logistic.R

claimants<-read.csv("/Volumes/Data/Course Content/DS content/Logistic Regression/claimants.csv")
#Finding null values
sum(is.na(claimants))
#Removing null values- na.omit(dataset)
claimants <- na.omit(claimants)
# Logistic Regression
#glm(y~x,family="bin....)
logit<-glm(ATTORNEY ~ factor(CLMSEX) + factor(CLMINSUR) + factor(SEATBELT) 
           + CLMAGE + LOSS,family= "binomial",data=claimants)
summary(logit)
# Confusion Matrix Table
#predict(modelobject,testdataset)
prob=predict(logit,type=c("response"),claimants)
prob

#table(dataframe1,dataframe2) ..to create 2X2 matrix
confusion<-table(prob>0.5,claimants$ATTORNEY)
confusion
# Model Accuracy
#adding diagonal elements in the confusion matrix
Accuracy<-sum(diag(confusion))/sum(confusion)
Accuracy
## ROC Curve

#Extract from the fitted model object the vector of fitted probabilities:
install.packages("ROCR")
install.packages("pROC")
library(ROCR)
library(pROC)
#prediction(probability values from model,y variable)
rocrpred<-prediction(prob,claimants$ATTORNEY)
rocrperf<-performance(rocrpred,'tpr','fpr')
plot(rocrperf,colorize=T,text.adj=c(-0.2,1.7))
#plot(rocrperf)
auc <- auc(claimants$ATTORNEY ~ prob)
auc