Customer Data Analysis
# load customer data
customers <- read.csv("data/customer_info.csv", stringsAsFactors = TRUE)
# load product data
products <- read.csv("data/product_info.csv", stringsAsFactors = TRUE)
# load customer product data
customer_products <- read.csv("data/customer_product.csv")#, stringsAsFactors = TRUE)
#Identify number of records
dim(customers)
## [1] 508932 4
## [1] 2 4
## [1] 508932 5
#view customer data
summary(customers)
## X customer_id age gender
## Min. : 1 C10000 : 1 Min. :21.00 female:199002
## 1st Qu.:127234 C100000: 1 1st Qu.:53.00 male :309930
## Median :254467 C100001: 1 Median :59.00
## Mean :254467 C100002: 1 Mean :58.47
## 3rd Qu.:381699 C100003: 1 3rd Qu.:64.00
## Max. :508932 C100004: 1 Max. :78.00
## (Other):508926
#view customer data
summary(products)
## product_id name price billing_cycle
## prd_1:1 annual_subscription :1 Min. : 125.0 Min. : 1.00
## prd_2:1 monthly_subscription:1 1st Qu.: 393.8 1st Qu.: 3.75
## Median : 662.5 Median : 6.50
## Mean : 662.5 Mean : 6.50
## 3rd Qu.: 931.2 3rd Qu.: 9.25
## Max. :1200.0 Max. :12.00
#view customer data
summary(customer_products)
## X customer_id product signup_date_time
## Min. : 1 Length:508932 Length:508932 Length:508932
## 1st Qu.:127234 Class :character Class :character Class :character
## Median :254467 Mode :character Mode :character Mode :character
## Mean :254467
## 3rd Qu.:381699
## Max. :508932
## cancel_date_time
## Length:508932
## Class :character
## Mode :character
##
##
##
#view customer data structure
str(customers)
## 'data.frame': 508932 obs. of 4 variables:
## $ X : int 1 2 3 4 5 6 7 8 9 10 ...
## $ customer_id: Factor w/ 508932 levels "C10000","C100000",..: 159281 159392 159503 159614 159725 159836 159947 160058 160169 160280 ...
## $ age : int 76 61 58 62 71 56 53 53 70 68 ...
## $ gender : Factor w/ 2 levels "female","male": 1 2 1 1 2 2 2 1 1 2 ...
#view customer data structure
str(products)
## 'data.frame': 2 obs. of 4 variables:
## $ product_id : Factor w/ 2 levels "prd_1","prd_2": 1 2
## $ name : Factor w/ 2 levels "annual_subscription",..: 1 2
## $ price : int 1200 125
## $ billing_cycle: int 12 1
#view customer data structure
str(customer_products)
## 'data.frame': 508932 obs. of 5 variables:
## $ X : int 1 2 3 4 5 6 7 8 9 10 ...
## $ customer_id : chr "C2448" "C2449" "C2450" "C2451" ...
## $ product : chr "prd_1" "prd_1" "prd_1" "prd_2" ...
## $ signup_date_time: chr "2017-01-01 10:35:09" "2017-01-01 11:39:29" "2017-01-01 11:42:00" "2017-01-01 13:32:08" ...
## $ cancel_date_time: chr NA "2021-09-05 10:00:02" "2019-01-13 16:24:55" NA ...
#Cleaning of Customer Data
#remove unwanted column
customers_clean <- customers[,c(2:4)]
#view summary of cleaned data
summary(customers_clean)
## customer_id age gender
## C10000 : 1 Min. :21.00 female:199002
## C100000: 1 1st Qu.:53.00 male :309930
## C100001: 1 Median :59.00
## C100002: 1 Mean :58.47
## C100003: 1 3rd Qu.:64.00
## C100004: 1 Max. :78.00
## (Other):508926
#view a subset of cleaned data
head(customers_clean)
## customer_id age gender
## 1 C2448 76 female
## 2 C2449 61 male
## 3 C2450 58 female
## 4 C2451 62 female
## 5 C2452 71 male
## 6 C2453 56 male
#view data structure of customer product data
str(customer_products)
