pilot
PAVAN
29 August 2018
Importing dataset
library(readxl)
dataset<- read_excel("C:/Users/pavan/Downloads/DATA SCIENCE SET.25.26 (1).xlsx")Data Exploration
summary(dataset)## User id Category Order Date
## Min. :9.000e+09 Length:221 Length:221
## 1st Qu.:9.123e+09 Class :character Class :character
## Median :1.000e+10 Mode :character Mode :character
## Mean :9.674e+09
## 3rd Qu.:1.000e+10
## Max. :1.000e+10
## Order Time bill value Age
## Min. :1899-12-31 03:33:55 Min. : 15 Min. :13.00
## 1st Qu.:1899-12-31 12:25:00 1st Qu.: 549 1st Qu.:23.00
## Median :1899-12-31 16:05:06 Median : 1500 Median :26.00
## Mean :1899-12-31 15:47:52 Mean : 3309 Mean :30.17
## 3rd Qu.:1899-12-31 20:00:00 3rd Qu.: 3000 3rd Qu.:32.00
## Max. :1899-12-31 23:40:00 Max. :50000 Max. :62.00
## Income Gender Profession Home Address
## Min. : 0 Length:221 Length:221 Length:221
## 1st Qu.: 22000 Class :character Class :character Class :character
## Median : 30000 Mode :character Mode :character Mode :character
## Mean : 30448
## 3rd Qu.: 45000
## Max. :100000
## Office Address Family members Languages Known
## Length:221 Min. :1.0 Length:221
## Class :character 1st Qu.:3.0 Class :character
## Mode :character Median :4.0 Mode :character
## Mean :4.1
## 3rd Qu.:5.0
## Max. :8.0str(dataset)## Classes 'tbl_df', 'tbl' and 'data.frame': 221 obs. of 13 variables:
## $ User id : num 9.12e+09 1.00e+10 9.12e+09 1.00e+10 9.12e+09 ...
## $ Category : chr "Others" "Apparel" "Jewellery" "Movies/Entertainment" ...
## $ Order Date : chr "43101" "43104" "43110" "43112" ...
## $ Order Time : POSIXct, format: "1899-12-31 21:30:00" "1899-12-31 22:55:00" ...
## $ bill value : num 2000 2650 25000 750 8700 1500 22000 2200 1400 20000 ...
## $ Age : num 26 30 24 24 28 21 26 24 39 24 ...
## $ Income : num 45000 22000 50000 23000 60000 23000 45000 38000 45000 50000 ...
## $ Gender : chr "Female" "Female" "Female" "Female" ...
## $ Profession : chr "Doctor" "Entreprenuer" "Engineer" "IT Professional" ...
## $ Home Address : chr "17.4886661,78.3897115,17z(J.N.T.U)" "17.400205, 78.487070" "17.5182524,78.4732573,13z(Bolarum)" "17.424879, 78.507825" ...
## $ Office Address : chr "17.4743872,78.2975193,14z(Serilingampally)" "17.408753, 78.438758" "17.3722658,78.48243,15z(Malakpet)" "17.485128, 78.511785" ...
## $ Family members : num 7 2 4 5 4 5 7 4 5 4 ...
## $ Languages Known: chr "Malyalam,Telugu and English" "English, Telugu" "Tamil,English and Telugu" "English, Telugu" ...library(dplyr)##
## Attaching package: 'dplyr'## The following objects are masked from 'package:stats':
##
## filter, lag## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, unionc<-dataset %>% group_by(`User id`,Income,Profession,Gender,`Family members`,) %>% summarise(count=n(),bills_made=list(`bill value`),total = sum(`bill value`),categories=list(Category))
x<-c[,c('Income','total')]wcss<-vector()
for(i in 1:10) wcss[i]<-sum(kmeans(x,i)$withinss)
plot(1:10,wcss)
kmeans<-kmeans(x,4,iter.max = 300,nstart = 10)
library(cluster)
clusplot(x,
kmeans$cluster,
lines = 0,
shade = TRUE,
color = TRUE,
labels = 2,
plotchar = FALSE,
span = TRUE,
main = paste('cluster of clients'),
xlab = "Total bill value",
ylab = "income")
From above cluster plot we can infer that cluster 1,4 are standard,they only purchase when needed.
Whereas cluster 3 we can call them as “Targets” as we target those who have high income and high spenders.
Coming to cluster 2 they have less income but they spend higher.
Individual consumption representation - Buyers consumption categories
library(ggplot2)
ggplot(c,aes(Profession))+geom_bar()+coord_flip()
By seeing above plot,Buyer whose profession is “Business”" purchase so many times compared to others.
