import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
%matplotlib inline

human_resources = sns.load_dataset("humanresources")
print(human_resources.head(n=4))

#Creating a PointPlot
sns.set(style="whitegrid")

g = sns.factorplot(x="time_spend_company", y="satisfaction_level", hue="salary", data=human_resources,
                   size=5, kind="point", palette="pastel", ci=95, dodge=True, join=False)
g.despine(left=True)
g.set_ylabels("Satisfaction Level")
g.set_xlabels("")
plt.title('Pointplot: Years in the Company against Satisfaction Level')
plt.show()

#Employees who have worked between 3-6 years tend to have the lowest satisfaction levels,
#vs those who have been there for <3 years, or more than 6 years.
#Data also indicates that in general, those in the high income salary do not tend to have higher satisfaction levels.
#Contrary to popular belief that money is a primary factor affecting employee job satisfaction!

sns.set(style="darkgrid")
cp = sns.factorplot(x="promotion_last_5years", y="satisfaction_level", hue="salary", data=human_resources,
                   size=5, kind="bar", palette="pastel", ci=95)
cp.despine(left=True)
cp.set_ylabels("Job Satisfaction")
cp.set_xlabels("Promotion in the last 5 years")
plt.title('Barplot: Career progression against Job Satisfaction')
plt.show()

#Generally, employees who got promoted in the last 5 years tend to indicate higher job satisfaction.
#This change was however minimal for those in the high salary bracket.
#Further indicating that money is not the primary factor affecting job satisfaction!

sns.set(style="darkgrid")
cp = sns.factorplot(x="salary", y="average_monthly_hours", hue="left", data=human_resources,
                   size=5, kind="bar", palette="pastel", ci=95)
cp.despine(left=True)
cp.set_ylabels("Average Monthly Hours at Work Place")
cp.set_xlabels("Salary Brackets")
plt.title('Barplot: Time spent at work place against Salary')
plt.show()
#Generally people who left in the low and medium income brackets tend to have worked longer hours.
#Except for those in the high salary brackets (once again!). In fact these group of people actually worked shorter hours.

# Showing the  information in a seaborn facet grid.
sns.set(style="ticks")  #Setting the overall aesthetic

# Tell seaborn about the structure of our data.
g = sns.FacetGrid(human_resources, row="department", col="salary")
# Name the plot type and the variable to be plotted using the structure.
g.map(plt.hist, "satisfaction_level", color="steelblue",  lw=0)
# Moving the plots apart to make room for our titles.
plt.subplots_adjust(top=0.9)
# Making a more informative axis name.
g.set_axis_labels('Satisfaction Levels')
plt.suptitle('Satisfaction Levels by Department and Salary')
# Removing excess lines around the plot.
sns.despine(trim=True)
plt.show()

#FacetGrid shows the satisfaction breakdown by department, and salary.
#Sales had the largest number of employees, followed by technical and support staff
#Data shows employees were moderate to adequately satisfied.

fig = plt.figure(figsize=(6,4))
ax=sns.kdeplot(human_resources.loc[(human_resources['left'] == 0),'average_monthly_hours'] , color='b',shade=True, label='Stayed')
ax=sns.kdeplot(human_resources.loc[(human_resources['left'] == 1),'average_monthly_hours'] , color='r',shade=True, label='Left')
plt.title('Average monthly hours worked')
#People who left either worked little hours, or worked more than 250 hours.
#Employees who stayed worked an average of 150-250 hours.