Pandas

Pandas is the most widely-used data manipulation and analysis library in the Python data science ecosystem, originally created by Wes McKinney in 2008 while working at AQR Capital Management and open-sourced in 2009. The name derives from the term Panel Data, and today the library has become an indispensable tool for data analysts, data scientists, quantitative finance researchers, and machine learning engineers worldwide. Built on top of NumPy arrays, Pandas provides two core data structures: the one-dimensional Series (a labeled array) and the two-dimensional DataFrame (a labeled tabular structure similar to a SQL table or Excel spreadsheet). These structures enable an extremely rich set of data processing capabilities including reading and writing data in CSV, Excel, JSON, SQL, Parquet, HDF5, and HTML formats; missing value detection and imputation (isnull, fillna, interpolate); data reshaping and pivoting (pivot_table, melt, stack, unstack); time series operations (resample, shift, rolling, expanding window functions); hierarchical multi-level indexing (MultiIndex); groupby aggregation operations supporting sum, mean, count, agg, transform, and apply; data merging and joining (merge, join, concat with inner, outer, left, right strategies); vectorized string processing via the .str accessor; and categorical data types (CategoricalDtype). Pandas 2.0 introduced an Apache Arrow-based backend, significantly improving memory efficiency and performance, with PyArrow as an optional backend data structure. As of 2026, Pandas has over 44,000 GitHub stars, over 100 million monthly PyPI downloads, and is used by millions of developers globally, serving as the bedrock of the Python data science stack.