Who is more likely to become a telco churn
Ron Lee
9/26/2018
About this analysis
Actually I tend to make the most of this clear-format data to practice my skills of EDA and model building,since it’s been a while that I haven’t work on data mining project.
Process of analyzing the data
1.1 Load the needed packages
library(tidyverse)
library(corrplot)
library(pROC)
library(caret)
library(RColorBrewer)
library(ggalluvial)
library(xgboost)1.2 Import the data and have a glimpse
churn<-suppressMessages(read_csv("WA_Fn-UseC_-Telco-Customer-Churn.csv"))
glimpse(churn)## Observations: 7,043
## Variables: 21
## $ customerID <chr> "7590-VHVEG", "5575-GNVDE", "3668-QPYBK", "77...
## $ gender <chr> "Female", "Male", "Male", "Male", "Female", "...
## $ SeniorCitizen <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, ...
## $ Partner <chr> "Yes", "No", "No", "No", "No", "No", "No", "N...
## $ Dependents <chr> "No", "No", "No", "No", "No", "No", "Yes", "N...
## $ tenure <int> 1, 34, 2, 45, 2, 8, 22, 10, 28, 62, 13, 16, 5...
## $ PhoneService <chr> "No", "Yes", "Yes", "No", "Yes", "Yes", "Yes"...
## $ MultipleLines <chr> "No phone service", "No", "No", "No phone ser...
## $ InternetService <chr> "DSL", "DSL", "DSL", "DSL", "Fiber optic", "F...
## $ OnlineSecurity <chr> "No", "Yes", "Yes", "Yes", "No", "No", "No", ...
## $ OnlineBackup <chr> "Yes", "No", "Yes", "No", "No", "No", "Yes", ...
## $ DeviceProtection <chr> "No", "Yes", "No", "Yes", "No", "Yes", "No", ...
## $ TechSupport <chr> "No", "No", "No", "Yes", "No", "No", "No", "N...
## $ StreamingTV <chr> "No", "No", "No", "No", "No", "Yes", "Yes", "...
## $ StreamingMovies <chr> "No", "No", "No", "No", "No", "Yes", "No", "N...
## $ Contract <chr> "Month-to-month", "One year", "Month-to-month...
## $ PaperlessBilling <chr> "Yes", "No", "Yes", "No", "Yes", "Yes", "Yes"...
## $ PaymentMethod <chr> "Electronic check", "Mailed check", "Mailed c...
## $ MonthlyCharges <dbl> 29.85, 56.95, 53.85, 42.30, 70.70, 99.65, 89....
## $ TotalCharges <dbl> 29.85, 1889.50, 108.15, 1840.75, 151.65, 820....
## $ Churn <chr> "No", "No", "Yes", "No", "Yes", "Yes", "No", ...There are 7043 observations and 21 variables,the response variable we want to predict is Churn and clearly we can observe from the glimpse info that the class of value of many variables is character, so we want to switch them to factor – the category value in R.
churn<- churn %>% mutate_if(is.character,as.factor) %>% select(-customerID)
churn$SeniorCitizen<-if_else(churn$SeniorCitizen==1,"Yes","No") %>% as.factor()
glimpse(churn)## Observations: 7,043
## Variables: 20
## $ gender <fct> Female, Male, Male, Male, Female, Female, Mal...
## $ SeniorCitizen <fct> No, No, No, No, No, No, No, No, No, No, No, N...
## $ Partner <fct> Yes, No, No, No, No, No, No, No, Yes, No, Yes...
## $ Dependents <fct> No, No, No, No, No, No, Yes, No, No, Yes, Yes...
## $ tenure <int> 1, 34, 2, 45, 2, 8, 22, 10, 28, 62, 13, 16, 5...
## $ PhoneService <fct> No, Yes, Yes, No, Yes, Yes, Yes, No, Yes, Yes...
## $ MultipleLines <fct> No phone service, No, No, No phone service, N...
## $ InternetService <fct> DSL, DSL, DSL, DSL, Fiber optic, Fiber optic,...
## $ OnlineSecurity <fct> No, Yes, Yes, Yes, No, No, No, Yes, No, Yes, ...
## $ OnlineBackup <fct> Yes, No, Yes, No, No, No, Yes, No, No, Yes, N...
## $ DeviceProtection <fct> No, Yes, No, Yes, No, Yes, No, No, Yes, No, N...
## $ TechSupport <fct> No, No, No, Yes, No, No, No, No, Yes, No, No,...
## $ StreamingTV <fct> No, No, No, No, No, Yes, Yes, No, Yes, No, No...
## $ StreamingMovies <fct> No, No, No, No, No, Yes, No, No, Yes, No, No,...
## $ Contract <fct> Month-to-month, One year, Month-to-month, One...
## $ PaperlessBilling <fct> Yes, No, Yes, No, Yes, Yes, Yes, No, Yes, No,...
## $ PaymentMethod <fct> Electronic check, Mailed check, Mailed check,...
## $ MonthlyCharges <dbl> 29.85, 56.95, 53.85, 42.30, 70.70, 99.65, 89....
## $ TotalCharges <dbl> 29.85, 1889.50, 108.15, 1840.75, 151.65, 820....
## $ Churn <fct> No, No, Yes, No, Yes, Yes, No, No, Yes, No, N...1.3 Take a look at the distribution of Churn and the relations between gender and churn
churn %>% ggplot(aes(x=Churn))+geom_bar(fill = c("#FC4E07", "#E7B800"))
churn %>% group_by(Churn,gender) %>% count() %>% ggplot(aes(x=Churn,y=n,fill=gender))+geom_col()+
ggrepel::geom_label_repel(aes(label=n))+labs(title="gender & churn",y="Number")
table(churn$Churn)/nrow(churn)##
## No Yes
## 0.7346301 0.2653699- For most of customers,they tend to stay at the platform
- There is no difference between gender,male and female leave the platform in the same possibility(50%)
1.4 Churn with the only numberic variables
According to the introduction of the data,most of the features are in factor format,only charges related features are quantitive
churn %>% ggplot(aes(x=MonthlyCharges,y=TotalCharges,color=Churn))+geom_point()+scale_color_brewer(palette="Set1")+geom_jitter()+ggforce::facet_zoom(x= MonthlyCharges>75,y= Churn == "Yes",split = T)
* People with less monthly charges tend to leave
* People with more total charges are more likely to stay * So,officials should pay more attention to those who spend more money and provide them with more discount
1.5 Missing Values
It appears that some missing values are mixed in the data which cause some warning messages when I executed the R codes,maybe we should look at it
sapply(churn, function(x){
return(sum(is.na(x)))
})## gender SeniorCitizen Partner Dependents
## 0 0 0 0
## tenure PhoneService MultipleLines InternetService
## 0 0 0 0
## OnlineSecurity OnlineBackup DeviceProtection TechSupport
## 0 0 0 0
## StreamingTV StreamingMovies Contract PaperlessBilling
## 0 0 0 0
## PaymentMethod MonthlyCharges TotalCharges Churn
## 0 0 11 0Only TotalCharges variable has missing ones,replacing 11 NAs with the median value would be a reasonal solution to that.