## 'data.frame': 508932 obs. of 5 variables:
## $ X : int 1 2 3 4 5 6 7 8 9 10 ...
## $ customer_id : chr "C2448" "C2449" "C2450" "C2451" ...
## $ product : chr "prd_1" "prd_1" "prd_1" "prd_2" ...
## $ signup_date_time: chr "2017-01-01 10:35:09" "2017-01-01 11:39:29" "2017-01-01 11:42:00" "2017-01-01 13:32:08" ...
## $ cancel_date_time: chr NA "2021-09-05 10:00:02" "2019-01-13 16:24:55" NA ...
#seeing that date time comes in as factor
# bring in lubridate to handle datetime
library(lubridate, warn.conflicts = FALSE)
#Cleaning of Customer Product Data
#remove unwanted column
customer_product_clean <- customer_products[,c(2:5)]
#convert the date time column to datetime type
customer_product_clean$signup_date_time <- ymd_hms(customer_product_clean$signup_date_time)
customer_product_clean$cancel_date_time <- ymd_hms(customer_product_clean$cancel_date_time)
#rename product column in customer_product to product_id
colnames(customer_product_clean)[2] <- "product_id"
#rename name column in product to subscription
colnames(products)[2] <- "subscription"
#combining data frames (customer product clean and products)
customer_consolidated_data = merge(x=customer_product_clean, y=products, by ="product_id", all.x = TRUE)
#combining data frames (customer consolidated data and customer data)
customer_consolidated_data = merge(x=customer_consolidated_data, y=customers_clean, by ="customer_id", all.x = TRUE)
head(customer_consolidated_data)
## customer_id product_id signup_date_time cancel_date_time
## 1 C10000 prd_1 2017-07-09 12:55:31 <NA>
## 2 C100000 prd_2 2021-02-13 11:41:49 <NA>
## 3 C100001 prd_1 2021-02-13 11:45:50 <NA>
## 4 C100002 prd_1 2021-02-13 11:51:02 <NA>
## 5 C100003 prd_1 2021-02-13 11:56:54 <NA>
## 6 C100004 prd_1 2021-02-13 12:05:35 <NA>
## subscription price billing_cycle age gender
## 1 annual_subscription 1200 12 46 female
## 2 monthly_subscription 125 1 54 female
## 3 annual_subscription 1200 12 27 female
## 4 annual_subscription 1200 12 47 female
## 5 annual_subscription 1200 12 63 female
## 6 annual_subscription 1200 12 40 male
str(customer_consolidated_data)
## 'data.frame': 508932 obs. of 9 variables:
## $ customer_id : chr "C10000" "C100000" "C100001" "C100002" ...
## $ product_id : chr "prd_1" "prd_2" "prd_1" "prd_1" ...
## $ signup_date_time: POSIXct, format: "2017-07-09 12:55:31" "2021-02-13 11:41:49" ...
## $ cancel_date_time: POSIXct, format: NA NA ...
## $ subscription : Factor w/ 2 levels "annual_subscription",..: 1 2 1 1 1 1 2 2 1 2 ...
## $ price : int 1200 125 1200 1200 1200 1200 125 125 1200 125 ...
## $ billing_cycle : int 12 1 12 12 12 12 1 1 12 1 ...
## $ age : int 46 54 27 47 63 40 69 48 60 60 ...
## $ gender : Factor w/ 2 levels "female","male": 1 1 1 1 1 2 1 1 2 1 ...
#duplicate dataframe for manipulation purpose
consolidated_modified <- customer_consolidated_data
#Signup analytic breakdown
consolidated_modified$yearIn <- strftime(consolidated_modified$signup_date_time, "%Y")
consolidated_modified$monthIn <- strftime(consolidated_modified$signup_date_time, "%h")
consolidated_modified$dayIn <- strftime(consolidated_modified$signup_date_time, "%d")
consolidated_modified$dayNameIn <- strftime(consolidated_modified$signup_date_time, "%a")
consolidated_modified$hourIn <- strftime(consolidated_modified$signup_date_time, "%H")
consolidated_modified$minuteIn <- strftime(consolidated_modified$signup_date_time, "%M")
consolidated_modified$AMPMIn <- strftime(consolidated_modified$signup_date_time, "%p")
#Signup analytic breakdown
consolidated_modified$yearOut <- strftime(consolidated_modified$cancel_date_time, "%Y")
consolidated_modified$monthOut <- strftime(consolidated_modified$cancel_date_time, "%h")
consolidated_modified$dayOut <- strftime(consolidated_modified$cancel_date_time, "%d")
consolidated_modified$dayNameOut <- strftime(consolidated_modified$cancel_date_time, "%a")
consolidated_modified$hourOut <- strftime(consolidated_modified$cancel_date_time, "%H")
consolidated_modified$minuteOut <- strftime(consolidated_modified$cancel_date_time, "%M")
consolidated_modified$AMPMOut <- strftime(consolidated_modified$cancel_date_time, "%p")
str(consolidated_modified)
