Skip to content

Project-Report.md

Latest commit

 

History

History
129 lines (125 loc) · 3.9 KB

Project-Report.md

File metadata and controls

129 lines (125 loc) · 3.9 KB
title MachineLearning
author Reynaldo
date 25 de septiembre de 2017
output html_document 
knitr::opts_chunk$set(echo = TRUE)

step 1: download zip file from website to new folder call: data

getwd()
if(!file.exists("./MLearning")) dir.create("./MLearning")
fileUrl <- "https://d396qusza40orc.cloudfront.net/predmachlearn/pml-training.csv"
download.file(fileUrl, destfile = "./Mlearning/TrainingData.csv")
fileUrl <- "https://d396qusza40orc.cloudfront.net/predmachlearn/pml-testing.csv"
download.file(fileUrl, destfile = "./Mlearning/TestingData.csv")

step 2: load data into R and get dimension of every new read

rm(list=ls()) 
TrD <- read.csv("./MLearning/TrainingData.csv",na.strings=c('','NA'))
TrD1<-TrD[,!apply(TrD,2,function(x) any(is.na(x)) )]
TrD2<-TrD1[,-c(1:7)]
TtD <- read.csv("./MLearning/TestingData.csv")
dim(TrD)
dim(TrD2)
head(TrD2)
dim(TtD)

step3: Load Packages

library(caret)
library(rpart)
library(rpart.plot)
library(randomForest)
library(corrplot)
library(rattle)
library(tidyverse)

##Step4: Create a partition with the training dataset

inTrain<- createDataPartition(TrD2$classe, p=0.7, list=FALSE)
TrainSet <- TrD2[inTrain, ]
TestSet  <- TrD2[-inTrain, ]
dim(TrainSet)
dim(TestSet)

remove identification only variables (columns 1 to 5)

TrainSet <- TrainSet[, -(1:5)]
TestSet  <- TestSet[, -(1:5)]
dim(TrainSet)
dim(TestSet)

##Step5: Correlation

corMatrix <- cor(TrainSet[, -48])
corrplot(corMatrix, order = "FPC", method = "color", type = "lower",title = 'Correlation among Traits', tl.cex = 0.8, tl.col = rgb(0, 0, 0))

##Step6: Prediction Model ###A.- Randon Forest Model fit

set.seed(12345)
Control.RF <- trainControl(method="cv", number=3, verboseIter=FALSE)
FitRF <- train(classe ~ ., data=TrainSet, method="rf",
                          trControl=Control.RF)
FitRF$finalModel

prediction on Test dataset

Predict.RF <- predict(FitRF, newdata=TestSet)
Confusion.Matrix.RF <- confusionMatrix(Predict.RF, TestSet$classe)
Confusion.Matrix.RF

Plot matrix results

plot(Confusion.Matrix.RF$table, col = Confusion.Matrix.RF$byClass, 
     main = paste("Accuracy for Random Forest Model =",
                  round(Confusion.Matrix.RF$overall['Accuracy'], 2)))

###B.- Method: Decision Trees model fit

set.seed(12345)
Fit.DT <- rpart(classe ~ ., data=TrainSet, method="class")
fancyRpartPlot(Fit.DT)

Prediction: Test dataset

Prediction.DT <- predict(Fit.DT, newdata=TestSet, type="class")
Confusion.Matrix.DT <- confusionMatrix(Prediction.DT, TestSet$classe)
Confusion.Matrix.DT

plot(Confusion.Matrix.DT$table, col = Confusion.Matrix.DT$byClass, 
     main = paste("Accuracy for Desicion Tree =",
                  round(Confusion.Matrix.DT$overall['Accuracy'], 2)))

###C.- Method: Generalized Boosted Model model fit

set.seed(12345)
Control.GBM <- trainControl(method = "repeatedcv", number = 5, repeats = 1)
Fit.GBM  <- train(classe ~ ., data=TrainSet, method = "gbm",
                    trControl = Control.GBM, verbose = FALSE)
Fit.GBM$finalModel

Prediction Test dataset

Predict.GBM <- predict(Fit.GBM, newdata=TestSet)
Confusion.Matrix.GBM <- confusionMatrix(Predict.GBM, TestSet$classe)
Confusion.Matrix.GBM

plot matrix results

plot(Confusion.Matrix.GBM$table, col =Confusion.Matrix.GBM$byClass, 
     main = paste("Accuracy for GBM =", round(Confusion.Matrix.GBM$overall['Accuracy'], 2)))

Step 7: Selected Model to the Test Data

Predict.Test <- predict(FitRF, newdata=TtD)
Predict.Test