Skip to main content
. 2023 Feb 3;14:1093905. doi: 10.3389/fpls.2023.1093905

Table 5.

Comparison of test results of different crown profile models based on LSTM and its variant algorithms.

Model R2 (Rank) MSE (Rank) MAE (Rank) ME (Rank) Sum Rank (Rank)
Vanila LSTM 0.6692 (5) 0.2942 (5) 0.3959 (5) 0.2028 (5) 20 (5)
LightGBM 0.7197 (2) 0.2493 (2) 0.3586 (2) 0.0607 (3) 9 (2)
Stacked LSTMs-LightGBM 0.7040 (3) 0.2633 (3) 0.3700 (3) 0.0379 (1) 10 (3)
Integrated LSTM-LightGBM 0.7308 (1) 0.2394 (1) 0.3537 (1) 0.0458 (2) 5 (1)
Hybrid LSTM-LightGBM 0.6900 (4) 0.2757 (4) 0.3780 (4) 0.0664 (4) 16 (4)

Rank is the rating of the test statistic (R2, ME, MAE and MSE), the smaller the rating value, the better the model prediction result. the best performance is highlighted in bold.