.Rhistory

library(nsga2R)
library(doParallel)
library(MLmetrics)
library(ggplot2)
library(parallel)
library(randomForest)
library(GenAlgo)
library(caret)
library(stringr)
library(devtools)
library(Metrics)
library(iRF)
library(cluster)
library(doSNOW)
library(doMC)
library(purrr)
library(dplyr)
library(SGL)
library(e1071)
library(MLmetrics)
library("e1071", lib.loc="~/R/win-library/3.1")
registerDoMC()
set.seed(100)
#Function used to set up paralell computing
resetCluster <- function(){
stopCluster(cl)
cores = detectCores(all.tests = FALSE, logical = TRUE)
cl<-makeCluster(cores-1, outfile="Log.txt")
registerDoSNOW(cl)
}
cores = detectCores(all.tests = FALSE, logical = TRUE)
#Reading data sources
data = load('data/completeDataset.csv')
#Reading data sources
data = load('data/completeDataset.RDS')
data <- data[-nearZeroVar(data)] #Here we remove any variables where the variance is close to zero
#Reading data sources
data = load('data/completeDataset.RDS')
#Reading data sources
data = readRDS('data/completeDataset.RDS')
data <- data[-nearZeroVar(data)] #Here we remove any variables where the variance is close to zero
expressionData = readRDS('data/expressionOnly.RDS')
metabolicExpression = readRDS('data/metabolic_gene_data.RDS')
biomassGrowthFlux = 'r_4041';
biomassIndex = 717;
growth = data$log2relT  #Extract the growth rates
strains = data$Row #Extract the strains names
data$Row <- NULL
data$log2relT <- NULL #
expressionData[,1] <- NULL #Remove the growth rate data
iRFIndicies = read.csv('data/Features_Extracted_Using_iRF.csv', header = F)
SGLIndicies = read.csv('data/Features_Extracted_Using_SGL.csv', header = F)
GeneticAlgoIndicies = read.csv('data/genetic_feature_selection_features.csv', header = FALSE,  stringsAsFactors=FALSE)
iRFData = data[, which(colnames(data) %in% unlist(iRFIndicies[,1])) ]
SGLData = data[, which(colnames(data) %in% unlist(SGLIndicies[,1]))]
fluxData <- data[, -which(colnames(data) %in% colnames(expressionData))] #flux data is all data without gene data
#Full data partitions
testingIndices<- unlist(read.csv('testing_index.csv', header = F))
testingIndices <- testingIndices + 1 # to account for index changes between R and Python
testingTarget <- growth[testingIndices]
trainingTarget <- growth[-testingIndices]
trainingData <- data[-testingIndices,]
testingData <-  data[testingIndices,]
normParam <- preProcess(trainingData)
normalisedTestData <- predict(normParam, testingData)
normalisedTrainingData <- predict(normParam, trainingData)
#Expression data partition
expressionTrainingData <- expressionData[-testingIndices,]
expressionTestingData <- expressionData[testingIndices,]
normParamExp <- preProcess(expressionTrainingData)
normalisedExpressionTrainingData <- predict(normParamExp, expressionTrainingData)
normalisedExpressionTestingData <- predict(normParamExp, expressionTestingData)
#Flux data partition
fluxTrainingData <- fluxData[-testingIndices,]
fluxTestingData <- fluxData[testingIndices,]
normParamFlux <- preProcess(fluxTrainingData)
normalisedFluxTrain <- predict(normParamFlux, fluxTrainingData)
normalisedFluxTest <- predict(normParamFlux, fluxTestingData)
#Metabolic expression data partition
metabolicExpressionTrainingData <- metabolicExpression[-testingIndices,]
metabolicExpressionTestingData <- metabolicExpression[testingIndices,]
normParamMetabolicExpression <- preProcess(metabolicExpressionTrainingData)
normalisedMetabolicExpressionTrain <- predict(normParamMetabolicExpression, metabolicExpressionTrainingData)
normalisedMetabolicExpressionTest <- predict(normParamMetabolicExpression, metabolicExpressionTestingData)
#SGL expression data partition
SGLTrainingData <- SGLData[-testingIndices,]
SGLTestingData <- SGLData[testingIndices,]
normParamSGL <- preProcess(SGLTrainingData)
normalisedSGLTrain <- predict(normParamSGL, SGLTrainingData)
normalisedSGLTest <- predict(normParamSGL, SGLTestingData)
#iRF expression data partition
iRFTrainingData <- iRFData[-testingIndices,]
iRFTestingData <- iRFData[testingIndices,]
normParamiRF <- preProcess(iRFTrainingData)
normalisediRFTrain <- predict(normParamiRF, iRFTrainingData)
normalisediRFTest <- predict(normParamiRF, iRFTestingData)
# Genetic algo features, need to check the normalisation is correct - come back to this
gen_features_index <- read.csv('data/genetic_feature_selection_features.csv', header = F)
normGenTrain <- vector(mode = "list", length = dim(gen_features_index)[1] - 1)
normGenTest <- vector(mode = "list", length = dim(gen_features_index)[1] - 1)
for (j in c(1:9)){
features <- word(unlist(gen_features_index[,j]))
print(features)
tr = data[-testingIndices, features]
te = data[testingIndices, features]
normParam <- preProcess(tr)
normGenTrain[[j]] = predict(normParam, tr)
normGenTest[[j]] = predict(normParam, te)
}