##################################
# Analysis of Bank Marketing Data set: https://archive.ics.uci.edu/ml/datasets/Bank+Marketing#
# Reference: [Moro et al., 2014] S. Moro, P. Cortez and P. Rita. A Data-Driven Approach to Predict the Success of Bank Telemarketing. Decision Support Systems, Elsevier, 62:22-31, June 2014
# Author: DIana Marek
# Date: December 2016
##################################
# list workspace
ls()

# Reset R'brain
rm(list=ls())

# check wd
getwd()

#set wd
setwd("/Users/dmarek/EducationSIB/Courses_2016/Nespresso_R_course")

# confirm wd
getwd()

# Read the data
bank_data<-read.csv2("bank.csv")
# bank_data<- read.csv("bank.csv", sep=";") # this is an alternative
# bank_data<- read.table("bank.csv", header=TRUE, sep=";") # this is an alternative
View(bank_data) # equivalent to doble click on the object name

#customers<-read.csv2('FR_Customers_sample_small.csv')
orders<-read.csv2('FR_Orders_sample_small.csv')
hist(as.numeric(orders$CHF_REVENUE))

# Structure, info about bank_data
class(bank_data)
head(bank_data) 
dim(bank_data)
str(bank_data)
colnames(bank_data)
class(bank_data$age)
class(bank_data$balance)
class(bank_data$job)
levels(bank_data$job)
table(bank_data$education) #counts occurrences for each level of the factor
with(bank_data,table(education)) # equivalent to the above command
prop.table(bank_data$education) #counts can be converted to proportions with prop.table()

# Basic stats
summary(bank_data)

# Subsetting
subset(bank_data,month=="dec") # keeps only the customers contacted in december
subset(bank_data,month=="dec" & (education=="tertiary" | education== "secondary")) # keeps only the customers contacted in december and with a secondary or tertiary education

# Customised summaries
by(bank_data$balance, bank_data$job, mean)
aggregate(data.frame(balance=bank_data$balance), list(job=bank_data$job), range) # to change labels of the output table 

#Adding data (column)
gender<-sample(c(1,2), dim(bank_data)[1], replace=TRUE) 
gender<-as.factor(gender)
bank_data_updated <- cbind (bank_data,gender)

#Exporting data
write.table(bank_data_updated, file="bank_data_updated.csv", quote=FALSE, sep=";",row.names=FALSE)


# Graphics
#Hist
#Raw scale
hist(bank_data$balance,breaks=100,xlab="Customer balance", 
     main= "Customer balance distribution", freq = FALSE, col= "green", ylim=c(0, 8e-04))
lines(density(bank_data$balance),col="orange3",lwd=2)


#log scale
hist(log10(bank_data$balance[bank_data$balance>0]),breaks=100,xlab="Customer balance (log scale)", 
     main= "Customer balance distribution (log scale)", freq = FALSE, col= "green"); 
lines(density(log10(bank_data$balance[bank_data$balance>0])),col="orange3",lwd=2, na.rm=FALSE)


#boxplot
boxplot(duration ~ contact, data=bank_data, main= "Call duration per contact communication type", 
        xlab = "Contact communication type", ylab= "Call duration (sec)")
points(duration ~ contact, data=bank_data, col="red", pch = 3)  #adds the actual data points onto the plot

boxplot(duration ~ contact, data=bank_data, main= "Call duration per contact communication type",
        ylab = "Contact communication type", xlab= "Call duration (sec)", 
        horizontal=TRUE, # horizontal plot
        las=0, # default style of axis labels (always parallel to axis), see more ?par
        cex.axis=1.2, cex.lab=1.5, cex.main=2) # diff magnification for axis annotation, labels and titles#

