View AttritionAnalysis on GitHub

Exploratory Data Analysis

Stuart Miller August 4, 2019

Setup Environment

# import libraries
library(knitr)
library(tidyverse)
library(naniar)
library(Hmisc)
library(GGally)
library(gridExtra)
library(RColorBrewer)
library(gplots)
library(corrplot)
library(ggthemes)

# import helper functions
source('../helper/data_munging.R')
source('../helper/visual.R')

# read in data
train <- read_csv('../data/CaseStudy2-data_train.csv')

Data Exploration

A data dictionary was not given. So start by exploring the variables to get a sense of what is included. A simple data dictionary will be built. See ‘./analysis/data/README.md’ for more details.

names(train)

##  [1] "ID"                       "Age"                     
##  [3] "Attrition"                "BusinessTravel"          
##  [5] "DailyRate"                "Department"              
##  [7] "DistanceFromHome"         "Education"               
##  [9] "EducationField"           "EmployeeCount"           
## [11] "EmployeeNumber"           "EnvironmentSatisfaction" 
## [13] "Gender"                   "HourlyRate"              
## [15] "JobInvolvement"           "JobLevel"                
## [17] "JobRole"                  "JobSatisfaction"         
## [19] "MaritalStatus"            "MonthlyIncome"           
## [21] "MonthlyRate"              "NumCompaniesWorked"      
## [23] "Over18"                   "OverTime"                
## [25] "PercentSalaryHike"        "PerformanceRating"       
## [27] "RelationshipSatisfaction" "StandardHours"           
## [29] "StockOptionLevel"         "TotalWorkingYears"       
## [31] "TrainingTimesLastYear"    "WorkLifeBalance"         
## [33] "YearsAtCompany"           "YearsInCurrentRole"      
## [35] "YearsSinceLastPromotion"  "YearsWithCurrManager"

Reponse Variables

Two models are requested

  • A model for salary, which is given as MonthlyIncome in the dataset.
  • A model for attrition, which is given as Attrition in the dataset.

Histogram of monthly income reveals that it is highly skewed to the right.

# histogram of MonthlyIncome
train %>% ggplot(aes(x = MonthlyIncome)) +
  geom_histogram(bins = 15)


# histogram of MonthlyIncome
train %>% ggplot(aes(x = log(MonthlyIncome))) +
  geom_histogram(bins = 15)

A bar plot of Attrition reveals that the response variable is highly imbalanced.

# histogram of Attrition
train %>% ggplot(aes(x = Attrition)) +
  geom_bar()

Explanatory - Continuous Variables

Exploration of the continuous variables.

# create a vector of numeric features
features.numeric <- c('DailyRate', 'DistanceFromHome', 'Age', 'HourlyRate', 'MonthlyIncome', 'MonthlyRate',
           'NumCompaniesWorked','PercentSalaryHike', 'TotalWorkingYears', 'TrainingTimesLastYear',
           'YearsAtCompany','YearsInCurrentRole','YearsSinceLastPromotion', 'YearsWithCurrManager')

Explore features in isolation and the relationship between features and the response variables.

Correlation to Response

The following table ranks features in order of descending correlation to MonthlyIncome.

train.numeric <- train %>% select(features.numeric)

correlation.matrix <- rcorr(as.matrix(train.numeric))
train.corToMI <- data.frame(flattenCorrMatrix(correlation.matrix$r, correlation.matrix$P)) %>% 
  filter(row == 'MonthlyIncome' | column == 'MonthlyIncome') %>%
  mutate(cor = abs(cor)) %>%
  arrange(-cor)
kable(train.corToMI)
row column cor p
MonthlyIncome TotalWorkingYears 0.7785112 0.0000000
MonthlyIncome YearsAtCompany 0.4913790 0.0000000
Age MonthlyIncome 0.4842883 0.0000000
MonthlyIncome YearsInCurrentRole 0.3618405 0.0000000
MonthlyIncome YearsWithCurrManager 0.3284875 0.0000000
MonthlyIncome YearsSinceLastPromotion 0.3159116 0.0000000
MonthlyIncome NumCompaniesWorked 0.1558943 0.0000038
MonthlyIncome MonthlyRate 0.0645941 0.0568449
MonthlyIncome PercentSalaryHike 0.0538659 0.1123575
MonthlyIncome TrainingTimesLastYear 0.0390146 0.2503303
DistanceFromHome MonthlyIncome 0.0066672 0.8443188
HourlyRate MonthlyIncome 0.0023912 0.9438534
DailyRate MonthlyIncome 0.0000879 0.9979342

Correlation Heatmap

TotalworkingYears, Age, YearsAtCompany, YearsInCurrentRole, and YearsWithCurrentManager.

heatmap.cor <- function(df){
  df %>%
    keep(is.numeric) %>%
    drop_na() %>%
    cor %>%
    corrplot( addCoef.col = 'white',
             number.digits = 2,
             number.cex = 0.5,
             method = 'square')
}
heatmap.cor(train.numeric)

Variance of Features

The following table ranks features in order of descending standard deviation.