## 'data.frame': 508932 obs. of 23 variables:
## $ customer_id : chr "C10000" "C100000" "C100001" "C100002" ...
## $ product_id : chr "prd_1" "prd_2" "prd_1" "prd_1" ...
## $ signup_date_time: POSIXct, format: "2017-07-09 12:55:31" "2021-02-13 11:41:49" ...
## $ cancel_date_time: POSIXct, format: NA NA ...
## $ subscription : Factor w/ 2 levels "annual_subscription",..: 1 2 1 1 1 1 2 2 1 2 ...
## $ price : int 1200 125 1200 1200 1200 1200 125 125 1200 125 ...
## $ billing_cycle : int 12 1 12 12 12 12 1 1 12 1 ...
## $ age : int 46 54 27 47 63 40 69 48 60 60 ...
## $ gender : Factor w/ 2 levels "female","male": 1 1 1 1 1 2 1 1 2 1 ...
## $ yearIn : chr "2017" "2021" "2021" "2021" ...
## $ monthIn : chr "Jul" "Feb" "Feb" "Feb" ...
## $ dayIn : chr "09" "13" "13" "13" ...
## $ dayNameIn : chr "Sun" "Sat" "Sat" "Sat" ...
## $ hourIn : chr "13" "11" "11" "11" ...
## $ minuteIn : chr "55" "41" "45" "51" ...
## $ AMPMIn : chr "PM" "AM" "AM" "AM" ...
## $ yearOut : chr NA NA NA NA ...
## $ monthOut : chr NA NA NA NA ...
## $ dayOut : chr NA NA NA NA ...
## $ dayNameOut : chr NA NA NA NA ...
## $ hourOut : chr NA NA NA NA ...
## $ minuteOut : chr NA NA NA NA ...
## $ AMPMOut : chr NA NA NA NA ...
#convert factorizable columns to factor
consolidated_modified$yearIn <- factor(consolidated_modified$yearIn)
consolidated_modified$yearOut <- factor(consolidated_modified$yearOut)
consolidated_modified$monthIn <- factor(consolidated_modified$monthIn)
consolidated_modified$monthOut <- factor(consolidated_modified$monthOut)
consolidated_modified$dayNameIn <- factor(consolidated_modified$dayNameIn)
consolidated_modified$dayNameOut <- factor(consolidated_modified$dayNameOut)
consolidated_modified$AMPMIn <- factor(consolidated_modified$AMPMIn)
consolidated_modified$AMPMOut <- factor(consolidated_modified$AMPMOut)
str(consolidated_modified)
