Skip to content

Commit 685b5c5

Browse files
fracpetefracpete
committed
added additional methods; updated version; added news item
1 parent bfe268d commit 685b5c5

10 files changed

+546
-56
lines changed

‎docs/maven.md

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -3,6 +3,6 @@ Add the following dependency to your `pom.xml`:
33
<dependency>
44
<groupId>net.sf.meka</groupId>
55
<artifactId>meka</artifactId>
6-
<version>1.9.7</version>
6+
<version>1.9.8</version>
77
</dependency>
88
```

‎docs/meka.classifiers.multilabel.MULAN.md

Lines changed: 5 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -8,6 +8,11 @@ http://mulan.sourceforge.net
88
Method Name
99
default: RAkEL1
1010

11+
* `-no-rename`
12+
13+
Don't rename attributes
14+
default: rename
15+
1116
* `-W <classifier name>`
1217

1318
Full name of base classifier.

‎docs/meka.classifiers.multilabel.Maniac.md

Lines changed: 4 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -100,6 +100,10 @@ J"org Wicker, Andrey Tyukin, Stefan Kramer: A Nonlinear Label Compression and Tr
100100

101101
Output additional statistics.
102102

103+
* `-use-qr`
104+
105+
Use QR decomposition to find coefficients
106+
103107
* `-output-debug-info`
104108

105109
If set, classifier is run in debug mode and

‎docs/meka.classifiers.multilabel.PLST.md

Lines changed: 4 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -97,6 +97,10 @@ For more information see:
9797

9898
Output additional statistics.
9999

100+
* `-use-qr`
101+
102+
Use QR decomposition to find coefficients
103+
100104
* `-output-debug-info`
101105

102106
If set, classifier is run in debug mode and

‎docs/meka.classifiers.multilabel.meta.ERFH.md

Lines changed: 157 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,157 @@
1+
## Synopsis
2+
Extremely Randomised Forest of HOMER trees.
3+
4+
## BibTeX
5+
```
6+
@inproceedings{Li2017,
7+
address = {Cham},
8+
author = {Li, Jinxia and Zheng, Yihan and Han, Chao and Wu, Qingyao and Chen, Jian},
9+
booktitle = {Intelligence Science and Big Data Engineering},
10+
editor = {Sun, Yi and Lu, Huchuan and Zhang, Lihe and Yang, Jian and Huang, Hua},
11+
pages = {450--460},
12+
publisher = {Springer International Publishing},
13+
title = {Extremely Randomized Forest with Hierarchy of Multi-label Classifiers},
14+
year = {2017},
15+
ISBN = {978-3-319-67777-4}
16+
}
17+
```
18+
## Options
19+
* `-T threshold`
20+
21+
Prediction threshold
22+
23+
* `-I <num>`
24+
25+
Sets the number of models (default 10)
26+
27+
* `-P <size percentage>`
28+
29+
Size of each bag, as a percentage of total training size (default 67)
30+
31+
* `-S <seed>`
32+
33+
Random number seed for sampling (default 1)
34+
35+
* `-W <classifier name>`
36+
37+
Full name of base classifier.
38+
(default: meka.classifiers.multilabel.BR)
39+
40+
* `-output-debug-info`
41+
42+
If set, classifier is run in debug mode and
43+
may output additional info to the console
44+
45+
* `-do-not-check-capabilities`
46+
47+
If set, classifier capabilities are not checked before classifier is built
48+
(use with caution).
49+
50+
* `-num-decimal-places`
51+
52+
The number of decimal places for the output of numbers in the model (default 2).
53+
54+
* `-batch-size`
55+
56+
The desired batch size for batch prediction (default 100).
57+
58+
**Options specific to classifier meka.classifiers.multilabel.BR:**
59+
60+
* `-W <classifier name>`
61+
62+
Full name of base classifier.
63+
(default: weka.classifiers.trees.J48)
64+
65+
* `-output-debug-info`
66+
67+
If set, classifier is run in debug mode and
68+
may output additional info to the console
69+
70+
* `-do-not-check-capabilities`
71+
72+
If set, classifier capabilities are not checked before classifier is built
73+
(use with caution).
74+
75+
* `-num-decimal-places`
76+
77+
The number of decimal places for the output of numbers in the model (default 2).
78+
79+
* `-batch-size`
80+
81+
The desired batch size for batch prediction (default 100).
82+
83+
**Options specific to classifier weka.classifiers.trees.J48:**
84+
85+
* `-U`
86+
87+
Use unpruned tree.
88+
89+
* `-O`
90+
91+
Do not collapse tree.
92+
93+
* `-C <pruning confidence>`
94+
95+
Set confidence threshold for pruning.
96+
(default 0.25)
97+
98+
* `-M <minimum number of instances>`
99+
100+
Set minimum number of instances per leaf.
101+
(default 2)
102+
103+
* `-R`
104+
105+
Use reduced error pruning.
106+
107+
* `-N <number of folds>`
108+
109+
Set number of folds for reduced error
110+
pruning. One fold is used as pruning set.
111+
(default 3)
112+
113+
* `-B`
114+
115+
Use binary splits only.
116+
117+
* `-S`
118+
119+
Do not perform subtree raising.
120+
121+
* `-L`
122+
123+
Do not clean up after the tree has been built.
124+
125+
* `-A`
126+
127+
Laplace smoothing for predicted probabilities.
128+
129+
