banter is a package for creating hierarchical acoustic event classifiers out of multiple call type detectors.
To install the latest version from GitHub:
# make sure you have devtools installed if(!require('devtools')) install.packages('devtools') # install package from GitHub devtools::install_github('ericarcher/banter')
The BANTER (Bio-Acoustic eveNT classifiER) model is initialized with a data frame of events. There is one row per event and it must have a column called
event.id which is a unique id for each event, and a column called
species which assigns each event to a given species. Every other column in the data.frame will be used as a predictor variable for the events.
In the package, an example data.frame is in the
train.data example data list as the
data(train.data) bant.mdl <- initBanterModel(train.data$events)
Next, detector data is added to the initialized BANTER model object. Each detector is a data.frame with a column called
event.id that associates the detected call with an event that the model was initialized with, and a
call.id column that provides a unique identifier for each call. Every other column will be used as a predictor variable for the calls.
In the package, example data.frames for three detectors are provided in the
$detectors element of the
train.data example data list. Here is an example of adding the burst pulse (
bant.mdl <- addBanterDetector( bant.mdl, data = train.data$detectors$bp, name = "bp", ntree = 10, sampsize = 1 )
addBanterDetector function can be called repeatedly to add additional detectors. Alternatively, if the detectors are all in a named list, they can be added at once:
bant.mdl <- addBanterDetector( bant.mdl, data = train.data$detectors, ntree = 10, sampsize = 1 )
Once all of the detectors have been added, then the full BANTER model is run:
bant.mdl <- runBanterModel(bant.mdl, ntree = 5000, sampsize = 3)
The model can be easily summarized:
randomForest model can be extracted for the event or detector models:
# extract event Random Forest model event.rf <- getBanterModel(bant.mdl, "event") # extract burst pulse (bp) Random Forest model bp.rf <- getBanterModel(bant.mdl, "bp")
These can then be visualized using other tools, such as those in the
To predict novel data, it must be in a list with the event data in the
$events element, and the detector data in a named list called
data(test.data) predict(bant.mdl, test.data)
Rankin, S., Archer, F., Keating, J. L., Oswald, J. N., Oswald, M., Curtis, A. and Barlow, J. (2017) Acoustic classification of dolphins in the California Current using whistles, echolocation clicks, and burst pulses. Mar Mam Sci, 33: 520-540. doi:10.1111/mms.12381
subsampleDetections()to draw a random number of detections per event
predict()so that species names in
new.dataare modified using
make.names()like they are in constructing the initial banter model.
summary()to display a matrix of all.
addBanterDetector()are now first checked to make sure they are syntactically valid and unique by comparing them with the results from
predict()are first checked to make sure at least one detector from model is present.
predict()will automatically have all species detector probabilities and detector proportions set to 0.
speciescolumn exists in
Any scripts or data that you put into this service are public.
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.