## 'data.frame': 508932 obs. of 23 variables:
## $ customer_id : chr "C10000" "C100000" "C100001" "C100002" ...
## $ product_id : chr "prd_1" "prd_2" "prd_1" "prd_1" ...
## $ signup_date_time: POSIXct, format: "2017-07-09 12:55:31" "2021-02-13 11:41:49" ...
## $ cancel_date_time: POSIXct, format: NA NA ...
## $ subscription : Factor w/ 2 levels "annual_subscription",..: 1 2 1 1 1 1 2 2 1 2 ...
## $ price : int 1200 125 1200 1200 1200 1200 125 125 1200 125 ...
## $ billing_cycle : int 12 1 12 12 12 12 1 1 12 1 ...
## $ age : int 46 54 27 47 63 40 69 48 60 60 ...
## $ gender : Factor w/ 2 levels "female","male": 1 1 1 1 1 2 1 1 2 1 ...
## $ yearIn : Factor w/ 5 levels "2017","2018",..: 1 5 5 5 5 5 5 5 5 5 ...
## $ monthIn : Factor w/ 12 levels "Apr","Aug","Dec",..: 6 4 4 4 4 4 4 4 4 4 ...
## $ dayIn : chr "09" "13" "13" "13" ...
## $ dayNameIn : Factor w/ 7 levels "Fri","Mon","Sat",..: 4 3 3 3 3 3 3 3 3 3 ...
## $ hourIn : chr "13" "11" "11" "11" ...
## $ minuteIn : chr "55" "41" "45" "51" ...
## $ AMPMIn : Factor w/ 2 levels "AM","PM": 2 1 1 1 1 2 2 2 2 2 ...
## $ yearOut : Factor w/ 5 levels "2017","2018",..: NA NA NA NA NA NA NA NA NA NA ...
## $ monthOut : Factor w/ 12 levels "Apr","Aug","Dec",..: NA NA NA NA NA NA NA NA NA NA ...
## $ dayOut : chr NA NA NA NA ...
## $ dayNameOut : Factor w/ 7 levels "Fri","Mon","Sat",..: NA NA NA NA NA NA NA NA NA NA ...
## $ hourOut : chr NA NA NA NA ...
## $ minuteOut : chr NA NA NA NA ...
## $ AMPMOut : Factor w/ 2 levels "AM","PM": NA NA NA NA NA NA NA NA NA NA ...
#Descriptive Information
#Average age of customers based on gender
with(consolidated_modified, by(age, gender, mean))
## gender: female
## [1] 58.34073
## ------------------------------------------------------------
## gender: male
## [1] 58.55798
#loading ggplot library
library(ggplot2)
ggplot(consolidated_modified,aes(y=age, x=gender, fill=gender )) + geom_boxplot() + coord_flip()+ scale_fill_brewer(palette = "Set1")
#Creating dataframe for cancellations of subscriptions
consolidated_cancellation <- consolidated_modified
consolidated_cancellation <-na.omit(consolidated_cancellation)
head(consolidated_cancellation)
## customer_id product_id signup_date_time cancel_date_time
## 11 C100009 prd_2 2021-02-13 12:22:02 2021-12-26 14:07:50
## 15 C100012 prd_1 2021-02-13 12:42:20 2021-02-23 16:45:25
## 17 C100014 prd_2 2021-02-13 13:01:18 2021-11-09 16:59:34
## 23 C10002 prd_1 2017-07-10 06:27:10 2019-01-03 09:51:28
## 26 C100022 prd_2 2021-02-13 14:11:18 2021-09-21 11:25:56
## 27 C100023 prd_1 2021-02-13 14:14:05 2021-06-12 15:07:44
## subscription price billing_cycle age gender yearIn monthIn dayIn
## 11 monthly_subscription 125 1 75 male 2021 Feb 13
## 15 annual_subscription 1200 12 43 male 2021 Feb 13
## 17 monthly_subscription 125 1 57 male 2021 Feb 13
## 23 annual_subscription 1200 12 58 female 2017 Jul 10
## 26 monthly_subscription 125 1 47 female 2021 Feb 13
## 27 annual_subscription 1200 12 50 female 2021 Feb 13
## dayNameIn hourIn minuteIn AMPMIn yearOut monthOut dayOut dayNameOut hourOut
## 11 Sat 12 22 PM 2021 Dec 26 Sun 14
## 15 Sat 12 42 PM 2021 Feb 23 Tue 16
## 17 Sat 13 01 PM 2021 Nov 09 Tue 16
## 23 Mon 07 27 AM 2019 Jan 03 Thu 09
## 26 Sat 14 11 PM 2021 Sep 21 Tue 12
## 27 Sat 14 14 PM 2021 Jun 12 Sat 16
## minuteOut AMPMOut
## 11 07 PM
## 15 45 PM
## 17 59 PM
## 23 51 AM
## 26 25 PM
## 27 07 PM
str(consolidated_cancellation)
