-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathRESULTS.txt
93 lines (86 loc) · 3.38 KB
/
RESULTS.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
Experiment: Effect of standard deviation data exclusion
Without std deviation:
2014-11-16T02:23:26-06:00 RESULT:
Inter avg: 1, Pre avg: 0
Inter correct: 31, Inter total: 31, Pre correct: 0, Pre total: 10
Overall Score: 0.756
NOTE: Above has higher score, but is worse result (see pre avg)
With std deviation (> 1 sigma excluded):
2014-11-16T02:28:44-06:00 RESULT:
Inter avg: 0.323, Pre avg: 1
Inter correct: 10, Inter total: 31, Pre correct: 10, Pre total: 10
Overall Score: 0.488
With std deviation (>2 sigma excluded):
2014-11-16T02:58:36-06:00 RESULT:
Inter avg: 0.387, Pre avg: 1
Inter correct: 12, Inter total: 31, Pre correct: 10, Pre total: 10
Overall Score: 0.537
Conclusion: >2 Sigma exclusion is used forwards.
Experiment: Effect of removing data scaling from 0-1
Without data scaling:
2014-11-16T03:38:26-06:00 RESULT:
Inter avg: 0.806, Pre avg: 0.6
Inter correct: 25, Inter total: 31, Pre correct: 6, Pre total: 10
Overall Score: 0.756
Note: Note sure about this, seems inter goes way up, but pre dropped. Overall average went up.
Note 2: It's possible this is higher due to percision being increased. Maybe this could be leveraged.
Conclusion: Continue using data scaling for now.
Experiment: Adding min and max values as a feature
With min and max as features (per file):
2014-11-16T03:48:59-06:00 RESULT:
Inter avg: 0.516, Pre avg: 0.6
Inter correct: 16, Inter total: 31, Pre correct: 6, Pre total: 10
Overall Score: 0.537
Conclusion: Not including min/max as features.
Experiment: Effect of removing absolute data cleaning
Without abs:
2014-11-16T04:20:59-06:00 Finished Dog_5_preictal_segment_0030.mat
2014-11-16T04:20:59-06:00 RESULT:
Inter avg: 1, Pre avg: 0
Inter correct: 31, Inter total: 31, Pre correct: 0, Pre total: 10
Overall Score: 0.756
Conclusion: Abs will continue
Experiment: More sigma values
With std deviation (>3 sigma exlcuded):
2014-11-16T04:52:50-06:00 RESULT:
Inter avg: 0.774, Pre avg: 0.8
Inter correct: 24, Inter total: 31, Pre correct: 8, Pre total: 10
Overall Score: 0.78
4 sigma = bad
3.5 sigma = shitty
Conclusion: 3 sigma forwards.
Experiment: Remove data scaling now that we are usign 3 sigma exclusion
without data scaling:
2014-11-16T14:31:02-06:00 RESULT:
Inter avg: 0.806, Pre avg: 0.6
Inter correct: 25, Inter total: 31, Pre correct: 6, Pre total: 10
Overall Score: 0.756
Conclusion: Wrecked pre score, barely improved inter score. Continuing use of data scaling
Experiment: Change min/max to the min max of the average, not the original data.
New min/max:
2014-11-18T01:17:38-06:00 RESULT:
Inter avg: 1, Pre avg: 0
Inter correct: 31, Inter total: 31, Pre correct: 0, Pre total: 10
Overall Score: 0.756
>2 sigma data exclusion with new min/max:
2014-11-18T02:05:42-06:00 RESULT:
Inter avg: 1, Pre avg: 0
Inter correct: 31, Inter total: 31, Pre correct: 0, Pre total: 10
Overall Score: 0.756
Disable abs?
terrible
Conclusion: removed scaling
Experiment: Trying ANNs, first independantly, then combined with SVM.
With ANN:
2014-11-18T15:18:11-06:00 RESULT:
Inter avg: 0.839, Pre avg: 0.5
Inter correct: 26, Inter total: 31, Pre correct: 5, Pre total: 10
Overall Score: 0.756
More samples:
eh
Experiment: Both ANN and SVM:
Result:
2014-11-18T19:09:55-06:00 RESULT:
Inter avg: 0.861, Pre avg: 0.4
Inter correct: 99, Inter total: 115, Pre correct: 4, Pre total: 10
Overall Score: 0.824