#plot with regression line and correlation, colouring by factor
plot(bank_data$age,bank_data$balance, type="p", col="red", main="Balance as function of customers' age" , xlab="Age (years)",
     ylab="Balance (euros)")
abline(lm(bank_data$balance~bank_data$age),col="blue")
legend("topright", legend= c("clients","fit"), pch= c("o","-"),col = c(2, 4),bg = "gray90")


plot(bank_data$age,bank_data$balance, type="p", main="Balance as function of customers' age and marital status" , xlab="Age (years)",
     ylab="Balance (euros)", col = c("red", "green3", "blue")[bank_data$marital])
legend("topright", legend= levels(bank_data$marital), pch= c("o","o","o"),col = c("red", "green3", "blue"),bg = "gray90")

shapiro.test(bank_data$balance) # data not normal 
shapiro.test(bank_data$age) # data not normal


plot(log10(bank_data$age[bank_data$balance>0]),log10(bank_data$balance[bank_data$balance>0]), type="p", col="red", main="Log balance as function of log customers' age" , xlab="Log10 age (years)",
     ylab="Log10 balance (euros)") # removing negative and null balances
abline(lm(log10(bank_data$balance[bank_data$balance>0])~log10(bank_data$age[bank_data$balance>0])),col="blue")
shapiro.test(log10(bank_data$balance[bank_data$balance>0])) # data not normal but better
shapiro.test(log10(bank_data$age)) # data not normal but better

cor.test(bank_data$age,bank_data$balance, method="spearman") # 0.08 signficant correlation
cor.test(bank_data$age,bank_data$balance) 


#####ggplot2
library("ggplot2")
###scatter plots
balance_age_plot<-ggplot(data=bank_data, aes(x=age, y=balance))
balance_age_plot + geom_point() 
balance_age_plot + geom_point(alpha=0.3)
balance_age_plot + geom_point(alpha=0.3, color="blue") 
balance_age_plot + geom_point(alpha=0.3, color="blue", size=3)

# Colouring and shaping by levels of the factor 
balance_age_plot + geom_point(alpha=0.3, aes(color=marital)) + ggtitle("Account balance as function of age and marital status") +
  theme(plot.title = element_text(size=10, face="bold", margin= margin(10,0,10,0))) + 
  theme(plot.title = element_text(hjust = 0.45)) 

#remove bakground panel
balance_age_plot + geom_point(alpha=0.3, aes(color=marital, shape=marital)) + ggtitle("Account balance as function of age and marital status") +
  theme(plot.title = element_text(size=10, face="bold", margin= margin(10,0,10,0))) + 
  theme(plot.title = element_text(hjust = 0.45))  + theme(panel.background = element_rect(fill = "white"))

#add regression line
balance_age_plot + geom_point(alpha=0.3, aes(color=marital, shape=marital)) + ggtitle("Account balance as function of age and marital status") +
  theme(plot.title = element_text(size=10, face="bold", margin= margin(10,0,10,0), hjust = 0.45)) + stat_smooth(method="lm")

#boxplot
duration_contact_boxplot<-ggplot(data=bank_data, aes(x=contact, y=duration)) 
duration_contact_boxplot + geom_boxplot()
duration_contact_boxplot + geom_boxplot(alpha=1)
duration_contact_boxplot + geom_boxplot(alpha=0)
duration_contact_boxplot + geom_boxplot(alpha=0) + geom_jitter(alpha = 0.3, color = "tomato")
duration_contact_boxplot + geom_jitter(alpha = 0.3, color = "tomato") + geom_boxplot(alpha=0) +
  ggtitle("Call duration as function of contact type")

#histogram
#frequencies
balance_histo<-ggplot(data=bank_data, aes(x=balance)) + geom_histogram() + 
  ggtitle("Account balance distribution")

#density
balance_histo_density<-ggplot(data=bank_data, aes(x=balance,..density..)) + geom_histogram(bins=70, colour="black",fill ="green") + 
  geom_density(color="orange") + ggtitle("Account balance distribution")

#saving ggplot
ggsave(balance_histo_density, file="hist_density_ggplot.pdf", width=10, height=5)