## 'data.frame': 112485 obs. of 23 variables:
## $ customer_id : chr "C100009" "C100012" "C100014" "C10002" ...
## $ product_id : chr "prd_2" "prd_1" "prd_2" "prd_1" ...
## $ signup_date_time: POSIXct, format: "2021-02-13 12:22:02" "2021-02-13 12:42:20" ...
## $ cancel_date_time: POSIXct, format: "2021-12-26 14:07:50" "2021-02-23 16:45:25" ...
## $ subscription : Factor w/ 2 levels "annual_subscription",..: 2 1 2 1 2 1 2 2 2 2 ...
## $ price : int 125 1200 125 1200 125 1200 125 125 125 125 ...
## $ billing_cycle : int 1 12 1 12 1 12 1 1 1 1 ...
## $ age : int 75 43 57 58 47 50 70 41 49 57 ...
## $ gender : Factor w/ 2 levels "female","male": 2 2 2 1 1 1 2 2 2 1 ...
## $ yearIn : Factor w/ 5 levels "2017","2018",..: 5 5 5 1 5 5 5 5 5 5 ...
## $ monthIn : Factor w/ 12 levels "Apr","Aug","Dec",..: 4 4 4 6 4 4 4 4 4 4 ...
## $ dayIn : chr "13" "13" "13" "10" ...
## $ dayNameIn : Factor w/ 7 levels "Fri","Mon","Sat",..: 3 3 3 2 3 3 3 4 4 4 ...
## $ hourIn : chr "12" "12" "13" "07" ...
## $ minuteIn : chr "22" "42" "01" "27" ...
## $ AMPMIn : Factor w/ 2 levels "AM","PM": 2 2 2 1 2 2 2 1 1 1 ...
## $ yearOut : Factor w/ 5 levels "2017","2018",..: 5 5 5 3 5 5 5 5 5 5 ...
## $ monthOut : Factor w/ 12 levels "Apr","Aug","Dec",..: 3 4 10 5 12 7 9 3 6 9 ...
## $ dayOut : chr "26" "23" "09" "03" ...
## $ dayNameOut : Factor w/ 7 levels "Fri","Mon","Sat",..: 4 6 6 5 6 3 5 5 1 4 ...
## $ hourOut : chr "14" "16" "16" "09" ...
## $ minuteOut : chr "07" "45" "59" "51" ...
## $ AMPMOut : Factor w/ 2 levels "AM","PM": 2 2 2 1 2 2 2 2 2 2 ...
## - attr(*, "na.action")= 'omit' Named int [1:396447] 1 2 3 4 5 6 7 8 9 10 ...
## ..- attr(*, "names")= chr [1:396447] "1" "2" "3" "4" ...
summary(consolidated_cancellation)
## customer_id product_id signup_date_time
## Length:112485 Length:112485 Min. :2017-01-01 10:14:28
## Class :character Class :character 1st Qu.:2018-07-11 14:54:08
## Mode :character Mode :character Median :2019-07-11 08:20:30
## Mean :2019-06-28 10:54:26
## 3rd Qu.:2020-05-26 17:16:48
## Max. :2021-12-26 10:37:59
##
## cancel_date_time subscription price
## Min. :2017-01-08 15:14:53 annual_subscription :69682 Min. : 125.0
## 1st Qu.:2020-03-03 15:25:12 monthly_subscription:42803 1st Qu.: 125.0
## Median :2020-12-29 13:24:10 Median :1200.0
## Mean :2020-09-29 23:06:47 Mean : 790.9
## 3rd Qu.:2021-07-22 10:59:54 3rd Qu.:1200.0
## Max. :2021-12-31 21:44:10 Max. :1200.0
##
## billing_cycle age gender yearIn monthIn
## Min. : 1.000 Min. :23.00 female:44229 2017:13376 Jan :14593
## 1st Qu.: 1.000 1st Qu.:53.00 male :68256 2018:21241 Mar :13696
## Median :12.000 Median :59.00 2019:32797 Apr :12897
## Mean : 7.814 Mean :58.42 2020:30315 Feb :12596
## 3rd Qu.:12.000 3rd Qu.:64.00 2021:14756 May :12161
## Max. :12.000 Max. :78.00 Jun :10347
## (Other):36195
## dayIn dayNameIn hourIn minuteIn
## Length:112485 Fri:14821 Length:112485 Length:112485
## Class :character Mon:23794 Class :character Class :character
## Mode :character Sat: 7781 Mode :character Mode :character
## Sun: 5907
## Thu:18489
## Tue:21616
## Wed:20077
## AMPMIn yearOut