# get varaince table
temp.table <- train %>%
  select(features.numeric) %>%
  summarise_all(funs(sd(.))) %>%
  rownames_to_column %>%
  gather(var, value, -rowname) %>%
  arrange(-value) %>%
  select(-one_of(c('rowname')))
# rename columns for clairity and print markdown table
names(temp.table) <- c('Feature','Standard Deviation')
kable(temp.table)
Feature Standard Deviation
MonthlyRate 7108.381928
MonthlyIncome 4597.695974
DailyRate 401.116280
HourlyRate 20.127163
Age 8.925950
DistanceFromHome 8.136704
TotalWorkingYears 7.513668
YearsAtCompany 6.021036
PercentSalaryHike 3.675440
YearsInCurrentRole 3.639317
YearsWithCurrManager 3.574441
YearsSinceLastPromotion 3.185872
NumCompaniesWorked 2.520443
TrainingTimesLastYear 1.272665

Density Plots of All Numeric Features

train %>% keep(is.numeric) %>% 
  gather() %>% 
  ggplot(aes(x = value)) + 
  facet_wrap(~ key, scales = 'free') + 
  geom_density()

Age exploration

There appears to be a relationship between Age and MonthlyIncome. A relationship between Age and Attrition is not clear.

train %>% ggplot(aes(x = Age)) + 
  geom_histogram(bins = 15)


train %>% ggplot(aes(x = Age)) + 
  geom_histogram(bins = 15) + 
  facet_wrap(~ Attrition)


train %>% ggplot(aes(x = Age, y = MonthlyIncome)) + 
  geom_point() + geom_smooth(method = 'lm')

DailyRate exploration

There does not appear to be a correlation between DailyRate and either outcome variable.

train %>% ggplot(aes(x = DailyRate)) + 
  geom_histogram(bins = 15)


train %>% ggplot(aes(x = DailyRate)) + 
  geom_histogram(bins = 15) + 
  facet_wrap(~ Attrition)


train %>% ggplot(aes(x = DailyRate, y = MonthlyIncome)) + 
  geom_point() + geom_smooth(method = 'lm')

DistanceFromHome exploration

There does not appear to be a correlation between DistanceFromHome and either outcome variable.

train %>% ggplot(aes(x = DistanceFromHome)) + 
  geom_histogram(bins = 15)


train %>% ggplot(aes(x = DistanceFromHome)) + 
  geom_histogram(bins = 15) + 
  facet_wrap(~ Attrition)


train %>% ggplot(aes(x = DistanceFromHome, y = MonthlyIncome)) + 
  geom_point() + geom_smooth(method = 'lm')

MonthlyRate exploration

There does not appear to be a correlation between MonthlyRate and either outcome variable.

train %>% ggplot(aes(x = MonthlyRate)) + 
  geom_histogram(bins = 15)


train %>% ggplot(aes(x = MonthlyRate)) + 
  geom_histogram(bins = 15) + 
  facet_wrap(~ Attrition)


train %>% ggplot(aes(x = MonthlyRate, y = MonthlyIncome)) + 
  geom_point() + geom_smooth(method = 'lm')

PercentSalaryHike exploration

There does not appear to be a correlation between PercentSalaryHike and either outcome variable.

train %>% ggplot(aes(x = PercentSalaryHike)) + 
  geom_histogram()

## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.


train %>% ggplot(aes(x = PercentSalaryHike)) + 
  geom_histogram() + 
  facet_wrap(~ Attrition)

## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.


train %>% ggplot(aes(x = PercentSalaryHike, y = MonthlyIncome)) + 
  geom_point() + geom_smooth(method = 'lm')

TotalWorkingYears exploration

There does appear to be a correlation between TotalWorkingYears and MonthlyIncome.

train %>% ggplot(aes(x = TotalWorkingYears)) + 
  geom_histogram(bins = 15)


train %>% ggplot(aes(x = TotalWorkingYears)) + 
  geom_histogram(bins = 15) + 
  facet_wrap(~ Attrition)


train %>% ggplot(aes(x = TotalWorkingYears, y = MonthlyIncome)) + 
  geom_point() + geom_smooth(method = 'lm')

TrainingTimesLastYear exploration

There does not appear to be a correlation between TrainingTimesLastYear and either outcome variable.

train %>% ggplot(aes(x = TrainingTimesLastYear)) + 
  geom_histogram(bins = 15)


train %>% ggplot(aes(x = TrainingTimesLastYear)) + 
  geom_histogram(bins = 15) + 
  facet_wrap(~ Attrition)


train %>% ggplot(aes(x = TrainingTimesLastYear, y = MonthlyIncome)) + 
  geom_point() + geom_smooth(method = 'lm')

