Is EDA mandatory for unsupervised problems

Question

I have seen many problems solving videos where in the users not performing Exploratory Data Analysis for unsupervised problems. Is this true? But I feel EDA is necessary as we are dealing with unsupervised problems. Please help me in understanding this.

Answer 1

When you are dealing with data there are no set of rules which you are supposed to follow step by step. And as far as EDA is concerned, it is always a good practice to explore your data before jumping into model making( but not mandatory).

The videos you have seen might be focused on the unsupervised learning portion, so they skipped the EDA part.

But I would highly recommend applying EDA first. It will help you understand the data well:-

• Missing values
• Imbalanced dataset
• Outliers
• Skewed distribution

and much more. You need to know these issues in advance in order to handle them and to get more accurate results.