monthOut dayOut dayNameOut
## AM:27507 2017: 1779 Dec :12043 Length:112485 Fri:16154
## PM:84978 2018: 6447 Nov :11270 Class :character Mon:16103
## 2019:15841 Oct :11138 Mode :character Sat:16168
## 2020:32500 Sep :10605 Sun:16041
## 2021:55918 Aug :10414 Thu:16119
## Jul : 9942 Tue:15795
## (Other):47073 Wed:16105
## hourOut minuteOut AMPMOut
## Length:112485 Length:112485 AM:22034
## Class :character Class :character PM:90451
## Mode :character Mode :character
##
##
##
##
#install.packages("patchwork")
library(patchwork)
summary(consolidated_modified$subscription)
## annual_subscription monthly_subscription
## 325649 183283
summary(consolidated_modified$yearIn)
## 2017 2018 2019 2020 2021
## 30896 58099 109876 145642 164419
plot1 <- ggplot(consolidated_modified,aes(x=subscription, fill=subscription)) + geom_bar(stat = "count")
plot2 <- ggplot(consolidated_modified,aes(x=yearIn)) + geom_bar(stat = "count")
plot1 + plot2
ggplot(consolidated_cancellation,aes(x=yearOut, fill= gender)) + geom_bar(stat = "count")
ggplot(consolidated_modified,aes(x=monthIn, fill = gender)) + geom_bar(stat = "count") + coord_flip()
ggplot(consolidated_cancellation,aes(x=monthOut, fill = gender)) + geom_bar(stat = "count") + coord_flip()
ggplot(consolidated_modified,aes(x=yearIn)) + geom_bar(stat = "count") + ggtitle("Signup Annual Frequency")
ggplot(consolidated_cancellation,aes(x=yearOut)) + geom_bar(stat = "count") + ggtitle("Cancellation Annual Frequency")
ggplot(consolidated_modified,aes(x=monthIn, fill=subscription)) + geom_bar(stat = "count") + ggtitle("Signup Monthly Frequency")
ggplot(consolidated_cancellation,aes(x=monthOut, fill=subscription)) + geom_bar(stat = "count") + ggtitle("Cancellation Monthly Frequency")
ggplot(consolidated_modified,aes(x=dayNameIn, fill=subscription)) + geom_bar(stat = "count") + ggtitle("Signup by Day of the Week")
ggplot(consolidated_cancellation,aes(x=dayNameOut, fill=subscription)) + geom_bar(stat = "count") + ggtitle("Cancellation by Day of the Week")
ggplot(consolidated_modified,aes(x=dayIn, fill=subscription)) + geom_bar(stat = "count") + ggtitle("Signup by Day of the Month")
ggplot(consolidated_cancellation,aes(x=dayOut, fill=subscription)) + geom_bar(stat = "count") + ggtitle("Cancellation by Day of the Month")
ggplot(consolidated_modified,aes(x=hourIn, fill=subscription)) + geom_bar(stat = "count") + ggtitle("Signup by Time of the Day")
ggplot(consolidated_cancellation,aes(x=hourOut, fill=subscription)) + geom_bar(stat = "count") + ggtitle("Cancellation by Time of the Day")
ggplot(consolidated_modified,aes(x=AMPMIn, fill=subscription)) + geom_bar(stat = "count") + ggtitle("Signup by Period of the Day")
ggplot(consolidated_cancellation,aes(x=AMPMOut, fill=subscription)) + geom_bar(stat = "count") + ggtitle("Cancellation by Period of the Day")
#create churn status column
consolidated_modified$churn <- ifelse(is.na(consolidated_modified$cancel_date_time),"No", "Yes")
#convert churn column type to factor
consolidated_modified$churn <- factor(consolidated_modified$churn)
str(consolidated_modified)
