Casing Damage Prediction Model Based on the Data-Driven Method

Table 3

Model application test prediction results.

Project	Algorithm	Training set		Test set		Number of factors	Parameters removed
		Damaged	Undamaged	Damaged	Undamaged
Actual	—	56	139	12	37	—	—
test1	XGBoost	56	139	11	35	23	—
	LightGBM	53	139	10	36
test2	XGBoost	54	138	9	34	21	Casing service time; Sand layer production time
	LightGBM	32	138	7	36
test3	XGBoost	56	139	11	35	21	Perforation bottom; Sand layer bottom
	LightGBM	53	139	10	36
test4	XGBoost	56	139	10	35	20	Perforation bottom; Sand layer bottom; Perforator phasing
	LightGBM	53	139	10	36
test5	XGBoost	56	139	10	35	20	Perforation bottom; Sand layer bottom; Qlmax
	LightGBM	53	139	10	36
test6	XGBoost	56	139	11	34	20	Perforation bottom; Sand layer bottom; Max water cut
	LightGBM	49	139	10	36
test7	XGBoost	56	139	11	35	20	Perforation bottom; Sand layer bottom; Max fluid production intensity
	LightGBM	54	139	10	36
test8	XGBoost	56	139	11	35	20	Perforation bottom; Sand layer bottom; Qlcum
	LightGBM	54	139	10	36
test9	XGBoost	55	139	10	35	20	Perforation bottom; Sand layer bottom; Qlmax/KH
	LightGBM	54	139	10	36
test10	XGBoost	56	139	11	35	20	Perforation bottom; Sand layer bottom; Qlmax/HN
	LightGBM	54	139	10	35
test11	XGBoost	54	138	10	33	20	Perforation bottom; Sand layer bottom; Casing service time
	LightGBM	47	137	9	35
test12	XGBoost	56	139	10	35	20	Perforation bottom; Sand layer bottom; Sand layer production time
	LightGBM	54	139	10	36
test13	XGBoost	55	139	10	35	19	Perforation bottom; Sand layer bottom; Max fluid production intensity; Qlcum
	LightGBM	54	139	10	35