Meta-IP: An Imbalanced Processing Model Based on Meta-Learning for IT Project Extension Forecasts

Table 6

The G-mean values were achieved with forecast models on balanced and imbalanced training datasets (2^nd-year) by various sampling methods. Bold figures reflect the row maximum.

Forecast model	Imbalanced	Naive bayesian	Bagging	SMOTE	SVM
LDA	39.1	72.3	83.9	75.4	82.1
LR	40.9	70.5	83.6	77.6	83.4
NN	42.3	70.9	84.4	77.8	81.9
SVM	41.3	73.3	85.9	76.5	82.5
RF	42.7	72.8	85.6	75.2	83.2