Pandas is the best Python framework I most probably use before the Machine learning process for data cleaning and data overview, where we do null value handling outlier treatment and for appropriately creating data.

The best thing about pandas is the compatibility with data sets that you can manipulate as excel files, csv, json, you can also handle lists or sqlalchemy dataframes, it is very important this part of the data with pandas if you want to send to call them elsewhere for example a web page.

What I like most about the pandas framework for Python is the ease of use and its great documentation. Currently, being pandas an extension of numpy, it has one of the best possible documentations.