Python and notebooks for data

A notebook is a document made of cells. Some cells hold Python code, which you run one at a time, and the result appears right below. Other cells hold plain notes. Jupyter is the standard format. Google Colab and Kaggle notebooks let you use it free in the browser, with nothing to install.

Notebooks are perfect for exploring. You try something, look at the output, and adjust. That speed is why almost every data scientist uses them.

pandas is the Python library for tables. Its main object is the DataFrame: rows and named columns, like a spreadsheet you control with code. You usually start with import pandas as pd, then df = pd.read_csv('sales.csv'). Now df.head() shows the first rows, df.shape tells you rows and columns, and df.info() lists each column and its type.

If you know spreadsheets, you already know the ideas. pandas just lets you repeat them exactly, on millions of rows.

Filter: df[df['country'] == 'Vietnam'] keeps only matching rows. Select: df[['date', 'amount']] keeps only those columns. Group: df.groupby('city')['amount'].sum() gives a total per city, like a pivot table. Join: pd.merge(orders, customers, on='customer_id') matches rows from two tables, like a lookup.

After a merge, always check the row count. If it jumped unexpectedly, the key was not unique and some rows were duplicated. That one habit catches a lot of silent mistakes.

A notebook is reproducible when someone else can run it top to bottom and get the same result. Because cells can be run in any order, it is easy to end up with a notebook that only works on your screen. Before you share, use “Restart and run all” and fix whatever breaks.

Keep the raw data untouched and do all cleaning in code. Write a short note above each step saying why. Future you is the person most grateful for this.