Logistic Regression with Confusion Matrix in R

 #Coded by Tawkir Ahmed 

library(ggplot2) #USed for plotting data

library(dplyr) #Used to extract columns in the data

library(rms) #Used to extract p-value from logistic model

library(aod)

library(caret)

# logistic model

rail <- read.csv(choose.files())   #training

test<- read.csv(choose.files())    #testing

labs <- attributes(rail)$OPS #need to think

labs

summary(rail)

# collapse all missing values to NA

rail$OPS <- factor(rail$OPS, levels=c("1", "2", "3", "4", "5"))

# run our regression model

model <- glm(OPS~ y1+ y2+ y3+ y4+ y5, 

               data=rail, family="binomial")

summary(model)

#Confusion matrix

# training

p<- predict(model, rail)

tab<- table(p, rail$OPS)

tab

1-sum(diag(tab))/sum(tab)

#testing

p1<- predict(model, test)

tab1<- table(p1, test$OPS)

tab1

1-sum(diag(tab1))/sum(tab1)

#end

library(e1071)

#data

confusionMatrix(rail$OPS, sample(rail$OPS))

newPrior <- c(.05, .8, .15, 0.5, 0.9)

names(newPrior) <- levels(rail$OPS)

cm <-confusionMatrix(rail$OPS, sample(rail$OPS))

#2

# extract the confusion matrix values as data.frame

cm_d <- as.data.frame(cm$table)

# confusion matrix statistics as data.frame

cm_st <-data.frame(cm$overall)

# round the values

cm_st$cm.overall <- round(cm_st$cm.overall,2)

# here we also have the rounded percentage values

cm_p <- as.data.frame(prop.table(cm$table))

cm_d$Perc <- round(cm_p$Freq*100,2)

#3

library(ggplot2)     # to plot

library(gridExtra)   # to put more

library(grid)        # plot together

# plotting the matrix

cm_d_p <-  ggplot(data = cm_d, aes(x = Prediction , y =  Reference, fill = Freq))+

  geom_tile() +

  geom_text(aes(label = paste("",Freq,",",Perc,"%")), color = 'red', size = 8) +

  theme_light() +

  guides(fill=FALSE) 

# plotting the stats

cm_st_p <-  tableGrob(cm_st)

# all together

grid.arrange(cm_d_p, cm_st_p,nrow = 1, ncol = 2, 

             top=textGrob("Confusion Matrix and Statistics",gp=gpar(fontsize=25,font=1)))

#search confusion matrix plot

###########################################################################