README.md

In mjafin/PredictiveModelPipeline: Feature selection and evaluation for predictive high dimensional models

PredictiveModelPipeline

High dimensional predictive modelling in R

Conda installation of dependencies

# get miniconda
wget https://repo.anaconda.com/miniconda/Miniconda2-latest-Linux-x86_64.sh
bash Miniconda2-latest-Linux-x86_64.sh  -b
cd miniconda2/bin

# add bioconda
./conda config --add channels defaults && ./conda config --add channels bioconda && ./conda config --add channels conda-forge

# install R and dependencies
./conda create -n predmod_r r-base r-yaml r-hmisc r-sda r-mixomics r-epibasix r-survival r-domc r-foreach r-cvtools -y

Basic usage

require(sda)
data(singh2002)
colnames(singh2002$x) = paste("V",1:dim(singh2002$x)[2],sep="") # need unique column / variable names
rownames(singh2002$x) = paste("S",1:dim(singh2002$x)[1],sep="") # need unique row / sample names
names(singh2002$y) = rownames(singh2002$x)

source("PredictiveModelPipeline.R")
mypredmodel = PredictiveModel("/path/to/data/analysis_testing.yaml") # analysis template with settings
mypredmodel$PrepareData(Xtrain=singh2002$x , ytrain=singh2002$y) # prepare CV folds by preprocessing methods
mypredmodel$MachineLearning() # all settings specified in yaml # perform machine learning (parallelised)
require(epibasix)
mypredmodel$Plot() # make AUC, Sens, Spec, Acc plots
#mypredmodel$FinalModelBuild() # build a final model but first select number of features based on plots, specify in yaml and reload yaml
#myout = mypredmodel$Predict(testdata) # test on external data. dimensionality / number of columns must match training matrix

mjafin/PredictiveModelPipeline documentation built on Feb. 11, 2020, 3:39 a.m.