churn$TotalCharges<-if_else(is.na(churn$TotalCharges),median(churn$TotalCharges,na.rm = T),churn$TotalCharges)1.6 Maybe it’s time to scan all the features and find the highly related variables
p1<-churn %>% group_by(Churn,SeniorCitizen) %>% count() %>% ggplot(aes(x=Churn,y=n,fill=SeniorCitizen))+geom_col()+ggrepel::geom_label_repel(aes(label=n))+labs(title="senior & churn",y="Number")
p2<-churn %>% group_by(Churn,Partner) %>% count() %>% ggplot(aes(x=Churn,y=n,fill=Partner))+geom_col()+ggrepel::geom_label_repel(aes(label=n))+labs(title="Partner & churn",y="Number")
p3<-churn %>% group_by(Churn,Dependents) %>% count() %>% ggplot(aes(x=Churn,y=n,fill=Dependents))+geom_col()+ggrepel::geom_label_repel(aes(label=n))+labs(title="Dependents & churn",y="Number")
p4<-churn %>% group_by(Churn,PhoneService) %>% count() %>% ggplot(aes(x=Churn,y=n,fill=PhoneService))+geom_col()+ggrepel::geom_label_repel(aes(label=n))+labs(title="PhoneService & churn",y="Number")
p5<-churn %>% group_by(Churn,MultipleLines) %>% count() %>% ggplot(aes(x=Churn,y=n,fill=MultipleLines))+geom_col()+ggrepel::geom_label_repel(aes(label=n))+labs(title="MultipleLines & churn",y="Number")
p6<-churn %>% group_by(Churn,InternetService) %>% count() %>% ggplot(aes(x=Churn,y=n,fill=InternetService))+geom_col()+ggrepel::geom_label_repel(aes(label=n))+labs(title="InternetService & churn",y="Number")
p7<-churn %>% group_by(Churn,OnlineSecurity) %>% count() %>% ggplot(aes(x=Churn,y=n,fill=OnlineSecurity))+geom_col()+ggrepel::geom_label_repel(aes(label=n))+labs(title="OnlineSecurity & churn",y="Number")
p8<-churn %>% group_by(Churn,DeviceProtection) %>% count() %>% ggplot(aes(x=Churn,y=n,fill=DeviceProtection))+geom_col()+ggrepel::geom_label_repel(aes(label=n))+labs(title="DeviceProtection & churn",y="Number")
p9<-churn %>% group_by(Churn,TechSupport) %>% count() %>% ggplot(aes(x=Churn,y=n,fill=TechSupport))+geom_col()+ggrepel::geom_label_repel(aes(label=n))+labs(title="TechSupport & churn",y="Number")
p10<-churn %>% group_by(Churn,PaperlessBilling) %>% count() %>% ggplot(aes(x=Churn,y=n,fill=PaperlessBilling))+geom_col()+ggrepel::geom_label_repel(aes(label=n))+labs(title="PaperlessBilling & churn",y="Number")
p11<-churn %>% group_by(Churn,PaymentMethod) %>% count() %>% ggplot(aes(x=Churn,y=n,fill=PaymentMethod))+geom_col()+ggrepel::geom_label_repel(aes(label=n))+labs(title="PaymentMethod & churn",y="Number")
p12<-churn %>% group_by(Churn,Contract) %>% count() %>% ggplot(aes(x=Churn,y=n,fill=Contract))+geom_col()+ggrepel::geom_label_repel(aes(label=n))+labs(title="Contract & churn",y="Number")
p13<-churn %>% group_by(Churn,StreamingMovies) %>% count() %>% ggplot(aes(x=Churn,y=n,fill=StreamingMovies))+geom_col()+ggrepel::geom_label_repel(aes(label=n))+labs(title="StreamingMovies & churn",y="Number")
p14<-churn %>% group_by(Churn,StreamingTV) %>% count() %>% ggplot(aes(x=Churn,y=n,fill=StreamingTV))+geom_col()+ggrepel::geom_label_repel(aes(label=n))+labs(title="StreamingTV & churn",y="Number")
gridExtra::grid.arrange(p1,p2,nrow=2)
gridExtra::grid.arrange(p3,p4,nrow=2)
gridExtra::grid.arrange(p5,p6,nrow=2)
gridExtra::grid.arrange(p7,p8,nrow=2)
gridExtra::grid.arrange(p9,p10,nrow=2)
gridExtra::grid.arrange(p11,p12,nrow=2)
gridExtra::grid.arrange(p13,p14,nrow=2)
I feel sorry to say that I should have defined a function to write less codes and make those codes more readable.If you have a glimpse of those stacked bar plots,customers with specific labels are more likely to become a yes-to-churn person
1.6 One more feature
churn %>% ggplot(aes(x=Churn,y=tenure,fill=Churn))+geom_boxplot()+ggtitle("tenure & churn")+stat_summary(fun.y = mean,geom = "point")+ggthemes::theme_economist_white()+scale_fill_brewer(palette = "RdGy")
1.7 Observe three features at the same time
- Payment methods in real life has strong connections to the amount of charges,but that’s only our hypothesis,pictures would tell you the truth.
- Signing a contract means you have a deal with the platform and you should follow the disciplines.
churn %>% ggplot(aes(x=PaymentMethod,y=MonthlyCharges,fill=Churn))+geom_boxplot()+scale_fill_brewer(palette = "Set3")+ggthemes::theme_solarized()+ggtitle("Monthly Charges & PaymentMethods")
churn %>% ggplot(aes(x=Contract,y=MonthlyCharges,fill=Churn))+geom_boxplot()+scale_fill_brewer(palette = "OrRd")+ggtitle("Monthly Charges & PaymentMethods")+ggthemes::theme_tufte()
churn %>% select_if(is.numeric) %>% cor() %>% corrplot.mixed()
- From the plots drawed above,Tenure and TotalCharges are highly related,it gives the proof that these two features play an vital role in determining which customer would leave.
p1<-churn %>% ggplot(aes(x=TotalCharges,y=Churn,fill=Churn))+ggridges::geom_density_ridges()+ggridges::theme_ridges()+scale_fill_brewer(palette = "Set2")
p2<-churn %>% ggplot(aes(x=MonthlyCharges,y=Churn,fill=Churn))+ggridges::geom_density_ridges()+ggridges::theme_ridges()+scale_fill_brewer(palette = "Set3")
gridExtra::grid.arrange(p1,p2)
- As for TotalCharges,ridge plots show no clear evidence that it can classify whether you churn or not,but monthlycharges plot explictly tells us that clients with higher MonthlyCharges are more likely to churn.
2.1 Feature importances
At first we realize that this churn data contains three quantitive features and the other is categorical,but are they all important?clearly not.To get the preliminary feature importance,we build a simple model with rpart function,which is a well-known CART decision tree algorithm.
rp.model<-rpart::rpart(Churn~.,data = churn,method = "class")
rpart.plot::rpart.plot(rp.model)
varImp(rp.model) %>% rownames_to_column(var="feature") %>% filter(Overall != 0) %>% arrange(desc(Overall)) %>% ggplot(aes(x=reorder(feature,Overall),y=Overall,fill=feature))+geom_col()+coord_flip()+ggtitle("Feature Importances")+theme_bw()+scale_fill_brewer(palette = "Paired")+theme(legend.position = "none")
- The importances of feature are shown in bar,the three numerical features are both essential predictors for churn.