YearsAtCompany exploration

There does appear to be a correlation between YearsAtCompany and MonthlyIncome.

train %>% ggplot(aes(x = YearsAtCompany)) + 
  geom_histogram(bins = 15)


train %>% ggplot(aes(x = YearsAtCompany)) + 
  geom_histogram(bins = 15) + 
  facet_wrap(~ Attrition)


train %>% ggplot(aes(x = YearsAtCompany, y = MonthlyIncome)) + 
  geom_point() + geom_smooth(method = 'lm')

YearsInCurrentRole exploration

There does appear to be a correlation between YearsInCurrentRole and MonthlyIncome.

train %>% ggplot(aes(x = YearsInCurrentRole)) + 
  geom_histogram(bins = 15)


train %>% ggplot(aes(x = YearsInCurrentRole)) + 
  geom_histogram(bins = 15) + 
  facet_wrap(~ Attrition)


train %>% ggplot(aes(x = YearsInCurrentRole, y = MonthlyIncome)) + 
  geom_point() + geom_smooth(method = 'lm')

YearsSinceLastPromotion exploration

There does appear to be a correlation between YearsSinceLastPromotion and MonthlyIncome.

train %>% ggplot(aes(x = YearsSinceLastPromotion)) + 
  geom_histogram(bins = 15)


train %>% ggplot(aes(x = YearsSinceLastPromotion)) + 
  geom_histogram(bins = 15) + 
  facet_wrap(~ Attrition)


train %>% ggplot(aes(x = YearsSinceLastPromotion, y = MonthlyIncome)) + 
  geom_point() + geom_smooth(method = 'lm')

YearsWithCurrManager exploration

There does not appear to be a correlation between YearsWithCurrManager and either outcome variable.

train %>% ggplot(aes(x = YearsWithCurrManager)) + 
  geom_histogram(bins = 15)


train %>% ggplot(aes(x = YearsWithCurrManager)) + 
  geom_histogram(bins = 15) + 
  facet_wrap(~ Attrition)


train %>% ggplot(aes(x = YearsWithCurrManager, y = MonthlyIncome)) + 
  geom_point() + geom_smooth(method = 'lm')

Multivariate Exploration

Scatter plot and correlation matrix gives a board overview of correlations between continuous variables.

  • MonthlyIncome appears to be correlated with Age and TotalWorkingYears
  • YearsAtCompany,YearsInCurrentRole,YearsSinceLastPromotion, YearsWithCurrManager, TotalWorkingYears, and Age appear to be correlated.
train %>%
  select(c(features.numeric)) %>%
  ggpairs()

Look at log of MonthlyIncome logMI because MonthlyIncome is right skewed

train <- train %>% mutate(logMI = log(MonthlyIncome))
train %>%
  select(c(features.numeric, 'logMI')) %>%
  ggpairs()

Close up of MonthlyIncome and highest correlated features

Taking the log of MonthlyIncome does not appear to improve linear correlation.

train %>%
  select(c('logMI','MonthlyIncome','Age','TotalWorkingYears','YearsAtCompany','YearsInCurrentRole',
           'YearsSinceLastPromotion','YearsWithCurrManager')) %>%
  ggpairs()

From the plot above, several of the variables appear to be linearly related: YearsAtCompany vs YearsWithCurrManager, YearsInCurrentRole and YearsSinceLastPromotion. This could be an interesting trend in the data.

p1 <- train %>% filter(YearsAtCompany > 0) %>%
  ggplot(aes(x = YearsWithCurrManager, y = YearsAtCompany)) + 
  geom_point() + geom_smooth(method = 'lm')
p2 <- train %>% filter(YearsAtCompany > 0) %>%
  ggplot(aes(x = YearsInCurrentRole, y = YearsAtCompany)) + 
  geom_point() + geom_smooth(method = 'lm')
p3 <- train %>% filter(YearsAtCompany > 0) %>%
  ggplot(aes(x = YearsSinceLastPromotion, y = YearsAtCompany)) + 
  geom_point() + geom_smooth(method = 'lm')
grid.arrange(p1,p2,p3, ncol = 2)


m.YearsAtCompany <- train %>% 
  filter(YearsAtCompany > 0) %>% 
  lm(log(YearsAtCompany) ~ YearsSinceLastPromotion +  YearsInCurrentRole + YearsWithCurrManager, data = .)

summary(m.YearsAtCompany)

## 
## Call:
## lm(formula = log(YearsAtCompany) ~ YearsSinceLastPromotion + 
##     YearsInCurrentRole + YearsWithCurrManager, data = .)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -1.13397 -0.36114  0.02062  0.30107  1.57553 
## 
## Coefficients:
##                         Estimate Std. Error t value Pr(>|t|)    
## (Intercept)             0.608360   0.024460  24.872  < 2e-16 ***
## YearsSinceLastPromotion 0.022474   0.005623   3.997 6.99e-05 ***
## YearsInCurrentRole      0.108390   0.006035  17.960  < 2e-16 ***
## YearsWithCurrManager    0.115189   0.005946  19.373  < 2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.4311 on 838 degrees of freedom
## Multiple R-squared:  0.767,  Adjusted R-squared:  0.7662 
## F-statistic: 919.7 on 3 and 838 DF,  p-value: < 2.2e-16

train %>% filter(YearsAtCompany > 0) %>%basic.fit.plots(., m.YearsAtCompany)

Explanatory - factor Variables

Exploration of the continuous variables.

