scheidan / RuminatoR

RuminatoR - an R package to classify feeding behavior of ruminants

RuminatoR allows to classify different behaviors of cows and other ruminants based on pressure signals measured in the headcollar.

This is a reimplementation based on Scheidegger (2008).

Installation

1. Install R and R-Studio (or any other editor).

2. Install remotes (type in the R command line)

install.packages("remotes")

3. Install RuminatoR (type in the R command line)

library(remotes)
install_github("scheidan/RuminatoR")

Usage

You can get help in R with ?train or ?classify.

Example

We assume the data are available as dataframes in the following structure:

head(trainings.data)
                 time pressure   activity
1 2008-09-04 20:34:50   1076.8 ruminating
2 2008-09-04 20:34:50   1043.6 ruminating
3 2008-09-04 20:34:50   1029.1 ruminating

head(new.data)
                 time pressure
1 2008-09-04 06:59:59   1085.4
2 2008-09-04 06:59:59   1069.1
3 2008-09-04 06:59:59   1049.8

library(RuminatoR)
library(ggplot2)  # optional, only for visualization

## --------------------------
## 1) train classifier

rf.classifier <- train(data.training,
                       min.amplitude=30, min.dt=6)

## ---------------------------------
## 2) apply classifier to new data

dat.class <- classify(rf.classifier, new.data,
                      max.dt.peak = 30, min.n.peaks = 10)

head(dat.class)
table(dat.class$activity.predicted)

## -----------
## 3) visualize result

## color line
ggplot(dat.class, aes(x=time, y=pressure, color=activity.predicted)) +
    geom_path(aes(group=1))

## color background
ggplot(dat.class) +
    ## background
    geom_rect(aes(NULL, NULL,
                  xmin=time-0.5, xmax=time+0.5,
                  ymin=-Inf, ymax=Inf,
                  fill=activity.predicted)) +
    ## add pressure
    geom_path(aes(x=time, y=pressure))

References

Scheidegger, A. (2008). Klassifikation des Fressverhaltens von Kühen. Degree thesis, Zurich University of Applied Sciences, Zurich.