churn$TechSupport<-gsub("No internet service","No",churn$TechSupport)
churn$OnlineSecurity<-gsub("No internet service","No",churn$OnlineSecurity)
churn$MultipleLines<-gsub("No phone service","No",churn$MultipleLines)
churn$OnlineBackup<-gsub("No internet service","No",churn$OnlineBackup)
churn$DeviceProtection<-gsub("No internet service","No",churn$DeviceProtection)
churn$StreamingTV<-gsub("No internet service","No",churn$StreamingTV)
churn$StreamingMovies<-gsub("No internet service","No",churn$StreamingMovies)
churn$PaymentMethod<-as.factor(sapply(churn$PaymentMethod, function(x){x=unlist(str_split(x," "))[1]}))
churn %>% group_by(OnlineSecurity,Contract,TechSupport,InternetService,PaymentMethod,Churn) %>% count() %>% rename(Freq=n) %>% ggplot(aes(axis1=OnlineSecurity,axis2=Contract,axis3=TechSupport,axis4=InternetService,axis5=PaymentMethod,y=Freq))+scale_x_discrete(limits = c("OnlineSecurity","Contract","TechSupport","InternetService","PaymentMethod"), expand = c(.1, .05))+geom_alluvium(aes(fill=Churn))+geom_stratum()+theme_minimal()+geom_text(stat = "stratum",label.strata = TRUE)+coord_flip()
2.2 Clean the data for model fitting
- scale all those numberical features
- transform all categorical features into numberical using model.matrix function
- randomly split the data into two parts(70% for training,30% for validation)
churn<-churn %>% select(-gender)
churn[,c("MonthlyCharges","TotalCharges")]<-sapply(churn[,c("MonthlyCharges","TotalCharges")],function(x){
return((x-min(x))/max(x))
})
trans.churn<-as.data.frame(sapply(churn %>% select(-c(MonthlyCharges,TotalCharges,tenure,Churn)), function(x){
return(model.matrix(~x-1,data = churn %>% select(-c(MonthlyCharges,TotalCharges,tenure,Churn))))
})) %>% bind_cols(churn %>% select(MonthlyCharges,TotalCharges,tenure,Churn))
glimpse(trans.churn)## Observations: 7,043
## Variables: 38
## $ SeniorCitizen.xNo <dbl> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
## $ SeniorCitizen.xYes <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, ...
## $ Partner.xNo <dbl> 0, 1, 1, 1, 1, 1, 1, 1, 0, 1, 0, ...
## $ Partner.xYes <dbl> 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, ...
## $ Dependents.xNo <dbl> 1, 1, 1, 1, 1, 1, 0, 1, 1, 0, 0, ...
## $ Dependents.xYes <dbl> 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 1, ...
## $ PhoneService.xNo <dbl> 1, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, ...
## $ PhoneService.xYes <dbl> 0, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, ...
## $ MultipleLines.xNo <dbl> 1, 1, 1, 1, 1, 0, 0, 1, 0, 1, 1, ...
## $ MultipleLines.xYes <dbl> 0, 0, 0, 0, 0, 1, 1, 0, 1, 0, 0, ...
## $ InternetService.xDSL <dbl> 1, 1, 1, 1, 0, 0, 0, 1, 0, 1, 1, ...
## $ InternetService.xFiber.optic <dbl> 0, 0, 0, 0, 1, 1, 1, 0, 1, 0, 0, ...
## $ InternetService.xNo <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, ...
## $ OnlineSecurity.xNo <dbl> 1, 0, 0, 0, 1, 1, 1, 0, 1, 0, 0, ...
## $ OnlineSecurity.xYes <dbl> 0, 1, 1, 1, 0, 0, 0, 1, 0, 1, 1, ...
## $ OnlineBackup.xNo <dbl> 0, 1, 0, 1, 1, 1, 0, 1, 1, 0, 1, ...
## $ OnlineBackup.xYes <dbl> 1, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, ...
## $ DeviceProtection.xNo <dbl> 1, 0, 1, 0, 1, 0, 1, 1, 0, 1, 1, ...
## $ DeviceProtection.xYes <dbl> 0, 1, 0, 1, 0, 1, 0, 0, 1, 0, 0, ...
## $ TechSupport.xNo <dbl> 1, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, ...
## $ TechSupport.xYes <dbl> 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, ...
## $ StreamingTV.xNo <dbl> 1, 1, 1, 1, 1, 0, 0, 1, 0, 1, 1, ...
## $ StreamingTV.xYes <dbl> 0, 0, 0, 0, 0, 1, 1, 0, 1, 0, 0, ...
## $ StreamingMovies.xNo <dbl> 1, 1, 1, 1, 1, 0, 1, 1, 0, 1, 1, ...
## $ StreamingMovies.xYes <dbl> 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, ...
## $ Contract.xMonth.to.month <dbl> 1, 0, 1, 0, 1, 1, 1, 1, 1, 0, 1, ...
## $ Contract.xOne.year <dbl> 0, 1, 0, 1, 0, 0, 0, 0, 0, 1, 0, ...
## $ Contract.xTwo.year <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, ...
## $ PaperlessBilling.xNo <dbl> 0, 1, 0, 1, 0, 0, 0, 1, 0, 1, 0, ...
## $ PaperlessBilling.xYes <dbl> 1, 0, 1, 0, 1, 1, 1, 0, 1, 0, 1, ...
## $ PaymentMethod.xBank <dbl> 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, ...
## $ PaymentMethod.xCredit <dbl> 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, ...
## $ PaymentMethod.xElectronic <dbl> 1, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, ...
## $ PaymentMethod.xMailed <dbl> 0, 1, 1, 0, 0, 0, 0, 1, 0, 0, 1, ...
## $ MonthlyCharges <dbl> 0.097684211, 0.325894737, 0.29978...
## $ TotalCharges <dbl> 0.001272338, 0.215399318, 0.01028...
## $ tenure <int> 1, 34, 2, 45, 2, 8, 22, 10, 28, 6...
## $ Churn <fct> No, No, Yes, No, Yes, Yes, No, No... ind<-sample(2,nrow(trans.churn),replace = T,prob = c(0.7,0.3))
train.churn<-trans.churn[ind==1,]
test.churn<-trans.churn[ind==2,]3.1 Model building:Random Forest
rf.model<-train(Churn~.,data = train.churn,method="rf",tuneGrid=expand.grid(mtry=c(2,3)),importance=T,metric = "ROC",ntree=500,trControl=trainControl(classProbs = T))
rf.model## Random Forest
##
## 4874 samples
## 37 predictor
## 2 classes: 'No', 'Yes'
##
## No pre-processing
## Resampling: Bootstrapped (25 reps)
## Summary of sample sizes: 4874, 4874, 4874, 4874, 4874, 4874, ...
## Resampling results across tuning parameters:
##
## mtry Accuracy Kappa
## 2 0.7927319 0.4068291
## 3 0.7926069 0.4314928
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was mtry = 2. varImp(rf.model) %>% plot(main="Features of random forest")
rf.pred<-predict(rf.model,test.churn)
confusionMatrix(rf.pred,reference = test.churn$Churn,mode="everything",positive = "Yes")## Confusion Matrix and Statistics
##
## Reference
## Prediction No Yes
## No 1492 305
## Yes 121 251
##
## Accuracy : 0.8036
## 95% CI : (0.7862, 0.8201)
## No Information Rate : 0.7437
## P-Value [Acc > NIR] : 2.889e-11
##
## Kappa : 0.4222
## Mcnemar's Test P-Value : < 2.2e-16
##
## Sensitivity : 0.4514
## Specificity : 0.9250
## Pos Pred Value : 0.6747
## Neg Pred Value : 0.8303
## Precision : 0.6747
## Recall : 0.4514
## F1 : 0.5409
## Prevalence : 0.2563
## Detection Rate : 0.1157
## Detection Prevalence : 0.1715
## Balanced Accuracy : 0.6882
##
## 'Positive' Class : Yes
## rf.roc<-roc(test.churn$Churn,predict(rf.model,test.churn,type = "prob")[,2],levels = levels(test.churn$Churn))3.2 model building:XGboost
xg.train<-xgb.DMatrix(train.churn %>% select(-Churn) %>% as.matrix(),label=train.churn$Churn)
xg.test<-xgb.DMatrix(as.matrix(test.churn %>% select(-Churn)),label=test.churn$Churn)
xg.model<-xgb.train(params = list(max_depth=7,eval_metric="error"),data = xg.train,watchlist = list(validation=xg.test),nrounds = 500,print_every_n = 50,early_stopping_rounds = 200)## [1] validation-error:-0.256339