For investigation of the categrical variables, we will look at variables correlated to monthly income. Monthly income is correlated with TotalworkingYears, Age, YearsAtCompany, YearsInCurrentRole, and YearsWithCurrentManager.

# create a vector of numeric features
features.factor <- c('BusinessTravel', 'Department', 'Education', 'EducationField', 'EmployeeNumber', 'EnvironmentSatisfaction', 'Gender', 'JobInvolvement', 'JobLevel', 'JobRole', 'JobSatisfaction', 'MaritalStatus', 'OverTime', 'PerformanceRating', 'RelationshipSatisfaction', 'StockOptionLevel', 'WorkLifeBalance')

# factor categorical variables
train[, features.factor] <- lapply(train[, features.factor], as.factor)

BusinessTravel

btMI <- train %>% 
  ggplot(aes(x = BusinessTravel,
             y = log(MonthlyIncome),
             fill = BusinessTravel)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btAg <- train %>% 
  ggplot(aes(x = BusinessTravel,
             y = Age,
             fill = BusinessTravel)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYAC <- train %>% 
  ggplot(aes(x = BusinessTravel,
             y = YearsAtCompany,
             fill = BusinessTravel)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYCM <- train %>% 
  ggplot(aes(x = BusinessTravel,
             y = YearsWithCurrManager,
             fill = BusinessTravel)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btTWY <- train %>% 
  ggplot(aes(x = BusinessTravel,
             y = TotalWorkingYears,
             fill = BusinessTravel)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

grid.arrange(btMI, btAg, btYAC, btYCM, btTWY)


btMI <- train %>% 
  ggplot(aes(x = Gender,
             y = log(MonthlyIncome),
             fill = Gender)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btAg <- train %>% 
  ggplot(aes(x = Gender,
             y = Age,
             fill = Gender)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYAC <- train %>% 
  ggplot(aes(x = Gender,
             y = YearsAtCompany,
             fill = Gender)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYCM <- train %>% 
  ggplot(aes(x = Gender,
             y = YearsWithCurrManager,
             fill = Gender)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btTWY <- train %>% 
  ggplot(aes(x = Gender,
             y = TotalWorkingYears,
             fill = Gender)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

grid.arrange(btMI, btAg, btYAC, btYCM, btTWY)


btMI <- train %>% 
  ggplot(aes(x = Department,
             y = log(MonthlyIncome),
             fill = Department)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btAg <- train %>% 
  ggplot(aes(x = Department,
             y = Age,
             fill = Department)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYAC <- train %>% 
  ggplot(aes(x = Department,
             y = YearsAtCompany,
             fill = Department)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYCM <- train %>% 
  ggplot(aes(x = Department,
             y = YearsWithCurrManager,
             fill = Department)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btTWY <- train %>% 
  ggplot(aes(x = Department,
             y = TotalWorkingYears,
             fill = Department)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

grid.arrange(btMI, btAg, btYAC, btYCM, btTWY)


btMI <- train %>% 
  ggplot(aes(x = Education,
             y = log(MonthlyIncome),
             fill = Education)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btAg <- train %>% 
  ggplot(aes(x = Education,
             y = Age,
             fill = Education)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYAC <- train %>% 
  ggplot(aes(x = Education,
             y = YearsAtCompany,
             fill = Education)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYCM <- train %>% 
  ggplot(aes(x = Education,
             y = YearsWithCurrManager,
             fill = Education)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btTWY <- train %>% 
  ggplot(aes(x = Education,
             y = TotalWorkingYears,
             fill = Education)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

grid.arrange(btMI, btAg, btYAC, btYCM, btTWY)


btMI <- train %>% 
  ggplot(aes(x = EducationField,
             y = log(MonthlyIncome),
             fill = EducationField)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btAg <- train %>% 
  ggplot(aes(x = EducationField,
             y = Age,
             fill = EducationField)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYAC <- train %>% 
  ggplot(aes(x = EducationField,
             y = YearsAtCompany,
             fill = EducationField)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYCM <- train %>% 
  ggplot(aes(x = EducationField,
             y = YearsWithCurrManager,
             fill = EducationField)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btTWY <- train %>% 
  ggplot(aes(x = EducationField,
             y = TotalWorkingYears,
             fill = EducationField)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

