 | — | Two-party | 0.988 | 0.953 | 0.933 | 0.922 | 0.874 | 0.843 | 66 m | Three-party | 0.992 | 0.975 | 0.957 | 0.938 | 0.921 | 0.911 | 79 m |
| | Centralized | 0.992 | 0.974 | 0.951 | 0.933 | 0.918 | 0.887 | — | Two-party | 0.992 | 0.977 | 0.956 | 0.943 | 0.920 | 0.906 | 84 m | Three-party | 0.994 | 0.981 | 0.968 | 0.952 | 0.938 | 0.926 | 98 m |
| | Centralized | 0.993 | 0.980 | 0.970 | 0.952 | 0.921 | 0.895 | — | Two-party | 0.993 | 0.985 | 0.976 | 0.964 | 0.940 | 0.924 | 108 m | Three-party | 0.994 | 0.984 | 0.973 | 0.963 | 0.952 | 0.944 | 122 m |
|
|