Table 3 Performance metrics across different model variants.

From: Hybrid boosted attention-based LightGBM framework for enhanced credit risk assessment in digital finance

Model Variant

RMSE

MSE

MAPE

R2

Full HBA-LGBM

11.53

133.07

4.44%

0.998

Without Attention Mechanism

14.21

202.35

6.10%

0.987

Without Hybrid Boosting

13.75

189.50

5.85%

0.990

Without Cost-Sensitive Learning

14.03

195.80

5.95%

0.988

Without Synthetic Data Augmentation

13.90

192.40

5.88%

0.989

Back to article page