grid.arrange(btMI, btAg, btYAC, btYCM, btTWY)


btMI <- train %>% 
  ggplot(aes(x = EnvironmentSatisfaction,
             y = log(MonthlyIncome),
             fill = EnvironmentSatisfaction)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btAg <- train %>% 
  ggplot(aes(x = EnvironmentSatisfaction,
             y = Age,
             fill = EnvironmentSatisfaction)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYAC <- train %>% 
  ggplot(aes(x = EnvironmentSatisfaction,
             y = YearsAtCompany,
             fill = EnvironmentSatisfaction)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYCM <- train %>% 
  ggplot(aes(x = EnvironmentSatisfaction,
             y = YearsWithCurrManager,
             fill = EnvironmentSatisfaction)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btTWY <- train %>% 
  ggplot(aes(x = EnvironmentSatisfaction,
             y = TotalWorkingYears,
             fill = EnvironmentSatisfaction)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

grid.arrange(btMI, btAg, btYAC, btYCM, btTWY)


btMI <- train %>% 
  ggplot(aes(x = JobInvolvement,
             y = log(MonthlyIncome),
             fill = JobInvolvement)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btAg <- train %>% 
  ggplot(aes(x = JobInvolvement,
             y = Age,
             fill = JobInvolvement)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYAC <- train %>% 
  ggplot(aes(x = JobInvolvement,
             y = YearsAtCompany,
             fill = JobInvolvement)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYCM <- train %>% 
  ggplot(aes(x = JobInvolvement,
             y = YearsWithCurrManager,
             fill = JobInvolvement)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btTWY <- train %>% 
  ggplot(aes(x = JobInvolvement,
             y = TotalWorkingYears,
             fill = JobInvolvement)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

grid.arrange(btMI, btAg, btYAC, btYCM, btTWY)


btMI <- train %>% 
  ggplot(aes(x = JobLevel,
             y = log(MonthlyIncome),
             fill = JobLevel)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btAg <- train %>% 
  ggplot(aes(x = JobLevel,
             y = Age,
             fill = JobLevel)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYAC <- train %>% 
  ggplot(aes(x = JobLevel,
             y = YearsAtCompany,
             fill = JobLevel)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYCM <- train %>% 
  ggplot(aes(x = JobLevel,
             y = YearsWithCurrManager,
             fill = JobLevel)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btTWY <- train %>% 
  ggplot(aes(x = JobLevel,
             y = TotalWorkingYears,
             fill = JobLevel)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

grid.arrange(btMI, btAg, btYAC, btYCM, btTWY)


btMI <- train %>% 
  ggplot(aes(x = JobRole,
             y = log(MonthlyIncome),
             fill = JobRole)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few() + 
  theme(axis.text.x = element_text(angle = 60, hjust = 1))

btAg <- train %>% 
  ggplot(aes(x = JobRole,
             y = Age,
             fill = JobRole)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few() + 
  theme(axis.text.x = element_text(angle = 60, hjust = 1))

btYAC <- train %>% 
  ggplot(aes(x = JobRole,
             y = YearsAtCompany,
             fill = JobRole)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few() + 
  theme(axis.text.x = element_text(angle = 60, hjust = 1))

btYCM <- train %>% 
  ggplot(aes(x = JobRole,
             y = YearsWithCurrManager,
             fill = JobRole)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few() + 
  theme(axis.text.x = element_text(angle = 60, hjust = 1))

btTWY <- train %>% 
  ggplot(aes(x = JobRole,
             y = TotalWorkingYears,
             fill = JobRole)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few() + 
  theme(axis.text.x = element_text(angle = 60, hjust = 1))

grid.arrange(btMI, btAg, btYAC, btYCM, btTWY)

## Warning in check_pal_n(n, max_n): This palette can handle a maximum of 8
## values.You have supplied 9.

## Warning in check_pal_n(n, max_n): This palette can handle a maximum of 8
## values.You have supplied 9.

## Warning in check_pal_n(n, max_n): This palette can handle a maximum of 8
## values.You have supplied 9.

## Warning in check_pal_n(n, max_n): This palette can handle a maximum of 8
## values.You have supplied 9.

## Warning in check_pal_n(n, max_n): This palette can handle a maximum of 8
## values.You have supplied 9.


btMI <- train %>% 
  ggplot(aes(x = JobSatisfaction,
             y = log(MonthlyIncome),
             fill = JobSatisfaction)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btAg <- train %>% 
  ggplot(aes(x = JobSatisfaction,
             y = Age,
             fill = JobSatisfaction)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYAC <- train %>% 
  ggplot(aes(x = JobSatisfaction,
             y = YearsAtCompany,
             fill = JobSatisfaction)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYCM <- train %>% 
  ggplot(aes(x = JobSatisfaction,
             y = YearsWithCurrManager,
             fill = JobSatisfaction)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btTWY <- train %>% 
  ggplot(aes(x = JobSatisfaction,
             y = TotalWorkingYears,
             fill = JobSatisfaction)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

