Machine Learning - Machine Learning Development Life Cycle - Exploratory Data Analysis (EDA) Tutorial

Exploratory Data Analysis (EDA)-

Univariate Analysis – Single Variable Analysis

import seaborn as sns

Taking titanic dataset

1] Categorical Data

1. CountPlot

sns.countplot(df[‘Survived’])

Or

df[‘Survived’].value_counts().plot(kind=’bar’)

2. PieChart

The same information (count plot) in terms of percentage

df[‘Survived’].value_counts().plot(kind=’pie’,autopct=’%.2f’)

2] Numerical Data

import matplotlib.pyplot as plt

1. Histogram

plt.hist(df[‘age’],bins=10)

2. Distplot

Improvement of the histogram, it shows probability density function, the line is called kernel density estimation

sns.distplot(df[‘age’])

3. Boxplot

It provides 5 number summary, i.e IQR

sns.boxplot(df[‘age])

Bivariate & Multivariate Analysis – Two or more Variable Analysis

1] Scatterplot (Numerical-Numerical Data)

sns.scatterplot(tips[‘total_bill’],tips[‘tips’],hue=tips[‘sex’],style=tips[‘smoker’],size=tips[‘size’])

2] Bar Plot (Numerical – Categorical)

sns.barplot(titanic[‘Pclass’],titanic[‘Fare])

the black line is the confidence interval.

3] BoxPlot (Numerical – Categorical)

sns.boxplot(titanic[‘Sex],titanic[‘Age])

4] Distplot (Numerical – Categorical)

5] HeatMap (Categorical – Categorical)

6] ClusterMap (Categorical – Categorical)

7] PairPlot

sns.pairplot(iris,hue=’species’)

8] LinePlot (Numerical – Numerical)

Used mostly in time