Table 2 The results of different methods with different number of neurons in the hidden layer.
From: A novel SVD-UKFNN algorithm for predicting current efficiency of aluminum electrolysis
Model | Number of neurons in the hidden layer | Max | Min | Mean | MAE | MSE | SSE | \(r^2\) |
|---|---|---|---|---|---|---|---|---|
BPNN | 6 | 4.249 | − 3.143 | 0.099 | 0.477 | 0.544 | 819.224 | 0.97185 |
9 | 3.500 | − 2.888 | − 0.160 | 0.452 | 0.508 | 765.076 | 0.97371 | |
12 | 3.216 | − 2.719 | − 0.125 | 0.402 | 0.441 | 663.037 | 0.97722 | |
SVD-UKFNN | 6 | 3.438 | − 2.335 | 0.012 | 0.057 | 0.028 | 42.643 | 0.99853 |
9 | 2.931 | − 1.356 | 0.014 | 0.052 | 0.020 | 29.860 | 0.99897 | |
12 | 2.472 | − 1.927 | 0.015 | 0.054 | 0.019 | 28.850 | 0.99901 | |
NSVD-UKFNN | 6 | 0.207 | − 0.718 | 0.026 | 0.027 | 0.001 | 1.955 | 0.99993 |
9 | 0.275 | − 0.419 | 0.024 | 0.025 | 0.001 | 1.343 | 0.99995 | |
12 | 0.343 | − 0.331 | 0.021 | 0.022 | 0.001 | 1.140 | 0.99996 |