grid.arrange(btMI, btAg, btYAC, btYCM, btTWY)


btMI <- train %>% 
  ggplot(aes(x = MaritalStatus,
             y = log(MonthlyIncome),
             fill = MaritalStatus)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btAg <- train %>% 
  ggplot(aes(x = MaritalStatus,
             y = Age,
             fill = MaritalStatus)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYAC <- train %>% 
  ggplot(aes(x = MaritalStatus,
             y = YearsAtCompany,
             fill = MaritalStatus)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYCM <- train %>% 
  ggplot(aes(x = MaritalStatus,
             y = YearsWithCurrManager,
             fill = MaritalStatus)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btTWY <- train %>% 
  ggplot(aes(x = MaritalStatus,
             y = TotalWorkingYears,
             fill = MaritalStatus)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

grid.arrange(btMI, btAg, btYAC, btYCM, btTWY)


btMI <- train %>% 
  ggplot(aes(x = OverTime,
             y = log(MonthlyIncome),
             fill = OverTime)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btAg <- train %>% 
  ggplot(aes(x = OverTime,
             y = Age,
             fill = OverTime)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYAC <- train %>% 
  ggplot(aes(x = OverTime,
             y = YearsAtCompany,
             fill = OverTime)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYCM <- train %>% 
  ggplot(aes(x = OverTime,
             y = YearsWithCurrManager,
             fill = OverTime)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btTWY <- train %>% 
  ggplot(aes(x = OverTime,
             y = TotalWorkingYears,
             fill = OverTime)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

grid.arrange(btMI, btAg, btYAC, btYCM, btTWY)


btMI <- train %>% 
  ggplot(aes(x = PerformanceRating,
             y = log(MonthlyIncome),
             fill = PerformanceRating)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btAg <- train %>% 
  ggplot(aes(x = PerformanceRating,
             y = Age,
             fill = PerformanceRating)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYAC <- train %>% 
  ggplot(aes(x = PerformanceRating,
             y = YearsAtCompany,
             fill = PerformanceRating)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYCM <- train %>% 
  ggplot(aes(x = PerformanceRating,
             y = YearsWithCurrManager,
             fill = PerformanceRating)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btTWY <- train %>% 
  ggplot(aes(x = PerformanceRating,
             y = TotalWorkingYears,
             fill = PerformanceRating)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

grid.arrange(btMI, btAg, btYAC, btYCM, btTWY)


btMI <- train %>% 
  ggplot(aes(x = RelationshipSatisfaction,
             y = log(MonthlyIncome),
             fill = RelationshipSatisfaction)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btAg <- train %>% 
  ggplot(aes(x = RelationshipSatisfaction,
             y = Age,
             fill = RelationshipSatisfaction)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYAC <- train %>% 
  ggplot(aes(x = RelationshipSatisfaction,
             y = YearsAtCompany,
             fill = RelationshipSatisfaction)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYCM <- train %>% 
  ggplot(aes(x = RelationshipSatisfaction,
             y = YearsWithCurrManager,
             fill = RelationshipSatisfaction)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btTWY <- train %>% 
  ggplot(aes(x = RelationshipSatisfaction,
             y = TotalWorkingYears,
             fill = RelationshipSatisfaction)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

grid.arrange(btMI, btAg, btYAC, btYCM, btTWY)


btMI <- train %>% 
  ggplot(aes(x = StockOptionLevel,
             y = log(MonthlyIncome),
             fill = StockOptionLevel)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btAg <- train %>% 
  ggplot(aes(x = StockOptionLevel,
             y = Age,
             fill = StockOptionLevel)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYAC <- train %>% 
  ggplot(aes(x = StockOptionLevel,
             y = YearsAtCompany,
             fill = StockOptionLevel)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYCM <- train %>% 
  ggplot(aes(x = StockOptionLevel,
             y = YearsWithCurrManager,
             fill = StockOptionLevel)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btTWY <- train %>% 
  ggplot(aes(x = StockOptionLevel,
             y = TotalWorkingYears,
             fill = StockOptionLevel)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

grid.arrange(btMI, btAg, btYAC, btYCM, btTWY)


btMI <- train %>% 
  ggplot(aes(x = WorkLifeBalance,
             y = log(MonthlyIncome),
             fill = WorkLifeBalance)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btAg <- train %>% 
  ggplot(aes(x = WorkLifeBalance,
             y = Age,
             fill = WorkLifeBalance)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYAC <- train %>% 
  ggplot(aes(x = WorkLifeBalance,
             y = YearsAtCompany,
             fill = WorkLifeBalance)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYCM <- train %>% 
  ggplot(aes(x = WorkLifeBalance,
             y = YearsWithCurrManager,
             fill = WorkLifeBalance)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btTWY <- train %>% 
  ggplot(aes(x = WorkLifeBalance,
             y = TotalWorkingYears,
             fill = WorkLifeBalance)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

