• Importance of visualization in the analytics industry
• How to decide on which chart to use
• Introduction and background to matplotlib and seaborn in python
• A series of graphs and visualization using python to answer relevant questions from a real-world data
• Best practices of storytelling using visualization

1. For exploratory data analysis popularly referred to as the data EDA
2. Communicate topline findings clearly to different stakeholders
3. Share unbiased representation of the data
4. Using visualization to support findings, insights, and recommendations

• Data charts, also called quantitative charts, depict numbers graphically to make a point. They include pie charts, bar graphs, stacked bar graphs, scatter plots, line charts, etc.
• Concept charts, also called non-quantitative charts, use words and images. Concept charts describe a situation, such as interaction, interrelationship, leverage or forces at work. Some common examples are process flow charts, gantt charts, matrix, etc.

# — — — — — — — — — — — — - Backend Layer, Figure Canvas Layer: Encompases area in which figures are drawn — — — — — — — — — — — — -
from matplotlib.backends.backend_agg import FigureCanvasAgg as FigureCanvas
#— — — — — — — — — — — — -From Artist layer, importing Artist Figure: Artist layer knows how to use a renderer to draw an object----------------------
from matplotlib.figure import Figure
fig=Figure()
canvas=FigureCanvas(fig) #-----passing object of artist layer to backend layer (Figure Canvas)
#— — — — — — — — — — — — -Using numpy to create random variables----------------------
import numpy as np
x=np.random.rand(100)
#— — — — — — — — — — — — -Plotting a histogram---------------------
ax=fig.add_subplot(111)
ax.hist(x,100)
ax.set_title('Normal Distribution')
fig.savefig('matplotlib_histogram.png')

import matplotlib.pyplot as plt
#— — — — — — — — — — — — -Using numpy to create random variables----------------------
import numpy as np
x=np.random.rand(100)
plt.hist(x,100) #-----100 refers to the number of bins
plt.title(‘Normal distribution Graph’)
plt.xlabel(‘Random numbers generated’)
plt.ylabel(‘Density’)
plt.show()

#— — — — — — — — — — — — -Reading dataset — — — — — — — — — — — -
import pandas as pd
suicide_data=pd.read_csv(‘C:/Users/91905/Desktop/master.csv’)
suicide_data.head()
#We can summarize data by multiple columns like country, year, age, #generation etc. For the sake of simplicity we will try and the #answer the following questions
#1. How total suicide number in the globe changed with time
#2. Total suicides by gender and age bucket separately
#3. Total suicides by gender across age buckets
#4. Proportion of suicides by generation
#5. Distribution of suicides till date using box plot and histogram
#— — — — — — — — — — — — -Since data is at different levels, we will try and summarize the data by year, --------------
#— — — — — — — — — — — — -sex, age, sex+age and generation--------------
year_summary=suicide_data.groupby('year').agg(tot_suicide=('suicides_no','sum')).sort_values(by='year',ascending=True)
gender_summary=suicide_data.groupby('sex').agg(tot_suicide=('suicides_no','sum')).sort_values(by='sex',ascending=True)
age_summary=suicide_data.groupby('age').agg(tot_suicide=('suicides_no','sum')).sort_values(by='tot_suicide',ascending=True)
#— — — — — — — — — — — — -Line Graph to see the trend over years-----------------
import matplotlib.pyplot as plt
import numpy as np
%matplotlib inline
year_summary.plot(kind='line', figsize=(10,6));
plt.title('Summary of Suicides by Year');
plt.xlabel('Year');
plt.ylabel('#Suicides');
#— — — — — — — — — — — — -Bar graph to compare suicides by gender and age-------------
gender_summary.plot(kind='bar', figsize=(10,6));
plt.title('Summary of Suicides by Gender');
plt.xlabel('Sex');
plt.ylabel('#Suicides');
age_summary.plot(kind='bar', figsize=(10,6));
plt.title('Summary of Suicides by Age');
plt.xlabel('Age');
plt.ylabel('#Suicides');
#— — — — — — — — — — — — -Total suicides by age and gender together----
gender_age_summary=suicide_data.groupby(['sex','age']).agg(tot_suicide=('suicides_no','sum')).unstack()
gender_age_summary.head()
gender_age_summary.plot(kind='bar', figsize=(10,6), stacked=True);
plt.title('Summary of Suicides by Genger & Age');
plt.xlabel('Age');
plt.ylabel('#Suicides');
#— — — — — — — — — — — — -Proportion of Suicide by Generation— — — — — — — — — — — — -
generation_summary=suicide_data.groupby('generation').agg(tot_suicide=('suicides_no','sum')).sort_values(by='tot_suicide',ascending=False).reset_index()
generation_summary.head(10)
#— — — — — — — — — — — — -Plotting pie chart— — — — — — — — — — — — -
fig1, ax1 = plt.subplots();
fig1.set_size_inches(8,6)
ax1.pie(generation_summary['tot_suicide'],labels=generation_summary['generation'],autopct='%1.0f%%',shadow=True);
ax1.axis('equal')  # Equal aspect ratio ensures that pie is drawn as a circle.
plt.title('% Suicides across Generation')
plt.show();
#— — — — — — — — — — — — -Histogram Plot— — — — — — — — — — — — -
year_summary.plot(kind='box', figsize=(10,6)); #----------To plot histogram change kind='box' to kind='hist'
plt.title('Distribution of suicide figures');

