tidy.ftrank: Tidy up multiple resampling based ranking results.

Description Usage Arguments Details Value Author(s) See Also Examples

View source: R/ftrank.R

Description

Convenience function to ease the output of summ.ftrank objects.

Usage

1
tidy.ftrank(frsum, lmod = NULL, sorting = "Stat", tidy = "DisId", nam = "AlgId", decreasing = TRUE, file = NULL)

Arguments

frsum

mfr.sum objects

lmod

List of objects to be printed out - Default all objects

sorting

Should the results be sorted according to an argument contained in frsum$frdef - Default sorted by variable name

tidy

Tidy the output according to an argument contained in frsum$frdef

nam

Concatenate the original column names with content in to an argument contained in frsum$frdef

decreasing

Sorting order if argument sorting is not NULL

file

Write results into one file or several files

Details

This function has been designed to provide maximum facilities to ease both screen printing and file write of complex list of resampling based feature ranking resulting from different comparisons and/or different FR settings. It uses heavily the information contained in the FR definition table (frdef in the frsum object) to: 1) merge different tables (1 FR method on 1 dataset) in a bigger table (several FR and/or datasets) 2) sort each ranking table given a statistical output and 3) group tables into subsets corresponding to FR technique or datasets. File writing option can be rather useful when 1), 2) or 3) must be repeated manually several times:

Value

List of tables or list of files.

Author(s)

David Enot [email protected]

See Also

summ.ftrank

Examples

 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
################################################################
### Example involving 5 discrimination tasks by 2 feature ranking techniques
data(abr1)
y   <- factor(abr1$fact$class)
x <- preproc(abr1$pos , y=y, method=c("log10","TICnorm"),add=1)[,110:500]  
## Select classes all pairwise problems involving class with 1 
dat <- dat.sel1(x, y, pwise="1",mclass=NULL,pars=valipars(sampling="boot",niter=2,nreps=5))

### Perform AUC and Random Forest ranking
resauc = lapply(dat, function(x) feat.rank.re(x,method="fs.auc"))
resrf = lapply(dat, function(x) feat.rank.re(x,method="fs.rf",ntree=100))

################################################################
## Aggregate all the models
mfr=ftrank.agg(resauc,resrf)
print(mfr$frdef)
## Compute mrp-val for all the FR models
frsum=summ.ftrank(mfr,qtl=.7)

## print the FR components
print(frsum$frdef)

## have a look at the first 5 variables in the second model
print(frsum$frsum[[2]][1:5,])

################################################################
##### Let's concentrate on the models 1 and 6 first
## No need to tidy here - no sorting of the results and 
## add the content of "Alg" in the columns names
res=tidy.ftrank(frsum,lmod=c(1,6),tidy=NULL,sorting=NULL,nam="Alg")

## because we did not tidy the ranking tables
## rankings from models 1 and 6 are concatenated
## in the first field of res 
## print out the first 5 lines
res[1:5,]

## Same as before but we sort the results according to "Stat" 
## in decreasing order
res=tidy.ftrank(frsum,lmod=c(1,6),tidy=NULL,sorting="Stat",nam="Alg",decreasing=TRUE)
res[1:5,]

## Same as before but we sort the results according to "Rank" 
## in increasing order (of course)
res=tidy.ftrank(frsum,lmod=c(1,6),tidy=NULL,sorting="Rank",nam="Alg",decreasing=FALSE)
res[1:5,]

################################################################
###### Tidy all the rankings according to the discrimination task
## in this case we set tidy="Dis"
## all models: lmod=NULL
## sorting according to Stat
## append the name contained in Alg to the column name

res=tidy.ftrank(frsum,lmod=NULL,tidy="Dis",sorting="Stat",nam="Alg")

## Discrimination task tags are the name for each field of res
names(res)
## same as before for discrimination'1~2' or 1
res[[1]][1:5,]
res$'1~2'[1:5,]

## discrimination'2~3' or 2 ...
res[[2]][1:5,]


################################################################
###### Tidy all the rankings according to the FR method
## in this case we set tidy="Alg" for e.g.
## all models: lmod=NULL
## sorting according to Stat
## append the name contained in Dis to the column name

res=tidy.ftrank(frsum,lmod=NULL,tidy="Alg",sorting="Stat",nam="Dis")

## FR technique method are the name for each field or res
names(res)

## the top 5 variables highlighted by AUC in the 5 comparisons ...
res$'fs.auc'[1:5,]

################################################################
###### Redirect the output to one or several files
## print each tables to a CSV files starting by "testtidy"
## followed by the string contained in tidy (here "Dis")
## Not run: tidy=tidy.ftrank(frsum,lmod=NULL,tidy="Dis",sorting="Stat",nam="Alg",file="testtidy")

wilsontom/FIEmspro documentation built on Feb. 19, 2018, 9:03 a.m.