## Will train until validation_error hasn't improved in 200 rounds.
##
## [51] validation-error:-0.256339
## [101] validation-error:-0.256339
## [151] validation-error:-0.256339
## [201] validation-error:-0.256339
## Stopping. Best iteration:
## [1] validation-error:-0.256339 xg.pred<-predict(xg.model,xg.test)
xg.roc<-roc(test.churn$Churn,xg.pred)
confusionMatrix(as.factor(ifelse(xg.pred >= 0.5,"Yes","No")),reference = test.churn$Churn,positive = "Yes",mode = "everything")## Confusion Matrix and Statistics
##
## Reference
## Prediction No Yes
## No 0 0
## Yes 1613 556
##
## Accuracy : 0.2563
## 95% CI : (0.2381, 0.2753)
## No Information Rate : 0.7437
## P-Value [Acc > NIR] : 1
##
## Kappa : 0
## Mcnemar's Test P-Value : <2e-16
##
## Sensitivity : 1.0000
## Specificity : 0.0000
## Pos Pred Value : 0.2563
## Neg Pred Value : NaN
## Precision : 0.2563
## Recall : 1.0000
## F1 : 0.4081
## Prevalence : 0.2563
## Detection Rate : 0.2563
## Detection Prevalence : 1.0000
## Balanced Accuracy : 0.5000
##
## 'Positive' Class : Yes
## xgb.plot.importance(xgb.importance(colnames(xg.train),model=xg.model),xlab = "Relative importance")
3.3 model building:GLM
glm.model<-glm(Churn~.,data = train.churn,family = "binomial")
glm.model %>% summary()##
## Call:
## glm(formula = Churn ~ ., family = "binomial", data = train.churn)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.9571 -0.6828 -0.3016 0.7377 3.3010
##
## Coefficients: (15 not defined because of singularities)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -1.052510 1.814597 -0.580 0.561898
## SeniorCitizen.xNo -0.367257 0.100976 -3.637 0.000276 ***
## SeniorCitizen.xYes NA NA NA NA
## Partner.xNo 0.048117 0.092538 0.520 0.603085
## Partner.xYes NA NA NA NA
## Dependents.xNo 0.115826 0.107094 1.082 0.279460
## Dependents.xYes NA NA NA NA
## PhoneService.xNo -0.046453 0.781103 -0.059 0.952577
## PhoneService.xYes NA NA NA NA
## MultipleLines.xNo -0.395069 0.212506 -1.859 0.063014 .
## MultipleLines.xYes NA NA NA NA
## InternetService.xDSL 1.531491 0.972364 1.575 0.115252
## InternetService.xFiber.optic 2.859383 1.920519 1.489 0.136524
## InternetService.xNo NA NA NA NA
## OnlineSecurity.xNo 0.259282 0.214765 1.207 0.227322
## OnlineSecurity.xYes NA NA NA NA
## OnlineBackup.xNo 0.108721 0.212081 0.513 0.608203
## OnlineBackup.xYes NA NA NA NA
## DeviceProtection.xNo -0.031567 0.211615 -0.149 0.881418
## DeviceProtection.xYes NA NA NA NA
## TechSupport.xNo 0.168748 0.218063 0.774 0.439020
## TechSupport.xYes NA NA NA NA
## StreamingTV.xNo -0.443858 0.393187 -1.129 0.258951
## StreamingTV.xYes NA NA NA NA
## StreamingMovies.xNo -0.426316 0.394729 -1.080 0.280133
## StreamingMovies.xYes NA NA NA NA
## Contract.xMonth.to.month 1.130290 0.200639 5.633 1.77e-08 ***
## Contract.xOne.year 0.427822 0.202842 2.109 0.034933 *
## Contract.xTwo.year NA NA NA NA
## PaperlessBilling.xNo -0.368835 0.089630 -4.115 3.87e-05 ***
## PaperlessBilling.xYes NA NA NA NA
## PaymentMethod.xBank 0.071515 0.136275 0.525 0.599731
## PaymentMethod.xCredit -0.084686 0.138396 -0.612 0.540599
## PaymentMethod.xElectronic 0.340802 0.116272 2.931 0.003378 **
## PaymentMethod.xMailed NA NA NA NA
## MonthlyCharges -3.276914 4.545037 -0.721 0.470917
## TotalCharges 2.536015 0.710147 3.571 0.000355 ***
## tenure -0.056685 0.007117 -7.965 1.65e-15 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 5679.7 on 4873 degrees of freedom
## Residual deviance: 4103.1 on 4851 degrees of freedom
## AIC: 4149.1
##
## Number of Fisher Scoring iterations: 6 glm.model_1<-MASS::stepAIC(glm.model,direction = "both")## Start: AIC=4149.07
## Churn ~ SeniorCitizen.xNo + SeniorCitizen.xYes + Partner.xNo +
## Partner.xYes + Dependents.xNo + Dependents.xYes + PhoneService.xNo +
## PhoneService.xYes + MultipleLines.xNo + MultipleLines.xYes +
## InternetService.xDSL + InternetService.xFiber.optic + InternetService.xNo +
## OnlineSecurity.xNo + OnlineSecurity.xYes + OnlineBackup.xNo +
## OnlineBackup.xYes + DeviceProtection.xNo + DeviceProtection.xYes +
## TechSupport.xNo + TechSupport.xYes + StreamingTV.xNo + StreamingTV.xYes +
## StreamingMovies.xNo + StreamingMovies.xYes + Contract.xMonth.to.month +
## Contract.xOne.year + Contract.xTwo.year + PaperlessBilling.xNo +
## PaperlessBilling.xYes + PaymentMethod.xBank + PaymentMethod.xCredit +
## PaymentMethod.xElectronic + PaymentMethod.xMailed + MonthlyCharges +
## TotalCharges + tenure
##
##
## Step: AIC=4149.07
## Churn ~ SeniorCitizen.xNo + SeniorCitizen.xYes + Partner.xNo +
## Partner.xYes + Dependents.xNo + Dependents.xYes + PhoneService.xNo +
## PhoneService.xYes + MultipleLines.xNo + MultipleLines.xYes +
## InternetService.xDSL + InternetService.xFiber.optic + InternetService.xNo +
## OnlineSecurity.xNo + OnlineSecurity.xYes + OnlineBackup.xNo +
## OnlineBackup.xYes + DeviceProtection.xNo + DeviceProtection.xYes +
## TechSupport.xNo + TechSupport.xYes + StreamingTV.xNo + StreamingTV.xYes +
## StreamingMovies.xNo + StreamingMovies.xYes + Contract.xMonth.to.month +
## Contract.xOne.year + Contract.xTwo.year + PaperlessBilling.xNo +
## PaperlessBilling.xYes + PaymentMethod.xBank + PaymentMethod.xCredit +
## PaymentMethod.xElectronic + MonthlyCharges + TotalCharges +
## tenure
##
##
## Step: AIC=4149.07
## Churn ~ SeniorCitizen.xNo + SeniorCitizen.xYes + Partner.xNo +
## Partner.xYes + Dependents.xNo + Dependents.xYes + PhoneService.xNo +
## PhoneService.xYes + MultipleLines.xNo + MultipleLines.xYes +
## InternetService.xDSL + InternetService.xFiber.optic + InternetService.xNo +
