油田生产数据选取22.4.1 观察实验

井类型相同，井数：35，计量数：18938

提取连续未选用属性效果：

fold	train loss	valid loss	AccuracyF	precisionFn	recallFn	f1-score
fold 1	0.026004	0.090267	97.19	90.47	89.52	89.99
fold 2	0.092705	0.094515	96.89	91.59	89.88	90.73
fold 3	0.083233	0.199184	92.44	86.03	88.63	87.31
fold 4	0.092850	0.104990	97.06	86.66	91.09	88.82

损失：

fold 1	fold 2
fold 3	fold 4

ROC曲线：
横轴 fpr
纵轴 tpr

fold 1	fold 2
fold 3	fold 4

同时显示选用与未选用
蓝色：未选用 1
红色：选用 0

fold 1	fold 2
fold 3	fold 4

fold	threshold	fpr	tpr
fold 1	0.8936	0.0449	0.9515
fold 2	0.7199	0.0337	0.9359
fold 3	0.7733	0.0697	0.8944
fold 4	0.6720	0.0511	0.9515

********** fold 1 **********
train_loss:0.026004 train_acc:99.4297
valid loss:0.090267 valid_acc:97.1905
AccuracyF 97.19 posLabel 13.96
precisionF 90.47 recallF 89.52 f1score 89.99
********** fold 2 **********
train_loss:0.092705 train_acc:96.5427
valid loss:0.094515 valid_acc:96.8948
AccuracyF 96.89 posLabel 16.58
precisionF 91.59 recallF 89.88 f1score 90.73
********** fold 3 **********
train_loss:0.083233 train_acc:97.3595
valid loss:0.199184 valid_acc:92.4377
AccuracyF 92.44 posLabel 30.25
precisionF 86.03 recallF 88.63 f1score 87.31
********** fold 4 **********
train_loss:0.092850 train_acc:96.8948
valid loss:0.104990 valid_acc:97.0638
AccuracyF 97.06 posLabel 13.46
precisionF 86.66 recallF 91.09 f1score 88.82

---

未提取连续未选用属性 效果：

fold	train loss	valid loss	AccuracyF	precisionFn	recallFn	f1-score
fold 1	0.309661	0.292365	90.77	75.84	50.75	60.81
fold 2	0.299104	0.323459	87.30	69.10	45.00	54.50
fold 3	0.250652	0.495778	76.74	83.68	25.83	39.47
fold 4	0.307507	0.401963	87.85	52.72	49.50	51.06

损失：

fold 1	fold 2
fold 3	fold 4

ROC曲线：

fold 1	fold 2
fold 3	fold 4

---

使用最佳？的threshhold
0.91088164 0.023952095808383235 0.9463846532218396
AccuracyF 96.39 posLabel 13.96
precisionF 87.59 recallF 86.68 f1score 87.13
AccuracyF 94.32 posLabel 9.91
precisionF 92.54 recallF 64.97 f1score 76.34
********** fold 1 **********
train_loss:0.038845 train_acc:98.5284
valid loss:0.106159 valid_acc:96.3878
0.6162127 0.06125 0.9522114895780376
AccuracyF 95.46 posLabel 17.98
precisionF 84.37 recallF 89.75 f1score 86.98
AccuracyF 94.59 posLabel 15.08
precisionF 88.10 recallF 78.62 f1score 83.09
********** fold 2 **********
train_loss:0.132937 train_acc:94.8880
valid loss:0.137882 valid_acc:95.4584
0.7286569 0.08633093525179857 0.8971291866028708
AccuracyF 90.14 posLabel 27.69
precisionF 85.20 recallF 80.36 f1score 82.71
AccuracyF 85.07 posLabel 17.81
precisionF 90.51 recallF 54.89 f1score 68.34
********** fold 3 **********
train_loss:0.111692 train_acc:96.0991
valid loss:0.248463 valid_acc:90.1352
0.6494336 0.066006600660066 0.9464631782945736
AccuracyF 95.31 posLabel 14.07
precisionF 78.83 recallF 86.63 f1score 82.55
AccuracyF 94.57 posLabel 10.20
precisionF 86.13 recallF 68.65 f1score 76.40
********** fold 4 **********
train_loss:0.141379 train_acc:94.7261
valid loss:0.147124 valid_acc:95.3105