Towards Optimization of Boosting Models for Formation Lithology Identification

Table 4

Performance matrix of 5-fold cross validation over the AdaBoost classifier, Gradient Tree Boosting classifier, XGBoost classifier, and stacked classifier in the Hangjinqi gas field.

	Precision	Recall	f1 score
AdaBoost
C	0.9	0.675	0.741
CS	0.836	0.81	0.821
FS	0.827	0.823	0.824
M	0.909	0.922	0.914
MS	0.858	0.817	0.835
PS	0.846	0.901	0.872
S	0.859	0.716	0.772
Avg	0.859	0.857	0.856
XGBoost
C	0.825	0.725	0.759
CS	0.822	0.792	0.806
FS	0.796	0.816	0.802
M	0.899	0.919	0.907
MS	0.853	0.761	0.803
PS	0.839	0.886	0.862
S	0.819	0.805	0.805
Avg	0.846	0.845	0.842
Gradient Tree Boosting
C	0.9	0.7	0.76
CS	0.832	0.793	0.812
FS	0.819	0.826	0.82
M	0.919	0.915	0.916
MS	0.854	0.823	0.836
PS	0.839	0.895	0.865
S	0.843	0.755	0.787
Avg	0.857	0.851	0.85
Stacking result
C	0.9	0.75	0.801
CS	0.856	0.795	0.823
FS	0.826	0.835	0.827
M	0.919	0.924	0.919
MS	0.863	0.831	0.847
PS	0.847	0.904	0.874
S	0.901	0.818	0.852
Avg	0.868	0.867	0.864