## OnlineSecurity.xNo + OnlineSecurity.xYes + OnlineBackup.xNo +
## OnlineBackup.xYes + DeviceProtection.xNo + DeviceProtection.xYes +
## TechSupport.xNo + TechSupport.xYes + StreamingTV.xNo + StreamingTV.xYes +
## StreamingMovies.xNo + StreamingMovies.xYes + Contract.xMonth.to.month +
## Contract.xOne.year + Contract.xTwo.year + PaperlessBilling.xNo +
## PaymentMethod.xBank + PaymentMethod.xCredit + PaymentMethod.xElectronic +
## MonthlyCharges + TotalCharges + tenure
##
##
## Step: AIC=4149.07
## Churn ~ SeniorCitizen.xNo + SeniorCitizen.xYes + Partner.xNo +
## Partner.xYes + Dependents.xNo + Dependents.xYes + PhoneService.xNo +
## PhoneService.xYes + MultipleLines.xNo + MultipleLines.xYes +
## InternetService.xDSL + InternetService.xFiber.optic + InternetService.xNo +
## OnlineSecurity.xNo + OnlineSecurity.xYes + OnlineBackup.xNo +
## OnlineBackup.xYes + DeviceProtection.xNo + DeviceProtection.xYes +
## TechSupport.xNo + TechSupport.xYes + StreamingTV.xNo + StreamingTV.xYes +
## StreamingMovies.xNo + StreamingMovies.xYes + Contract.xMonth.to.month +
## Contract.xOne.year + PaperlessBilling.xNo + PaymentMethod.xBank +
## PaymentMethod.xCredit + PaymentMethod.xElectronic + MonthlyCharges +
## TotalCharges + tenure
##
##
## Step: AIC=4149.07
## Churn ~ SeniorCitizen.xNo + SeniorCitizen.xYes + Partner.xNo +
## Partner.xYes + Dependents.xNo + Dependents.xYes + PhoneService.xNo +
## PhoneService.xYes + MultipleLines.xNo + MultipleLines.xYes +
## InternetService.xDSL + InternetService.xFiber.optic + InternetService.xNo +
## OnlineSecurity.xNo + OnlineSecurity.xYes + OnlineBackup.xNo +
## OnlineBackup.xYes + DeviceProtection.xNo + DeviceProtection.xYes +
## TechSupport.xNo + TechSupport.xYes + StreamingTV.xNo + StreamingTV.xYes +
## StreamingMovies.xNo + Contract.xMonth.to.month + Contract.xOne.year +
## PaperlessBilling.xNo + PaymentMethod.xBank + PaymentMethod.xCredit +
## PaymentMethod.xElectronic + MonthlyCharges + TotalCharges +
## tenure
##
##
## Step: AIC=4149.07
## Churn ~ SeniorCitizen.xNo + SeniorCitizen.xYes + Partner.xNo +
## Partner.xYes + Dependents.xNo + Dependents.xYes + PhoneService.xNo +
## PhoneService.xYes + MultipleLines.xNo + MultipleLines.xYes +
## InternetService.xDSL + InternetService.xFiber.optic + InternetService.xNo +
## OnlineSecurity.xNo + OnlineSecurity.xYes + OnlineBackup.xNo +
## OnlineBackup.xYes + DeviceProtection.xNo + DeviceProtection.xYes +
## TechSupport.xNo + TechSupport.xYes + StreamingTV.xNo + StreamingMovies.xNo +
## Contract.xMonth.to.month + Contract.xOne.year + PaperlessBilling.xNo +
## PaymentMethod.xBank + PaymentMethod.xCredit + PaymentMethod.xElectronic +
## MonthlyCharges + TotalCharges + tenure
##
##
## Step: AIC=4149.07
## Churn ~ SeniorCitizen.xNo + SeniorCitizen.xYes + Partner.xNo +
## Partner.xYes + Dependents.xNo + Dependents.xYes + PhoneService.xNo +
## PhoneService.xYes + MultipleLines.xNo + MultipleLines.xYes +
## InternetService.xDSL + InternetService.xFiber.optic + InternetService.xNo +
## OnlineSecurity.xNo + OnlineSecurity.xYes + OnlineBackup.xNo +
## OnlineBackup.xYes + DeviceProtection.xNo + DeviceProtection.xYes +
## TechSupport.xNo + StreamingTV.xNo + StreamingMovies.xNo +
## Contract.xMonth.to.month + Contract.xOne.year + PaperlessBilling.xNo +
## PaymentMethod.xBank + PaymentMethod.xCredit + PaymentMethod.xElectronic +
## MonthlyCharges + TotalCharges + tenure
##
##
## Step: AIC=4149.07
## Churn ~ SeniorCitizen.xNo + SeniorCitizen.xYes + Partner.xNo +
## Partner.xYes + Dependents.xNo + Dependents.xYes + PhoneService.xNo +
## PhoneService.xYes + MultipleLines.xNo + MultipleLines.xYes +
## InternetService.xDSL + InternetService.xFiber.optic + InternetService.xNo +
## OnlineSecurity.xNo + OnlineSecurity.xYes + OnlineBackup.xNo +
## OnlineBackup.xYes + DeviceProtection.xNo + TechSupport.xNo +
## StreamingTV.xNo + StreamingMovies.xNo + Contract.xMonth.to.month +
## Contract.xOne.year + PaperlessBilling.xNo + PaymentMethod.xBank +
## PaymentMethod.xCredit + PaymentMethod.xElectronic + MonthlyCharges +
## TotalCharges + tenure
##
##
## Step: AIC=4149.07
## Churn ~ SeniorCitizen.xNo + SeniorCitizen.xYes + Partner.xNo +
## Partner.xYes + Dependents.xNo + Dependents.xYes + PhoneService.xNo +
## PhoneService.xYes + MultipleLines.xNo + MultipleLines.xYes +
## InternetService.xDSL + InternetService.xFiber.optic + InternetService.xNo +
## OnlineSecurity.xNo + OnlineSecurity.xYes + OnlineBackup.xNo +
## DeviceProtection.xNo + TechSupport.xNo + StreamingTV.xNo +
## StreamingMovies.xNo + Contract.xMonth.to.month + Contract.xOne.year +
## PaperlessBilling.xNo + PaymentMethod.xBank + PaymentMethod.xCredit +
## PaymentMethod.xElectronic + MonthlyCharges + TotalCharges +
## tenure
##
##
## Step: AIC=4149.07
## Churn ~ SeniorCitizen.xNo + SeniorCitizen.xYes + Partner.xNo +
## Partner.xYes + Dependents.xNo + Dependents.xYes + PhoneService.xNo +
## PhoneService.xYes + MultipleLines.xNo + MultipleLines.xYes +
## InternetService.xDSL + InternetService.xFiber.optic + InternetService.xNo +
## OnlineSecurity.xNo + OnlineBackup.xNo + DeviceProtection.xNo +
## TechSupport.xNo + StreamingTV.xNo + StreamingMovies.xNo +
## Contract.xMonth.to.month + Contract.xOne.year + PaperlessBilling.xNo +
## PaymentMethod.xBank + PaymentMethod.xCredit + PaymentMethod.xElectronic +
## MonthlyCharges + TotalCharges + tenure
##
##
## Step: AIC=4149.07
## Churn ~ SeniorCitizen.xNo + SeniorCitizen.xYes + Partner.xNo +
## Partner.xYes + Dependents.xNo + Dependents.xYes + PhoneService.xNo +
## PhoneService.xYes + MultipleLines.xNo + MultipleLines.xYes +
## InternetService.xDSL + InternetService.xFiber.optic + OnlineSecurity.xNo +
## OnlineBackup.xNo + DeviceProtection.xNo + TechSupport.xNo +
## StreamingTV.xNo + StreamingMovies.xNo + Contract.xMonth.to.month +
## Contract.xOne.year + PaperlessBilling.xNo + PaymentMethod.xBank +
## PaymentMethod.xCredit + PaymentMethod.xElectronic + MonthlyCharges +
## TotalCharges + tenure
##
##
## Step: AIC=4149.07
## Churn ~ SeniorCitizen.xNo + SeniorCitizen.xYes + Partner.xNo +
## Partner.xYes + Dependents.xNo + Dependents.xYes + PhoneService.xNo +
## PhoneService.xYes + MultipleLines.xNo + InternetService.xDSL +
## InternetService.xFiber.optic + OnlineSecurity.xNo + OnlineBackup.xNo +
## DeviceProtection.xNo + TechSupport.xNo + StreamingTV.xNo +
## StreamingMovies.xNo + Contract.xMonth.to.month + Contract.xOne.year +
## PaperlessBilling.xNo + PaymentMethod.xBank + PaymentMethod.xCredit +
## PaymentMethod.xElectronic + MonthlyCharges + TotalCharges +
## tenure
##
##
## Step: AIC=4149.07
## Churn ~ SeniorCitizen.xNo + SeniorCitizen.xYes + Partner.xNo +
## Partner.xYes + Dependents.xNo + Dependents.xYes + PhoneService.xNo +
## MultipleLines.xNo + InternetService.xDSL + InternetService.xFiber.optic +
## OnlineSecurity.xNo + OnlineBackup.xNo + DeviceProtection.xNo +
## TechSupport.xNo + StreamingTV.xNo + StreamingMovies.xNo +
## Contract.xMonth.to.month + Contract.xOne.year + PaperlessBilling.xNo +
## PaymentMethod.xBank + PaymentMethod.xCredit + PaymentMethod.xElectronic +
## MonthlyCharges + TotalCharges + tenure
##
##
## Step: AIC=4149.07
## Churn ~ SeniorCitizen.xNo + SeniorCitizen.xYes + Partner.xNo +
## Partner.xYes + Dependents.xNo + PhoneService.xNo + MultipleLines.xNo +
## InternetService.xDSL + InternetService.xFiber.optic + OnlineSecurity.xNo +
## OnlineBackup.xNo + DeviceProtection.xNo + TechSupport.xNo +
## StreamingTV.xNo + StreamingMovies.xNo + Contract.xMonth.to.month +
## Contract.xOne.year + PaperlessBilling.xNo + PaymentMethod.xBank +
## PaymentMethod.xCredit + PaymentMethod.xElectronic + MonthlyCharges +
## TotalCharges + tenure
##
##
## Step: AIC=4149.07
## Churn ~ SeniorCitizen.xNo + SeniorCitizen.xYes + Partner.xNo +
## Dependents.xNo + PhoneService.xNo + MultipleLines.xNo + InternetService.xDSL +
## InternetService.xFiber.optic + OnlineSecurity.xNo + OnlineBackup.xNo +
## DeviceProtection.xNo + TechSupport.xNo + StreamingTV.xNo +
## StreamingMovies.xNo + Contract.xMonth.to.month + Contract.xOne.year +
## PaperlessBilling.xNo + PaymentMethod.xBank + PaymentMethod.xCredit +
## PaymentMethod.xElectronic + MonthlyCharges + TotalCharges +
## tenure
##
##
## Step: AIC=4149.07
## Churn ~ SeniorCitizen.xNo + Partner.xNo + Dependents.xNo + PhoneService.xNo +
## MultipleLines.xNo + InternetService.xDSL + InternetService.xFiber.optic +
## OnlineSecurity.xNo + OnlineBackup.xNo + DeviceProtection.xNo +
## TechSupport.xNo + StreamingTV.xNo + StreamingMovies.xNo +
## Contract.xMonth.to.month + Contract.xOne.year + PaperlessBilling.xNo +
## PaymentMethod.xBank + PaymentMethod.xCredit + PaymentMethod.xElectronic +
## MonthlyCharges + TotalCharges + tenure
##
## Df Deviance AIC
## - PhoneService.xNo 1 4103.1 4147.1
## - DeviceProtection.xNo 1 4103.1 4147.1
## - OnlineBackup.xNo 1 4103.3 4147.3
## - Partner.xNo 1 4103.3 4147.3
## - PaymentMethod.xBank 1 4103.3 4147.3
## - PaymentMethod.xCredit 1 4103.4 4147.4
## - MonthlyCharges 1 4103.6 4147.6
## - TechSupport.xNo 1 4103.7 4147.7
## - StreamingMovies.xNo 1 4104.2 4148.2
## - Dependents.xNo 1 4104.2 4148.2
## - StreamingTV.xNo 1 4104.3 4148.3
## - OnlineSecurity.xNo 1 4104.5 4148.5
## <none> 4103.1 4149.1
## - InternetService.xFiber.optic 1 4105.3 4149.3
## - InternetService.xDSL 1 4105.5 4149.5
## - MultipleLines.xNo 1 4106.5 4150.5
## - Contract.xOne.year 1 4107.7 4151.7
## - PaymentMethod.xElectronic 1 4111.7 4155.7
## - SeniorCitizen.xNo 1 4116.3 4160.3
## - TotalCharges 1 4116.4 4160.4
## - PaperlessBilling.xNo 1 4120.1 4164.1
## - Contract.xMonth.to.month 1 4139.4 4183.4
## - tenure 1 4177.0 4221.0
##
## Step: AIC=4147.07
## Churn ~ SeniorCitizen.xNo + Partner.xNo + Dependents.xNo + MultipleLines.xNo +
## InternetService.xDSL + InternetService.xFiber.optic + OnlineSecurity.xNo +
## OnlineBackup.xNo + DeviceProtection.xNo + TechSupport.xNo +
## StreamingTV.xNo + StreamingMovies.xNo + Contract.xMonth.to.month +
## Contract.xOne.year + PaperlessBilling.xNo + PaymentMethod.xBank +
## PaymentMethod.xCredit + PaymentMethod.xElectronic + MonthlyCharges +
## TotalCharges + tenure
##
## Df Deviance AIC
## - DeviceProtection.xNo 1 4103.1 4145.1
## - Partner.xNo 1 4103.3 4145.3
## - PaymentMethod.xBank 1 4103.3 4145.3
## - PaymentMethod.xCredit 1 4103.4 4145.4
## - Dependents.xNo 1 4104.2 4146.2
## - OnlineBackup.xNo 1 4104.6 4146.6
## <none> 4103.1 4147.1
## - TechSupport.xNo 1 4105.6 4147.6
## + PhoneService.xNo 1 4103.1 4149.1
## + PhoneService.xYes 1 4103.1 4149.1
## - Contract.xOne.year 1 4107.7 4149.7
## - OnlineSecurity.xNo 1 4109.3 4151.3
## - PaymentMethod.xElectronic 1 4111.7 4153.7
## - MonthlyCharges 1 4113.8 4155.8
## - StreamingMovies.xNo 1 4114.7 4156.7
## - StreamingTV.xNo 1 4115.5 4157.5
## - MultipleLines.xNo 1 4115.7 4157.7
## - SeniorCitizen.xNo 1 4116.3 4158.3
## - TotalCharges 1 4116.4 4158.4
## - PaperlessBilling.xNo 1 4120.1 4162.1
## - Contract.xMonth.to.month 1 4139.4 4181.4
## - InternetService.xFiber.optic 1 4146.5 4188.5
## - InternetService.xDSL 1 4150.7 4192.7
## - tenure 1 4177.0 4219.0
##
## Step: AIC=4145.11
## Churn ~ SeniorCitizen.xNo + Partner.xNo + Dependents.xNo + MultipleLines.xNo +
## InternetService.xDSL + InternetService.xFiber.optic + OnlineSecurity.xNo +
## OnlineBackup.xNo + TechSupport.xNo + StreamingTV.xNo + StreamingMovies.xNo +
## Contract.xMonth.to.month + Contract.xOne.year + PaperlessBilling.xNo +
## PaymentMethod.xBank + PaymentMethod.xCredit + PaymentMethod.xElectronic +
## MonthlyCharges + TotalCharges + tenure
##
## Df Deviance AIC
## - Partner.xNo 1 4103.4 4143.4
## - PaymentMethod.xBank 1 4103.4 4143.4
## - PaymentMethod.xCredit 1 4103.5 4143.5
## - Dependents.xNo 1 4104.3 4144.3
## - OnlineBackup.xNo 1 4104.7 4144.7
## <none> 4103.1 4145.1
## - TechSupport.xNo 1 4105.8 4145.8
## + DeviceProtection.xNo 1 4103.1 4147.1
## + DeviceProtection.xYes 1 4103.1 4147.1
## + PhoneService.xNo 1 4103.1 4147.1
## + PhoneService.xYes 1 4103.1 4147.1
## - Contract.xOne.year 1 4107.7 4147.7
## - OnlineSecurity.xNo 1 4109.6 4149.6
## - PaymentMethod.xElectronic 1 4111.7 4151.7
## - MonthlyCharges 1 4114.7 4154.7
## - StreamingMovies.xNo 1 4114.8 4154.8
## - StreamingTV.xNo 1 4115.6 4155.6
## - MultipleLines.xNo 1 4116.0 4156.0
## - SeniorCitizen.xNo 1 4116.4 4156.4
## - TotalCharges 1 4116.5 4156.5
## - PaperlessBilling.xNo 1 4120.2 4160.2
## - Contract.xMonth.to.month 1 4139.5 4179.5
## - InternetService.xFiber.optic 1 4149.5 4189.5
## - InternetService.xDSL 1 4151.5 4191.5
## - tenure 1 4177.0 4217.0
##
## Step: AIC=4143.38
## Churn ~ SeniorCitizen.xNo + Dependents.xNo + MultipleLines.xNo +
## InternetService.xDSL + InternetService.xFiber.optic + OnlineSecurity.xNo +
## OnlineBackup.xNo + TechSupport.xNo + StreamingTV.xNo + StreamingMovies.xNo +
## Contract.xMonth.to.month + Contract.xOne.year + PaperlessBilling.xNo +
## PaymentMethod.xBank + PaymentMethod.xCredit + PaymentMethod.xElectronic +
## MonthlyCharges + TotalCharges + tenure
##
## Df Deviance AIC
## - PaymentMethod.xBank 1 4103.6 4141.6
## - PaymentMethod.xCredit 1 4103.8 4141.8
## - OnlineBackup.xNo 1 4104.9 4142.9
## <none> 4103.4 4143.4
## - Dependents.xNo 1 4105.4 4143.4
## - TechSupport.xNo 1 4106.1 4144.1
## + Partner.xNo 1 4103.1 4145.1
## + Partner.xYes 1 4103.1 4145.1
## + DeviceProtection.xNo 1 4103.3 4145.3
## + DeviceProtection.xYes 1 4103.3 4145.3
## + PhoneService.xNo 1 4103.4 4145.4
## + PhoneService.xYes 1 4103.4 4145.4
## - Contract.xOne.year 1 4108.0 4146.0
## - OnlineSecurity.xNo 1 4109.9 4147.9
## - PaymentMethod.xElectronic 1 4111.9 4149.9
## - MonthlyCharges 1 4115.0 4153.0
## - StreamingMovies.xNo 1 4115.0 4153.0
## - StreamingTV.xNo 1 4115.8 4153.8
## - MultipleLines.xNo 1 4116.2 4154.2
## - SeniorCitizen.xNo 1 4116.4 4154.4
## - TotalCharges 1 4116.8 4154.8
## - PaperlessBilling.xNo 1 4120.4 4158.4
## - Contract.xMonth.to.month 1 4139.7 4177.7
## - InternetService.xFiber.optic 1 4149.7 4187.7
## - InternetService.xDSL 1 4151.8 4189.8
## - tenure 1 4178.7 4216.7
##
## Step: AIC=4141.63
## Churn ~ SeniorCitizen.xNo + Dependents.xNo + MultipleLines.xNo +
## InternetService.xDSL + InternetService.xFiber.optic + OnlineSecurity.xNo +
## OnlineBackup.xNo + TechSupport.xNo + StreamingTV.xNo + StreamingMovies.xNo +
## Contract.xMonth.to.month + Contract.xOne.year + PaperlessBilling.xNo +
## PaymentMethod.xCredit + PaymentMethod.xElectronic + MonthlyCharges +
## TotalCharges + tenure
##
## Df Deviance AIC
## - PaymentMethod.xCredit 1 4104.7 4140.7
## - OnlineBackup.xNo 1 4105.2 4141.2
## <none> 4103.6 4141.6
## - Dependents.xNo 1 4105.7 4141.7
## - TechSupport.xNo 1 4106.4 4142.4
## + PaymentMethod.xBank 1 4103.4 4143.4
## + PaymentMethod.xMailed 1 4103.4 4143.4
## + Partner.xNo 1 4103.4 4143.4
## + Partner.xYes 1 4103.4 4143.4
## + DeviceProtection.xNo 1 4103.6 4143.6
## + DeviceProtection.xYes 1 4103.6 4143.6
## + PhoneService.xNo 1 4103.6 4143.6
## + PhoneService.xYes 1 4103.6 4143.6
## - Contract.xOne.year 1 4108.2 4144.2
## - OnlineSecurity.xNo 1 4110.2 4146.2
## - PaymentMethod.xElectronic 1 4114.5 4150.5
## - MonthlyCharges 1 4115.1 4151.1
## - StreamingMovies.xNo 1 4115.3 4151.3
## - StreamingTV.xNo 1 4116.2 4152.2
## - MultipleLines.xNo 1 4116.6 4152.6
## - SeniorCitizen.xNo 1 4116.6 4152.6
## - TotalCharges 1 4116.8 4152.8
## - PaperlessBilling.xNo 1 4120.9 4156.9
## - Contract.xMonth.to.month 1 4139.8 4175.8
## - InternetService.xFiber.optic 1 4150.4 4186.4
## - InternetService.xDSL 1 4152.7 4188.7
## - tenure 1 4179.6 4215.6
##
## Step: AIC=4140.68
## Churn ~ SeniorCitizen.xNo + Dependents.xNo + MultipleLines.xNo +
## InternetService.xDSL + InternetService.xFiber.optic + OnlineSecurity.xNo +
## OnlineBackup.xNo + TechSupport.xNo + StreamingTV.xNo + StreamingMovies.xNo +
## Contract.xMonth.to.month + Contract.xOne.year + PaperlessBilling.xNo +
## PaymentMethod.xElectronic + MonthlyCharges + TotalCharges +
## tenure
##
## Df Deviance AIC
## - OnlineBackup.xNo 1 4106.3 4140.3
## <none> 4104.7 4140.7
## - Dependents.xNo 1 4106.7 4140.7
## - TechSupport.xNo 1 4107.4 4141.4
## + PaymentMethod.xCredit 1 4103.6 4141.6
## + PaymentMethod.xBank 1 4103.8 4141.8
## + Partner.xNo 1 4104.4 4142.4
## + Partner.xYes 1 4104.4 4142.4
## + DeviceProtection.xYes 1 4104.7 4142.7
## + DeviceProtection.xNo 1 4104.7 4142.7
## + PhoneService.xNo 1 4104.7 4142.7
## + PhoneService.xYes 1 4104.7 4142.7
## + PaymentMethod.xMailed 1 4104.7 4142.7
## - Contract.xOne.year 1 4109.4 4143.4
## - OnlineSecurity.xNo 1 4111.2 4145.2
## - MonthlyCharges 1 4116.2 4150.2
## - StreamingMovies.xNo 1 4116.2 4150.2
## - StreamingTV.xNo 1 4117.2 4151.2
## - MultipleLines.xNo 1 4117.5 4151.5