## 'data.frame': 508932 obs. of 24 variables:
## $ customer_id : chr "C10000" "C100000" "C100001" "C100002" ...
## $ product_id : chr "prd_1" "prd_2" "prd_1" "prd_1" ...
## $ signup_date_time: POSIXct, format: "2017-07-09 12:55:31" "2021-02-13 11:41:49" ...
## $ cancel_date_time: POSIXct, format: NA NA ...
## $ subscription : Factor w/ 2 levels "annual_subscription",..: 1 2 1 1 1 1 2 2 1 2 ...
## $ price : int 1200 125 1200 1200 1200 1200 125 125 1200 125 ...
## $ billing_cycle : int 12 1 12 12 12 12 1 1 12 1 ...
## $ age : int 46 54 27 47 63 40 69 48 60 60 ...
## $ gender : Factor w/ 2 levels "female","male": 1 1 1 1 1 2 1 1 2 1 ...
## $ yearIn : Factor w/ 5 levels "2017","2018",..: 1 5 5 5 5 5 5 5 5 5 ...
## $ monthIn : Factor w/ 12 levels "Apr","Aug","Dec",..: 6 4 4 4 4 4 4 4 4 4 ...
## $ dayIn : chr "09" "13" "13" "13" ...
## $ dayNameIn : Factor w/ 7 levels "Fri","Mon","Sat",..: 4 3 3 3 3 3 3 3 3 3 ...
## $ hourIn : chr "13" "11" "11" "11" ...
## $ minuteIn : chr "55" "41" "45" "51" ...
## $ AMPMIn : Factor w/ 2 levels "AM","PM": 2 1 1 1 1 2 2 2 2 2 ...
## $ yearOut : Factor w/ 5 levels "2017","2018",..: NA NA NA NA NA NA NA NA NA NA ...
## $ monthOut : Factor w/ 12 levels "Apr","Aug","Dec",..: NA NA NA NA NA NA NA NA NA NA ...
## $ dayOut : chr NA NA NA NA ...
## $ dayNameOut : Factor w/ 7 levels "Fri","Mon","Sat",..: NA NA NA NA NA NA NA NA NA NA ...
## $ hourOut : chr NA NA NA NA ...
## $ minuteOut : chr NA NA NA NA ...
## $ AMPMOut : Factor w/ 2 levels "AM","PM": NA NA NA NA NA NA NA NA NA NA ...
## $ churn : Factor w/ 2 levels "No","Yes": 1 1 1 1 1 1 1 1 1 1 ...
#look at some characeristics
summary(consolidated_modified$subscription)
## annual_subscription monthly_subscription
## 325649 183283
#looking at the dependent variable (binary - Churn or not)
summary(consolidated_modified$churn)
## No Yes
## 396447 112485
#isolating data for tree classification
churn_data <- consolidated_modified[,c(5,8,9,10,11,12,13,14,15,16,24)]
str(churn_data)
