TLBC-package: Two-Level Behavior Classification

In TLBC: Two-Level Behavior Classification

Description

Contains functions for training and applying two-level random forest and hidden Markov models for human behavior classification from raw tri-axial accelerometer and/or GPS data.

This code works with csv data from Actigraph accelerometers (please export in RAW format, without timestamps), and/or with GPS data processed by the PALMS GPS cleaning software.

The TLBC classifier uses six behavior labels:

• Sitting

• Standing Still

• Standing Moving

• Walking/Running

• Bicycling

• Vehicle

Function classify uses a pre-learned TLBC model to classify accelerometer and/or GPS data with behavior labels. Pre-trained models that have been trained on three UCSD datasets are available for download.

Function trainModel trains a TLBC model from annotated accelerometer and/or GPS data.

Function calcPerformance computes the accuracy of predictions made on a given dataset.

Function looXval performs leave-one-out cross-validation on a dataset.

Details

Package:	TLBC
Type:	Package
Version:	1.0
Date:	2015-05-29
License:	GPL-2

Author(s)

Katherine Ellis <[e-mail]>

See Also

randomForest, HMM

Examples

## Not run: 

# train a new model
myAnnotations="~/myStudy/annotations.csv"
myAccel="~/myStudy/HipGT3X+"
myGPS="~/myStudy/GPS.csv"
WS=60
myModel="~/myStudy/myModel.RData"
trainModel(annotations=myAnnotations, accelerometers=myAccel, GPS=myGPS, winSize=WS, 
modelName=myModel)

# classify using a model computed yourself
myAccel="~/myStudy/HipGT3X+"
myGPS="~/myStudy/GPS.csv"
myModel="~/myStudy/myModel.RData"
myPredictions="~/myStudy/myModelPredictions"
classify(accelerometers=myAccel, GPS=myGPS, modelName=myModel, saveDir=myPredictions)

# compute the performance of a model on a dataset
myAnnotations="~/myStudy/annotations.csv"
myPredictions="~/myStudy/myModelPredictions"
WS=60
calcPerformance(annotations=myAnnotations, predictions=myPredictions, winSize=WS)

# perform leave-one-out cross-validation on a dataset
myAnnotations="~/myStudy/annotations.csv"
myAccel="~/myStudy/HipGT3X+"
WS=60
myPredictions="~/myStudy/looXvalPredictions"
looXval(annotations=myAnnotations, accelerometers=myAccel, winSize=WS, saveDir=myPredictions)


## End(Not run)

TLBC documentation built on May 1, 2019, 8:44 p.m.