## - SeniorCitizen.xNo 1 4117.6 4151.6
## - TotalCharges 1 4118.3 4152.3
## - PaymentMethod.xElectronic 1 4121.4 4155.4
## - PaperlessBilling.xNo 1 4121.6 4155.6
## - Contract.xMonth.to.month 1 4141.1 4175.1
## - InternetService.xFiber.optic 1 4151.0 4185.0
## - InternetService.xDSL 1 4153.1 4187.1
## - tenure 1 4182.4 4216.4
##
## Step: AIC=4140.28
## Churn ~ SeniorCitizen.xNo + Dependents.xNo + MultipleLines.xNo +
## InternetService.xDSL + InternetService.xFiber.optic + OnlineSecurity.xNo +
## TechSupport.xNo + StreamingTV.xNo + StreamingMovies.xNo +
## Contract.xMonth.to.month + Contract.xOne.year + PaperlessBilling.xNo +
## PaymentMethod.xElectronic + MonthlyCharges + TotalCharges +
## tenure
##
## Df Deviance AIC
## <none> 4106.3 4140.3
## - Dependents.xNo 1 4108.4 4140.4
## - TechSupport.xNo 1 4108.7 4140.7
## + OnlineBackup.xNo 1 4104.7 4140.7
## + OnlineBackup.xYes 1 4104.7 4140.7
## + PaymentMethod.xCredit 1 4105.2 4141.2
## + PaymentMethod.xBank 1 4105.3 4141.3
## + PhoneService.xNo 1 4105.8 4141.8
## + PhoneService.xYes 1 4105.8 4141.8
## + Partner.xNo 1 4106.0 4142.0
## + Partner.xYes 1 4106.0 4142.0
## + DeviceProtection.xNo 1 4106.2 4142.2
## + DeviceProtection.xYes 1 4106.2 4142.2
## + PaymentMethod.xMailed 1 4106.3 4142.3
## - Contract.xOne.year 1 4110.9 4142.9
## - OnlineSecurity.xNo 1 4112.1 4144.1
## - SeniorCitizen.xNo 1 4119.1 4151.1
## - TotalCharges 1 4119.3 4151.3
## - StreamingMovies.xNo 1 4120.4 4152.4
## - MultipleLines.xNo 1 4120.9 4152.9
## - StreamingTV.xNo 1 4121.5 4153.5
## - MonthlyCharges 1 4122.7 4154.7
## - PaymentMethod.xElectronic 1 4122.8 4154.8
## - PaperlessBilling.xNo 1 4122.9 4154.9
## - Contract.xMonth.to.month 1 4142.5 4174.5
## - InternetService.xDSL 1 4158.5 4190.5
## - InternetService.xFiber.optic 1 4162.0 4194.0
## - tenure 1 4184.7 4216.7 glm.pred<-predict(glm.model_1,test.churn,type = "response")
summary(glm.model_1)##
## Call:
## glm(formula = Churn ~ SeniorCitizen.xNo + Dependents.xNo + MultipleLines.xNo +
## InternetService.xDSL + InternetService.xFiber.optic + OnlineSecurity.xNo +
## TechSupport.xNo + StreamingTV.xNo + StreamingMovies.xNo +
## Contract.xMonth.to.month + Contract.xOne.year + PaperlessBilling.xNo +
## PaymentMethod.xElectronic + MonthlyCharges + TotalCharges +
## tenure, family = "binomial", data = train.churn)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.9466 -0.6803 -0.3033 0.7429 3.3094
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -0.94692 0.41716 -2.270 0.023212 *
## SeniorCitizen.xNo -0.35765 0.10000 -3.577 0.000348 ***
## Dependents.xNo 0.14037 0.09751 1.440 0.150001
## MultipleLines.xNo -0.39965 0.10484 -3.812 0.000138 ***
## InternetService.xDSL 1.49991 0.20924 7.168 7.59e-13 ***
## InternetService.xFiber.optic 2.85335 0.38172 7.475 7.72e-14 ***
## OnlineSecurity.xNo 0.25793 0.10690 2.413 0.015832 *
## TechSupport.xNo 0.17140 0.11116 1.542 0.123104
## StreamingTV.xNo -0.45017 0.11559 -3.894 9.84e-05 ***
## StreamingMovies.xNo -0.43019 0.11467 -3.752 0.000176 ***
## Contract.xMonth.to.month 1.12336 0.19993 5.619 1.92e-08 ***
## Contract.xOne.year 0.42420 0.20261 2.094 0.036284 *
## PaperlessBilling.xNo -0.36329 0.08942 -4.063 4.85e-05 ***
## PaymentMethod.xElectronic 0.34128 0.08352 4.086 4.38e-05 ***
## MonthlyCharges -3.29958 0.81430 -4.052 5.08e-05 ***
## TotalCharges 2.48507 0.70264 3.537 0.000405 ***
## tenure -0.05716 0.00699 -8.177 2.90e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 5679.7 on 4873 degrees of freedom
## Residual deviance: 4106.3 on 4857 degrees of freedom
## AIC: 4140.3
##
## Number of Fisher Scoring iterations: 6 confusionMatrix(ifelse(glm.pred>=0.5,"Yes","No") %>% as.factor(),test.churn$Churn,positive = "Yes",mode="everything")## Confusion Matrix and Statistics
##
## Reference
## Prediction No Yes
## No 1461 247
## Yes 152 309
##
## Accuracy : 0.816
## 95% CI : (0.7991, 0.8321)
## No Information Rate : 0.7437
## P-Value [Acc > NIR] : 7.891e-16
##
## Kappa : 0.4889
## Mcnemar's Test P-Value : 2.528e-06
##
## Sensitivity : 0.5558
## Specificity : 0.9058
## Pos Pred Value : 0.6703
## Neg Pred Value : 0.8554
## Precision : 0.6703
## Recall : 0.5558
## F1 : 0.6077
## Prevalence : 0.2563
## Detection Rate : 0.1425
## Detection Prevalence : 0.2125
## Balanced Accuracy : 0.7308
##
## 'Positive' Class : Yes
## glm.roc<-roc(test.churn$Churn,glm.pred)3.2 ROC comparison
plot.roc(xg.roc,col="black",legacy.axes = T)
plot.roc(rf.roc,col="blue",add = T)
plot.roc(glm.roc,add=TRUE,col="red")
legend("bottom", c("XGBoost","Random Forest", "Logistic"),
lty = c(1,1), lwd = c(2, 2), col = c("black", "red", "blue"), cex = 0.75)
4 Conclusion
In this markdown,we deeply ananlyzed the churn data from kaggle in various methods and machine learning algotithms.Maybe we should walk you through the procesure once again just in case you miss.
1. read in the data and have a glimpse at it
2. transform the character features into categorical
3. EDA about all the other features regarding to the predictor–Churn
4. using rpart decision tree algorithm to roughly see the importance of each feature
5. change all categorical features into numerical ones with built-in function model.matrix
6. fit three different machine learning models using xgboost,random forest and logistic regression
What can we get from the roc comparision plot
random forest make the best prediction
in general,xgboost always does a good job in prediction,but in this special case,it fails.Maybe I should find it out
tenure,monthlycharges and totalcharges are the top3 most important features