grid.arrange(btMI, btAg, btYAC, btYCM, btTWY)


btMI <- train %>% 
  ggplot(aes(x = Attrition,
             y = log(MonthlyIncome),
             fill = Attrition)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btAg <- train %>% 
  ggplot(aes(x = Attrition,
             y = Age,
             fill = Attrition)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYAC <- train %>% 
  ggplot(aes(x = Attrition,
             y = YearsAtCompany,
             fill = Attrition)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btYCM <- train %>% 
  ggplot(aes(x = Attrition,
             y = YearsWithCurrManager,
             fill = Attrition)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

btTWY <- train %>% 
  ggplot(aes(x = Attrition,
             y = TotalWorkingYears,
             fill = Attrition)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

grid.arrange(btMI, btAg, btYAC, btYCM, btTWY)

All Numeric Features vs Attrition

features.numeric

##  [1] "DailyRate"               "DistanceFromHome"       
##  [3] "Age"                     "HourlyRate"             
##  [5] "MonthlyIncome"           "MonthlyRate"            
##  [7] "NumCompaniesWorked"      "PercentSalaryHike"      
##  [9] "TotalWorkingYears"       "TrainingTimesLastYear"  
## [11] "YearsAtCompany"          "YearsInCurrentRole"     
## [13] "YearsSinceLastPromotion" "YearsWithCurrManager"

train %>% 
  ggplot(aes(x = DailyRate)) +
  geom_histogram(bins = 15) +
  facet_wrap( ~ Attrition, scales = 'free') + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()


train %>% 
  ggplot(aes(x = DistanceFromHome)) +
  geom_histogram(bins = 15) +
  facet_wrap( ~ Attrition, scales = 'free') + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()


train %>% 
  ggplot(aes(x = Age)) +
  geom_histogram(bins = 15) +
  facet_wrap( ~ Attrition, scales = 'free') + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()


train %>% 
  ggplot(aes(x = HourlyRate)) +
  geom_histogram(bins = 15) +
  facet_wrap( ~ Attrition, scales = 'free') + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()


train %>% 
  ggplot(aes(x = MonthlyIncome)) +
  geom_histogram(bins = 15) +
  facet_wrap( ~ Attrition, scales = 'free') + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()


train %>% 
  ggplot(aes(x = MonthlyRate)) +
  geom_histogram(bins = 15) +
  facet_wrap( ~ Attrition, scales = 'free') + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()


train %>% 
  ggplot(aes(x = NumCompaniesWorked)) +
  geom_histogram(bins = 15) +
  facet_wrap( ~ Attrition, scales = 'free') + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()


train %>% 
  ggplot(aes(x = PercentSalaryHike)) +
  geom_histogram(bins = 15) +
  facet_wrap( ~ Attrition, scales = 'free') + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()


train %>% 
  ggplot(aes(x = TotalWorkingYears)) +
  geom_histogram(bins = 15) +
  facet_wrap( ~ Attrition, scales = 'free') + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()


train %>% 
  ggplot(aes(x = TrainingTimesLastYear)) +
  geom_histogram(bins = 15) +
  facet_wrap( ~ Attrition, scales = 'free') + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()


train %>% 
  ggplot(aes(x = YearsAtCompany)) +
  geom_histogram(bins = 15) +
  facet_wrap( ~ Attrition, scales = 'free') + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()


train %>% 
  ggplot(aes(x = YearsInCurrentRole)) +
  geom_histogram(bins = 15) +
  facet_wrap( ~ Attrition, scales = 'free') + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()


train %>% 
  ggplot(aes(x = YearsSinceLastPromotion)) +
  geom_histogram(bins = 15) +
  facet_wrap( ~ Attrition, scales = 'free') + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()


train %>% 
  ggplot(aes(x = YearsWithCurrManager)) +
  geom_histogram(bins = 15) +
  facet_wrap( ~ Attrition, scales = 'free') + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

All Factor Features vs Attrition

features.factor

##  [1] "BusinessTravel"           "Department"              
##  [3] "Education"                "EducationField"          
##  [5] "EmployeeNumber"           "EnvironmentSatisfaction" 
##  [7] "Gender"                   "JobInvolvement"          
##  [9] "JobLevel"                 "JobRole"                 
## [11] "JobSatisfaction"          "MaritalStatus"           
## [13] "OverTime"                 "PerformanceRating"       
## [15] "RelationshipSatisfaction" "StockOptionLevel"        
## [17] "WorkLifeBalance"

train %>% ggplot(aes(x = BusinessTravel, fill = Attrition)) +
  geom_bar(position = 'fill') +
  coord_flip()


train %>% ggplot(aes(x = Department, fill = Attrition)) +
  geom_bar(position = 'fill') +
  coord_flip()