## 'data.frame': 508932 obs. of 11 variables:
## $ subscription: Factor w/ 2 levels "annual_subscription",..: 1 2 1 1 1 1 2 2 1 2 ...
## $ age : int 46 54 27 47 63 40 69 48 60 60 ...
## $ gender : Factor w/ 2 levels "female","male": 1 1 1 1 1 2 1 1 2 1 ...
## $ yearIn : Factor w/ 5 levels "2017","2018",..: 1 5 5 5 5 5 5 5 5 5 ...
## $ monthIn : Factor w/ 12 levels "Apr","Aug","Dec",..: 6 4 4 4 4 4 4 4 4 4 ...
## $ dayIn : chr "09" "13" "13" "13" ...
## $ dayNameIn : Factor w/ 7 levels "Fri","Mon","Sat",..: 4 3 3 3 3 3 3 3 3 3 ...
## $ hourIn : chr "13" "11" "11" "11" ...
## $ minuteIn : chr "55" "41" "45" "51" ...
## $ AMPMIn : Factor w/ 2 levels "AM","PM": 2 1 1 1 1 2 2 2 2 2 ...
## $ churn : Factor w/ 2 levels "No","Yes": 1 1 1 1 1 1 1 1 1 1 ...
## subscription age gender yearIn
## annual_subscription :325649 Min. :21.00 female:199002 2017: 30896
## monthly_subscription:183283 1st Qu.:53.00 male :309930 2018: 58099
## Median :59.00 2019:109876
## Mean :58.47 2020:145642
## 3rd Qu.:64.00 2021:164419
## Max. :78.00
##
## monthIn dayIn dayNameIn hourIn
## Mar : 54664 Length:508932 Fri: 66426 Length:508932
## Apr : 52914 Class :character Mon:107779 Class :character
## Jan : 52805 Mode :character Sat: 37735 Mode :character
## May : 51936 Sun: 28989
## Feb : 48388 Thu: 81831
## Jun : 47908 Tue: 95682
## (Other):200317 Wed: 90490
## minuteIn AMPMIn churn
## Length:508932 AM:120577 No :396447
## Class :character PM:388355 Yes:112485
## Mode :character
##
##
##
##
#convert some column type to factor
churn_data$dayIn <- factor(churn_data$dayIn)
churn_data$hourIn <- factor(churn_data$hourIn)
churn_data$minuteIn <- factor(churn_data$minuteIn)
set.seed(427) #setting seed
train_sample <- sample(508932, 470000)
#view train sample data type structure
str(train_sample)
## int [1:470000] 101589 334339 301126 19531 25124 477442 34056 279323 303645 34196 ...
#splitting data for training and testing
churn_train <- churn_data[train_sample,] #training data
churn_test <- churn_data[-train_sample,] #test data
# check the proportion of class variable
prop.table(summary(churn_train$churn))
## No Yes
## 0.779017 0.220983
prop.table(summary(churn_test$churn))
## No Yes
## 0.7785113 0.2214887
#training a model on the data
#load C50 library
#install.packages("C50")
library(C50)
churn_model <- C5.0(churn_train[-11],churn_train$churn,control = C5.0Control(minCases = 2000))
churn_model
##
## Call:
## C5.0.default(x = churn_train[-11], y = churn_train$churn, control
## = C5.0Control(minCases = 2000))
##
## Classification Tree
## Number of samples: 470000
## Number of predictors: 10
##
## Tree size: 3
##
## Non-standard options: attempt to group attributes, minimum number of cases: 2000
#display detailed info about the tree
summary(churn_model)
##
## Call:
## C5.0.default(x = churn_train[-11], y = churn_train$churn, control
## = C5.0Control(minCases = 2000))
##
##
## C5.0 [Release 2.07 GPL Edition] Sat Jul 23 03:35:43 2022
## -------------------------------
##
## Class specified by attribute `outcome'
##
## Read 470000 cases (11 attributes) from undefined.data
##
## Decision tree:
##
## subscription = annual_subscription: No (300770/64386)
## subscription = monthly_subscription:
## :...yearIn = 2017: Yes (4914/2260)
## yearIn in {2018,2019,2020,2021}: No (164316/36822)
##
##
## Evaluation on training data (470000 cases):
##
## Decision Tree
## ----------------
## Size Errors
##
## 3 103468(22.0%) <<
##
##
## (a) (b) <-classified as
## ----- -----
## 363878 2260 (a): class No
## 101208 2654 (b): class Yes
##
##
## Attribute usage:
##
## 100.00% subscription
## 36.01% yearIn
##
##
## Time: 9.6 secs
#plot the tree
#plot(churn_model) #gives unresolved error
#install.packages("gmodels")
library(gmodels)
## Step 3: Evaluating model performance by using test data
# create a factor vector of predictions on test data
churn_pred <- predict(churn_model,churn_test)
# cross tabulation of predicted versus actual classes
CrossTable(churn_test$churn, churn_pred,
prop.chisq = FALSE, prop.c = FALSE, prop.r = FALSE,
dnn = c('actual default','predicted default'))
##
##
## Cell Contents
## |-------------------------|
## | N |
## | N / Table Total |
## |-------------------------|
##
##
## Total Observations in Table: 38932
##
##
## | predicted default
## actual default | No | Yes | Row Total |
## ---------------|-----------|-----------|-----------|
## No | 30116 | 193 | 30309 |
## | 0.774 | 0.005 | |
## ---------------|-----------|-----------|-----------|
## Yes | 8389 | 234 | 8623 |
## | 0.215 | 0.006 | |
## ---------------|-----------|-----------|-----------|
## Column Total | 38505 | 427 | 38932 |
## ---------------|-----------|-----------|-----------|
##
##