* `-J`
130+
131+
Do not use MDL correction for info gain on numeric attributes.
132+
133+
* `-Q <seed>`
134+
135+
Seed for random data shuffling (default 1).
136+
137+
* `-doNotMakeSplitPointActualValue`
138+
139+
Do not make split point actual value.
140+
141+
* `-output-debug-info`
142+
143+
If set, classifier is run in debug mode and
144+
may output additional info to the console
145+
146+
* `-do-not-check-capabilities`
147+
148+
If set, classifier capabilities are not checked before classifier is built
149+
(use with caution).
150+
151+
* `-num-decimal-places`
152+
153+
The number of decimal places for the output of numbers in the model (default 2).
154+
155+
* `-batch-size`
156+
157+
The desired batch size for batch prediction (default 100).

‎docs/meka.classifiers.multilabel.meta.HOMER.md

Lines changed: 156 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,156 @@
1+
## Synopsis
2+
HOMER tree algorithm. For more information see:
3+
Tsoumakas, Grigorios, Katakis, Ioannis, Vlahavas, Ioannis: Effective and efficient multilabel classification in domains with large number of labels. In: Proc. ECML/PKDD 2008 Workshop on Mining Multidimensional Data (MMD’08), 53--59, 2008.
4+
5+
## BibTeX
6+
```
7+
@inproceedings{Tsoumakas2008,
8+
author = {Tsoumakas, Grigorios and Katakis, Ioannis and Vlahavas, Ioannis},
9+
booktitle = {Proc. ECML/PKDD 2008 Workshop on Mining Multidimensional Data (MMD’08)},
10+
organization = {sn},
11+
pages = {53--59},
12+
title = {Effective and efficient multilabel classification in domains with large number of labels},
13+
volume = {21},
14+
year = {2008}
15+
}
16+
```
17+
## Options
18+
* `-k K`
19+
20+
The number of partitions per level.
21+
22+
* `-S seed`
23+
24+
The seed to set.
25+
26+
* `-ls class`
27+
28+
The label splitter class to use.
29+
30+
* `-t threshold`
31+
32+
The threshold for the multi-label classifier distribution
33+
34+
* `-W <classifier name>`
35+
36+
Full name of base classifier.
37+
(default: meka.classifiers.multilabel.BR)
38+
39+
* `-output-debug-info`
40+
41+
If set, classifier is run in debug mode and
42+
may output additional info to the console
43+
44+
* `-do-not-check-capabilities`
45+
46+
If set, classifier capabilities are not checked before classifier is built
47+
(use with caution).
48+
49+
* `-num-decimal-places`
50+
51+
The number of decimal places for the output of numbers in the model (default 2).
52+
53+
* `-batch-size`
54+
55+
The desired batch size for batch prediction (default 100).
56+
57+
**Options specific to classifier meka.classifiers.multilabel.BR:**
58+
59+
* `-W <classifier name>`
60+
61+
Full name of base classifier.
62+
(default: weka.classifiers.trees.J48)
63+
64+
* `-output-debug-info`
65+
66+
If set, classifier is run in debug mode and
67+
may output additional info to the console
68+
69+
* `-do-not-check-capabilities`
70+
71+
If set, classifier capabilities are not checked before classifier is built
72+
(use with caution).
73+
74+
* `-num-decimal-places`
75+
76+
The number of decimal places for the output of numbers in the model (default 2).
77+
78+
* `-batch-size`
79+
80+
The desired batch size for batch prediction (default 100).
81+
82+
**Options specific to classifier weka.classifiers.trees.J48:**
83+
84+
* `-U`
85+
86+
Use unpruned tree.
87+
88+
* `-O`
89+
90+
Do not collapse tree.
91+
92+
* `-C <pruning confidence>`
93+
94+
Set confidence threshold for pruning.
95+
(default 0.25)
96+
97+
* `-M <minimum number of instances>`
98+
99+
Set minimum number of instances per leaf.
100+
(default 2)
101+
102+
* `-R`
103+
104+
Use reduced error pruning.
105+
106+
* `-N <number of folds>`
107+
108+
Set number of folds for reduced error
109+
pruning. One fold is used as pruning set.
110+
(default 3)
111+
112+
* `-B`
113+
114+
Use binary splits only.
115+
116+
* `-S`
117+
118+
Do not perform subtree raising.
119+
120+
* `-L`
121+
122+
Do not clean up after the tree has been built.
123+
124+
* `-A`
125+
126+
Laplace smoothing for predicted probabilities.
127+
128+
* `-J`
129+
130+
Do not use MDL correction for info gain on numeric attributes.
131+
132+
* `-Q <seed>`
133+
134+
Seed for random data shuffling (default 1).
135+
136+
* `-doNotMakeSplitPointActualValue`
137+
138+
Do not make split point actual value.
139+
140+
* `-output-debug-info`
141+
142+
If set, classifier is run in debug mode and
143+
may output additional info to the console
144+
145+
* `-do-not-check-capabilities`
146+
147+
If set, classifier capabilities are not checked before classifier is built
148+
(use with caution).
149+
150+
* `-num-decimal-places`
151+
152+
The number of decimal places for the output of numbers in the model (default 2).
153+
154+
* `-batch-size`
155+
156+
The desired batch size for batch prediction (default 100).

0 commit comments

Comments
 (0)