The size of the outliers label is too large

[choonghyunryu]

The size of the label for outliers in the fviz_cluster function is too large.
After performing DBSCAN clustering with the following code, the outliers label is rendered too large in the plot.

trans_usa <- scale(USArrests)

dc <- fpc::dbscan(trans_usa, eps = 1.24ss, MinPts = NCOL(trans_usa) + 1)
fviz_cluster(dc, data= trans_usa)

I can not see any arguments in the help page that change the size of the outliers label.
The version of the factoextra package is 1.0.5.
The operating system is CentOS 7, R 3.4.1 environment

[duttashi]

@choonghyunryu use labelsize to control the size of labels. By the way, the minimum reproducible example you posted is not entirely reproducible. The problem in your example is, eps1.24ss where ss in unrecognizable. Removing it, like dc <- fpc::dbscan(trans_usa, eps = 1.24, MinPts = NCOL(trans_usa) + 1) makes your example reproducible.

On R console, type `help("fviz_cluster") to see the usage.

Try the following,

trans_usa <- scale(USArrests)
dc <- fpc::dbscan(trans_usa, eps = 1.24, MinPts = NCOL(trans_usa) + 1)
fviz_cluster(dc, data= trans_usa, labelsize = 2)

[choonghyunryu]

@duttashi Thank you for your kind answer.
I'am sorry. eps1.24ss was the wrong typing.

I used the labelsize argument.
However, if you use the labelsize argument to change the size of the label of the outliers (excluded from the cluster) to a smaller size, the labels of the other observations contained in the cluster do not seem very small.

Is there any way to a smaller size of the outlier label and make the labels of other observations visible?

[duttashi]

@choonghyunryu its very difficult for me to understand what you've written.
"However, if you use the labelsize argument to change the size of the label of the outliers (excluded from the cluster) to a smaller size, the labels of the other observations contained in the cluster do not seem very small. Is there any way to a smaller size of the outlier label and make the labels of other observations visible?"
Please provide a minimum reprex of what you are looking for.

[choonghyunryu]

@duttashi Sorry. I am not fluent in English and communicate using a translate application.
In the visualization of the result below, the size of the label within the polygon is too small.
I hope that the values of the labelsize argument are common to labels outside of polygons and polygons.

trans_usa <- scale(USArrests)
dc <- fpc::dbscan(trans_usa, eps = 1.24, MinPts = NCOL(trans_usa) + 1)
fviz_cluster(dc, data= trans_usa, labelsize = 2)

[duttashi]

@choonghyunryu the size of the label within the polygon can be adjusted. Did you even try to see the signature for fviz_cluster() which specifies the labelsize argument? I think not.

So if you change the value in labelsize from 2 to say 8 or 10, the fontsize changes accordingly.

library(factoextra)
trans_usa <- scale(USArrests)
dc <- fpc::dbscan(trans_usa, eps = 1.24, MinPts = NCOL(trans_usa) + 1)
fviz_cluster(dc, data= trans_usa, labelsize = 8)

And why should the size of outlier label and the label for other objects within or outside the polygon should be the same? If the font size for the outlier label is the same as the font size for a non-outlier label than distinguishing outlier from non-outlier will be difficult. And this will obviate the entire purpose of visualizing outliers.

[choonghyunryu]

@duttashi Thank you for your quick response.
I wanted to distinguish between outliers and cluster observations, but not too big an outlier label.
Thank you for your suggestion.

[kassambara]

Thank you @duttashi for your input.

Fixed now, new arguments outlier.pointsize and outlier.labelsize added in fviz_cluster() to customize outliers detected with DBSCAN.

You need to install the latest developmental version of factoextra from Github

[duttashi]

@kassambara and @choonghyunryu 👍

[choonghyunryu]

Thank you @kassambara and @duttashi