There appears to be some difference between education levels and attition.

train %>% ggplot(aes(x = Education, fill = Attrition)) +
  geom_bar(position = 'fill') +
  coord_flip()

There appears to be some difference between education fields and attition.

train %>% ggplot(aes(x = EducationField, fill = Attrition)) +
  geom_bar(position = 'fill') +
  coord_flip()

There appears to be some difference between environment satisfaction and attition.

train %>% ggplot(aes(x = EnvironmentSatisfaction, fill = Attrition)) +
  geom_bar(position = 'fill') +
  coord_flip()


train %>% ggplot(aes(x = Gender, fill = Attrition)) +
  geom_bar(position = 'fill') +
  coord_flip()

Job involvement appears to have a big impact on attiriton

train %>% ggplot(aes(x = JobInvolvement, fill = Attrition)) +
  geom_bar(position = 'fill') +
  coord_flip()

There is a correlation between job level and attrition

train %>% ggplot(aes(x = JobLevel, fill = Attrition)) +
  geom_bar(position = 'fill') +
  coord_flip()

There is a correlation between job role and attrition

train %>% ggplot(aes(x = JobRole, fill = Attrition)) +
  geom_bar(position = 'fill') +
  coord_flip()

Job satisfaction appears to affect attrition

train %>% ggplot(aes(x = JobSatisfaction, fill = Attrition)) +
  geom_bar(position = 'fill') +
  coord_flip()

Marital status also correlates with attrition

train %>% ggplot(aes(x = MaritalStatus, fill = Attrition)) +
  geom_bar(position = 'fill') +
  coord_flip()

There appears to be a difference in attition related to overtime

train %>% ggplot(aes(x = OverTime, fill = Attrition)) +
  geom_bar(position = 'fill') +
  coord_flip()


train %>% ggplot(aes(x = PerformanceRating, fill = Attrition)) +
  geom_bar(position = 'fill') +
  coord_flip()


train %>% ggplot(aes(x = RelationshipSatisfaction, fill = Attrition)) +
  geom_bar(position = 'fill') +
  coord_flip()

Stock option level appears to play a role in attition

train %>% ggplot(aes(x = StockOptionLevel, fill = Attrition)) +
  geom_bar(position = 'fill') +
  coord_flip()

Work life balance appears to play a role in attirition

train %>% ggplot(aes(x = WorkLifeBalance, fill = Attrition)) +
  geom_bar(position = 'fill') +
  coord_flip()

All Numeric Features vs Attrition

p1 <- train %>% 
  ggplot(aes(x = Attrition,
             y = log(DailyRate),
             fill = Attrition)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

p2 <- train %>% 
  ggplot(aes(x = Attrition,
             y = DistanceFromHome,
             fill = Attrition)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

p3 <- train %>% 
  ggplot(aes(x = Attrition,
             y = Age,
             fill = Attrition)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

p4 <- train %>% 
  ggplot(aes(x = Attrition,
             y = HourlyRate,
             fill = Attrition)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

p5 <- train %>% 
  ggplot(aes(x = Attrition,
             y = MonthlyIncome,
             fill = Attrition)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

p6 <- train %>% 
  ggplot(aes(x = Attrition,
             y = MonthlyRate,
             fill = Attrition)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

p7 <- train %>% 
  ggplot(aes(x = Attrition,
             y = NumCompaniesWorked,
             fill = Attrition)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

p8 <- train %>% 
  ggplot(aes(x = Attrition,
             y = PercentSalaryHike,
             fill = Attrition)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

p9 <- train %>% 
  ggplot(aes(x = Attrition,
             y = TotalWorkingYears,
             fill = Attrition)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

p10 <- train %>% 
  ggplot(aes(x = Attrition,
             y = TrainingTimesLastYear,
             fill = Attrition)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

p11 <- train %>% 
  ggplot(aes(x = Attrition,
             y = YearsAtCompany,
             fill = Attrition)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

p12 <- train %>% 
  ggplot(aes(x = Attrition,
             y = YearsInCurrentRole,
             fill = Attrition)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

p13 <- train %>% 
  ggplot(aes(x = Attrition,
             y = YearsSinceLastPromotion,
             fill = Attrition)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()

p14 <- train %>% 
  ggplot(aes(x = Attrition,
             y = YearsWithCurrManager,
             fill = Attrition)) +
  geom_boxplot() + 
  scale_fill_few(palette = 'Dark') + 
  theme_few()
grid.arrange(p1, p2, p3, p4, p5, p6, p7,
             p8, p9, p10, p11, p12, p13, p14)