油田生产数据选取进展22.3.12 观察实验

未进行去噪处理
原数据按照时间排序

fold	train loss	valid loss	AccuracyF	precisionFn	recallFn	f1-score
fold 1	0.002866	0.035296	92.29	51.39	90.24	65.49
fold 2	0.002113	0.005586	98.62	100.00	91.46	95.54
fold 3	0.002704	0.006520	98.62	100.00	93.81	96.80
fold 4	0.002653	0.011558	96.05	98.12	90.23	94.01
fold 5	0.002135	0.005831	97.83	91.04	92.42	91.73

********** fold 1 **********
train_loss:0.002866 train_acc:99.4071
valid loss:0.035296 valid_acc:92.2925
AccuracyF 92.29 posLabel 14.23
precisionF 51.39 recallF 90.24 f1score 65.49
********** fold 2 **********
train_loss:0.002113 train_acc:99.6542
valid loss:0.005586 valid_acc:98.6166
AccuracyF 98.62 posLabel 14.82
precisionF 100.00 recallF 91.46 f1score 95.54
********** fold 3 **********
train_loss:0.002704 train_acc:99.4071
valid loss:0.006520 valid_acc:98.6166
AccuracyF 98.62 posLabel 20.95
precisionF 100.00 recallF 93.81 f1score 96.80
********** fold 4 **********
train_loss:0.002653 train_acc:99.3577
valid loss:0.011558 valid_acc:96.0474
AccuracyF 96.05 posLabel 31.62
precisionF 98.12 recallF 90.23 f1score 94.01
********** fold 5 **********
train_loss:0.002135 train_acc:99.7530
valid loss:0.005831 valid_acc:97.8261
AccuracyF 97.83 posLabel 13.24
precisionF 91.04 recallF 92.42 f1score 91.73

---

剔除f1-score变化最大的第一部分数据：

fold	train loss	valid loss	AccuracyF	precisionFn	recallFn	f1-score
fold 1	0.003632	0.036504	96.84	94.59	85.37	89.74
fold 2	0.002111	0.007875	97.63	98.10	91.15	94.50
fold 3	0.001736	0.007074	98.62	98.27	97.70	97.98
fold 4	0.002340	0.008007	97.83	92.31	90.91	91.60

********** fold 1 **********
train_loss:0.003632 train_acc:99.5389
valid loss:0.036504 valid_acc:96.8379
AccuracyF 96.84 posLabel 14.62
precisionF 94.59 recallF 85.37 f1score 89.74
********** fold 2 **********
train_loss:0.002111 train_acc:99.8024
valid loss:0.007875 valid_acc:97.6285
AccuracyF 97.63 posLabel 20.75
precisionF 98.10 recallF 91.15 f1score 94.50
********** fold 3 **********
train_loss:0.001736 train_acc:99.8682
valid loss:0.007074 valid_acc:98.6166
AccuracyF 98.62 posLabel 34.19
precisionF 98.27 recallF 97.70 f1score 97.98
********** fold 4 **********
train_loss:0.002340 train_acc:99.8682
valid loss:0.008007 valid_acc:97.8261
AccuracyF 97.83 posLabel 12.85
precisionF 92.31 recallF 90.91 f1score 91.60

---

fold	train loss	valid loss	AccuracyF	precisionFn	recallFn	f1-score
fold 1	0.004499	0.029143	94.94	83.33	89.29	86.21
fold 2	0.002107	0.009347	96.84	98.98	91.51	95.10
fold 3	0.003310	0.010244	97.79	93.55	91.58	92.55

********** fold 1 **********
train_loss:0.004499 train_acc:99.2101
valid loss:0.029143 valid_acc:94.9447
AccuracyF 94.94 posLabel 18.96
precisionF 83.33 recallF 89.29 f1score 86.21
********** fold 2 **********
train_loss:0.002107 train_acc:99.6840
valid loss:0.009347 valid_acc:96.8404
AccuracyF 96.84 posLabel 30.96
precisionF 98.98 recallF 91.51 f1score 95.10
********** fold 3 **********
train_loss:0.003310 train_acc:99.6051
valid loss:0.010244 valid_acc:97.7883
AccuracyF 97.79 posLabel 14.69
precisionF 93.55 recallF 91.58 f1score 92.55

---

去掉第一部分数据、以及提取信息

fold	train loss	valid loss	AccuracyF	precisionFn	recallFn	f1-score
fold 1	0.039308	0.073821	84.06	36.51	34.33	35.38
fold 2	0.039042	0.064052	78.37	55.00	52.38	53.66
fold 3	0.032341	0.085093	69.83	84.38	14.92	25.35
fold 4	0.045185	0.061185	85.96	46.53	70.15	55.95

********** fold 1 **********
train_loss:0.039308 train_acc:86.7173
valid loss:0.073821 valid_acc:84.0607
AccuracyF 84.06 posLabel 11.95
precisionF 36.51 recallF 34.33 f1score 35.38
********** fold 2 **********
train_loss:0.039042 train_acc:86.2113
valid loss:0.064052 valid_acc:78.3681
AccuracyF 78.37 posLabel 22.77
precisionF 55.00 recallF 52.38 f1score 53.66
********** fold 3 **********
train_loss:0.032341 train_acc:88.8678
valid loss:0.085093 valid_acc:69.8292
AccuracyF 69.83 posLabel 6.07
precisionF 84.38 recallF 14.92 f1score 25.35
********** fold 4 **********
train_loss:0.045185 train_acc:84.1872
valid loss:0.061185 valid_acc:85.9583
AccuracyF 85.96 posLabel 19.17
precisionF 46.53 recallF 70.15 f1score 55.95

fold	origin etp	nf etp	origin gn	nf gn	nc etp	nc gn
fold 1	98.969072	97.237569	98.461538	96.174863	0.	23.423423
fold 2	95.477387	94.623656	95.477387	92.146597	6.60793	5.333333
fold 3	89.622642	86.699507	83.700441	80	13.77551	12.621359
fold 4	83.333333	88.442211	77.55102	88.442211	38.78327	33.464567
fold 5	96.938776	97.802198	94.059406	97.26776	0.	31.111111

entropy
原始数据
k-fold f1 scored:[0.98969072 0.95477387 0.89622642 0.83333333 0.96938776]
剔除第一部分数据
k-fold f1 scored:[0.97237569 0.94623656 0.86699507 0.88442211 0.97802198]
gini
原始数据
k-fold f1 scored:[0.98461538 0.95477387 0.83700441 0.7755102 0.94059406]
剔除第一部分数据
k-fold f1 scored:[0.96174863 0.92146597 0.8 0.88442211 0.9726776 ]

去掉提取信息：
entropy
k-fold f1 scored:[0. 0.0660793 0.1377551 0.3878327 0. ]
gini
k-fold f1 scored:[0.23423423 0.05333333 0.12621359 0.33464567 0.31111111]