Useful tips — use sns.set(color_codes=True) before using the seaborn functionality. This adds a nice background to your visualization.

import seaborn as sns
sns.set(color_codes=True)
#— — — — — — — — — — — — -Suicides by year--------------------------
year_summary=suicide_data.groupby('year').agg(tot_suicide=('suicides_no','sum')).sort_values \
                                    (by='year',ascending=True).reset_index()
year_summary.head()
#— — — — — — — — — — — — -Using Seaborn to plot trend of suicide with time-------------------------------
fig, ax = plt.subplots(figsize=(10,6));
plt.title('#Suicides across Years')
sns.lineplot(year_summary['year'],year_summary['tot_suicide']);
#— — — — — — — — — — — — -Summary by Gender------------------------------------------
gender_summary=suicide_data.groupby('sex').agg(tot_suicide=('suicides_no','sum')).sort_values(by='sex',ascending=True).reset_index()
fig, ax = plt.subplots(figsize=(10,6));
plt.title('#Suicides across Years')
ax=sns.barplot(gender_summary['sex'],gender_summary['tot_suicide']);
ax.set(xlabel='Gender', ylabel='#Suicides');
plt.show()
#— — — — — — — — — — — — -Summary by Gender/age------------------------------------------
gender_age_summary=suicide_data.groupby(['sex','age']).agg(tot_suicide=('suicides_no','sum')).reset_index()
gender_age_summary.head()
fig, ax = plt.subplots(figsize=(10,6));
plt.title('#Suicides across Years')
ax=sns.barplot(x=gender_age_summary['sex'],y=gender_age_summary['tot_suicide'],hue=gender_age_summary['age']);
ax.set(xlabel='Gender', ylabel='#Suicides');
plt.show()
gender_age_summary1=suicide_data.groupby(['sex','age']).agg(tot_suicide=('suicides_no','sum')).unstack()
gender_age_summary1.head()
#— — — — — — — — — — — — -Stack bar--------------------------------
sns.set()
gender_age_summary1.plot(kind='bar', stacked=True)
#— — — — — — — — — — — — -Checking correlation between suicide, population and gdp per capita
sns.pairplot(suicide_data[['suicides_no', 'population', 'gdp_per_capita ($)']], size=4, aspect=1);
#— — — — — — — — — — — — -Boxplot— — — — — — — — — — — — -
sns.boxplot(gender_age_summary['sex'],gender_age_summary['tot_suicide']);

1. Define the problem statement clearly and use a top-down approach to break it into multiple hypotheses. In the above case study, the problem statement involves understanding the suicide rates across different countries and factors influencing them. So before jumping into a visualization list down all possible factors that might affect suicide rates, e.g. age, gender, GDP, population, growth, generation, etc.
2. Identify the key factors that might add value to your story. Think of how the factors can be interconnected to get a bigger picture of what is happening. e.g. instead of looking at suicides by age and gender separately try comparing the suicide rate across both the groups. You can derive quality insights like, Male individuals with age between 18 and 25 have higher suicide rates
3. Once you decide on the factors ensure you use the chart selection framework to decide which graph suits the best
4. Ensure consistency in formatting across all your graphs which includes color, text size, title size, axis spacing, legend position and alignment of chart objects
5. Below every visualization call out a few findings; This will help you stitch together a story for your stakeholder