EDA - Exploratory data analysis: Becoming more familiar with the data

What is EDA about?

What is EDA about?

Understanding the data by

• Understanding the data dictionary

• How to deal with missing data dataframe.info()

• Finding outliers

  dataframe.describe()

• Which features to keep and discard

Positive vs. Negative correlation

• 1.00 is a perfect correlation (age vs. age) is 100% correlated to itself. 0.00 means there is no correlation.

• Comparing cp to target has a +0.43. So, it has a positive potential correlation to the target

  • As cp goes up. The target value also increases. As cp incraeses, the target (has heart disease) goes up

• exang to target has a -0.44. So, it has a negative potential correlation to the target

  • As exange goes down. The target value will go up. As exang goes up the target (has heart disease) goes down

Explain ways to use EDA

• dataframe.describe() …count, mean, stdDev, min, max

• dataframe['target'].value_counts() will show you if, the data is balanced for dep var

• Finding questions to ask the SME's

• Using a correlation matrix dataframe.corr() to see how each variable is correlated to every other variable

• Visualizing correlation matrix using a heat map to visually see how variables relate to each other

Explain what a crosstab is used for

• To compare different a feature variable against the target variable

• A crosstab compares two variables and puts them in a matrix

• pd.crosstab(dataframe['target'], dataframe.sex)