Similary students,Engineers,IT Professionals and doctors.
ggplot(dataset,aes(Category))+geom_bar()+coord_flip()
Seeing this representation you can clearly say certain categories like Grocery,Others,Events,Apparel are purchased frequently.
gender_dis<-dataset %>% group_by(Gender) %>% summarise(total = sum(`bill value`))
gender_dis## # A tibble: 2 x 2
## Gender total
## <chr> <dbl>
## 1 Female 420269
## 2 Male 311028Female purchase more products than male. Hence proved
library(ggplot2)
ggplot(dataset,aes(x=Age,y=`bill value`,colour = "blue"))+geom_point()+xlab("Age") + ylab("Amount per purchase")+theme_dark()
Age between 20 and 30 are the consumers who purchase a lot.
Area wise purchase distribution- Heat map
library(ggmap)
library(tidyr)
library(dplyr)
address<-separate(dataset,'Home Address',sep=",",c("latitude","longitude","nameofplace"))## Warning: Expected 3 pieces. Missing pieces filled with `NA` in 140 rows
## [2, 4, 13, 22, 23, 27, 30, 32, 33, 36, 38, 39, 40, 42, 43, 44, 46, 49, 51,
## 52, ...].d<-address %>% group_by(`User id`,latitude,longitude) %>% summarise(total = sum(`bill value`),count = n())
d$longitude<-as.numeric(d$longitude)
d$latitude<-as.numeric(d$latitude)hyd_center = as.numeric(geocode("Hyderabad"))## Information from URL : http://maps.googleapis.com/maps/api/geocode/json?address=Hyderabad&sensor=falsehydmap = ggmap(get_googlemap(center= c(lon = 78.45667, lat = 17.45504), scale=2, zoom=12), extent="panel")## Map from URL : http://maps.googleapis.com/maps/api/staticmap?center=17.45504,78.45667&zoom=12&size=640x640&scale=2&maptype=terrain&sensor=falsehydmap + geom_point(aes(x=longitude, y=latitude), data= d, col="orange", alpha=0.4,size = d$count) + scale_size_continuous(range=range(d$count))+ggtitle("Area Wise Purchase Distribution")## Warning: Removed 3 rows containing missing values (geom_point).
From above near Banjara Hills and kukatpally area has more number of purchases.
Time of Purchase
library(tidyr)
time<-separate(dataset,'Order Time',sep = " ",c("date","order_time"))
t<-time[,c("bill value","order_time")]
t<-separate(t,order_time,sep = ":",c("hours","min","sec"))
t$hours<-as.numeric(t$hours)
t$min<-as.numeric(t$min)
t<-select(t,-c("sec"))
t$hours<-ifelse(t$min>=30,t$hours+1,t$min)
t<-select(t,-c("min"))clustering
wcss2<-vector()
for(i in 1:10) wcss2[i]<-sum(kmeans(t,i)$withinss)
plot(1:10,wcss2)
kmeans2<-kmeans(t,3,iter.max=300,nstart =10)
library(cluster)
clusplot(t,
kmeans2$cluster,
lines = 0,
shade = TRUE,
color = TRUE,
labels = 4,
plotchar = FALSE,
span = TRUE,
main = paste('Time of purchasing cluster'))
age_wise<-time[,c("Age","order_time")]
age_wise<-separate(age_wise,order_time,sep = ":",c("hours","min","sec"))
age_wise$hours<-as.numeric(age_wise$hours)
age_wise$min<-as.numeric(age_wise$min)
age_wise<-select(age_wise,-c("sec"))
age_wise$hours<-ifelse(age_wise$min>=30,age_wise$hours+1,age_wise$min)
age_wise<-select(age_wise,-c("min"))wcss3<-vector()
for(i in 1:10) wcss3[i]<-sum(kmeans(age_wise,i)$withinss)
plot(1:10,wcss3)
kmeans3<-kmeans(age_wise,4,iter.max=300,nstart =10)
library(cluster)
clusplot(age_wise,
kmeans3$cluster,
lines = 0,
shade = TRUE,
color = TRUE,
labels = 4,
plotchar = FALSE,
span = TRUE,
main = paste('Time of purchasing cluster'))
Correlation between Categories
library(vcd)## Loading required package: gridassocstats(table(dataset$Category,dataset$Gender))## X^2 df P(> X^2)
## Likelihood Ratio 45.425 14 3.4793e-05
## Pearson 41.475 14 1.4970e-04
##
## Phi-Coefficient : NA
## Contingency Coeff.: 0.398
## Cramer's V : 0.433There is a strong correlation between category and gender
ggplot(dataset,aes(Gender,Category))+geom_jitter()
Female purchases Apparel,Grocery category products more compared to male. While Male purchase Movies/Entertainment,Transit category products more.