Scribd
Upload
207 views

R Assignment Classification of Ocean Microbes

1) The document describes using machine learning algorithms like decision trees, random forests, and SVM to classify different types of ocean microbes (crypto, nano, pico, synecho, ultra) based on characteristics like size and chlorophyll content. 2) It finds that random forests achieves the highest accuracy at 92% while decision trees and SVM achieve slightly lower accuracy around 92% and 92%. 3) Feature importance is also calculated which finds that characteristics like PE, chlorophyll content, and size are the most important predictors for classification.

Uploaded by

api-302999229
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
207 views

R Assignment Classification of Ocean Microbes

1) The document describes using machine learning algorithms like decision trees, random forests, and SVM to classify different types of ocean microbes (crypto, nano, pico, synecho, ultra) based on characteristics like size and chlorophyll content. 2) It finds that random forests achieves the highest accuracy at 92% while decision trees and SVM achieve slightly lower accuracy around 92% and 92%. 3) Feature importance is also calculated which finds that characteristics like PE, chlorophyll content, and size are the most important predictors for classification.

Uploaded by

api-302999229
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 13

RAssignment:ClassificationofOcean

Microbes
Chia
Loaddataintoworkspace.
library("tree")
library("ggplot2")
seaflow<read.csv("seaflow_21min.csv",header=T)
Plotpeagainstchl_small
##file_idtimecell_idd1
##Min.:203.0Min.:12.0Min.:0Min.:1328
##1stQu.:204.01stQu.:174.01stQu.:74861stQu.:7296
##Median:206.0Median:362.0Median:14995Median:17728
##Mean:206.2Mean:341.5Mean:15008Mean:17039
##3rdQu.:208.03rdQu.:503.03rdQu.:224013rdQu.:24512
##Max.:209.0Max.:643.0Max.:32081Max.:54048
##d2fsc_smallfsc_perpfsc_big
##Min.:32Min.:10005Min.:0Min.:32384
##1stQu.:95841stQu.:313411stQu.:134961stQu.:32400
##Median:18512Median:35483Median:18069Median:32400
##Mean:17437Mean:34919Mean:17646Mean:32405
##3rdQu.:246563rdQu.:391843rdQu.:222433rdQu.:32416
##Max.:54688Max.:65424Max.:63456Max.:32464
##pechl_smallchl_bigpop
##Min.:0Min.:3485Min.:0crypto:102
##1stQu.:16351stQu.:225251stQu.:2800nano:12698
##Median:2421Median:30512Median:7744pico:20860
##Mean:5325Mean:30164Mean:8328synecho:18146
##3rdQu.:58543rdQu.:382993rdQu.:12880ultra:20537
##Max.:58675Max.:64832Max.:57184

Partitions

library("caret")
##Loadingrequiredpackage:lattice
train_index<createDataPartition(1:nrow(seaflow),1,.5,list=F)
trainset<seaflow[train_index,]
testset<seaflow[train_index,]
DecisonTree
library("rpart")
fol<formula(pop~fsc_small+fsc_perp+fsc_big+pe+chl_big+chl_small)
model_dt<rpart(fol,method="class",data=trainset)
print(model_dt)
##n=36172
##
##node),split,n,loss,yval,(yprob)
##*denotesterminalnode
##
##1)root3617225756pico(0.00140.170.290.250.28)
##2)pe<5001.52631215954pico(00.220.3900.39)
##4)chl_small<32273.5115012055pico(00.000350.8200.18)*
##5)chl_small>=32273.5148116703ultra(00.390.06200.55)
##10)chl_small>=41054.55382771nano(00.86000.14)*
##11)chl_small<41054.594292092ultra(00.130.09700.78)*
##3)pe>=5001.59860794synecho(0.00520.0540.00590.920.015)
##6)chl_small>=37876668145nano(0.0760.7800.0730.067)*
##7)chl_small<378769192175synecho(00.00130.00630.980.011)*
predict_dt<predict(model_dt,newdata=testset)
prepop_dt<cbind(rownames(predict_dt),colnames(predict_dt)[apply(predict_dt,1,which.max)])
accuracy_dt<(sum(prepop_dt[,2]==seaflow[prepop_dt[,1],'pop']))/dim(prepop_dt)[1]
##Loadingrequiredpackage:grid

RadomForest
library("randomForest")
##randomForest4.612
##TyperfNews()toseenewfeatures/changes/bugfixes.
model_rf<randomForest(fol,data=trainset)
#model_rf<randomForest(fol,data=trainset,importance=T,keep.forest=T)
#testing
predict_rf<predict(model_rf,newdata=testset)
accuracy_rf<(sum(predict_rf==seaflow[labels(predict_rf),'pop']))/length(predict_rf)
print(accuracy_rf)
##[1]0.9202676
#Ginicoef
#"Thehigherthenumber,themoretheginiimpurityscoredecreasesbybranchingonthisvar
iable,indicatingthatthevariableismoreimportant"
importance(model_rf)

##MeanDecreaseGini
##fsc_small2705.5014
##fsc_perp2098.5750
##fsc_big206.2317
##pe8802.4334
##chl_big4923.9151
##chl_small8113.0261
SVM
library("e1071")
#SVM
#model<svm(fol,data=trainingdata)
model_svm<svm(fol,data=trainset)
#testing
predict_svm<predict(model_svm,newdata=testset)
accuracy_svm<(sum(predict_svm==seaflow[labels(predict_svm),'pop']))/length(predict_svm)
print(accuracy_svm)
##[1]0.919687
ConfusionMatrices
##true
##predcryptonanopicosynechoultra
##nano515204138870
##pico01949101978
##synecho0934904294
##ultra0115491807286
##true
##predcryptonanopicosynechoultra
##crypto512010
##nano0556801331
##pico001008101383
##synecho03390774
##ultra079536018510
##true
##predcryptonanopicosynechoultra
##crypto491020
##nano1565001369
##pico0010036281367
##synecho155590478
##ultra071235328484

##ConfusionMatrixandStatistics
##
##Reference
##Predictioncryptonanopicosynechoultra
##crypto491020
##nano1565001369
##pico0010036281367
##synecho155590478
##ultra071235328484
##
##OverallStatistics
##
##Accuracy:0.9197
##95%CI:(0.9168,0.9225)
##NoInformationRate:0.2887
##PValue[Acc>NIR]:<2.2e16
##
##Kappa:0.8917
##Mcnemar'sTestPValue:NA
##
##StatisticsbyClass:
##
##Class:cryptoClass:nanoClass:picoClass:synecho
##Sensitivity0.9607840.88720.96090.9964
##Specificity0.9999170.98760.94580.9975
##PosPredValue0.9423080.93840.87800.9924
##NegPredValue0.9999450.97620.98350.9988
##Prevalence0.0014100.17610.28870.2510
##DetectionRate0.0013550.15620.27750.2501
##DetectionPrevalence0.0014380.16650.31600.2520
##BalancedAccuracy0.9803510.93740.95340.9969
##Class:ultra
##Sensitivity0.8295
##Specificity0.9589
##PosPredValue0.8883
##NegPredValue0.9345
##Prevalence0.2828
##DetectionRate0.2346
##DetectionPrevalence0.2641
##BalancedAccuracy0.8942
DiscrteVariable?

You might also like