Pandas Interview Questions and Answers

Answer: Python library for data manipulation and analysis

Pandas is a powerful Python library providing data structures and tools for data manipulation and analysis. Built on NumPy, it's essential for data science.

Answer: Using array operations instead of loops

Vectorization uses NumPy array operations instead of Python loops. Dramatically faster. Use built-in methods, broadcasting, avoid apply/iterrows when possible.

Answer: Groups data by column values for aggregation

groupby() splits data into groups based on criteria. Apply aggregation functions (mean, sum, count). Foundation of split-apply-combine pattern in Pandas.

Answer: Series and DataFrame

Series (1D labeled array) and DataFrame (2D labeled data structure) are the two primary Pandas data structures. DataFrame is most commonly used.

Answer: Avoid iteration; use vectorization

Best: avoid iteration with vectorization. If needed: itertuples() > iterrows() > loop. itertuples() fastest. iterrows() slow due to Series creation. Never use loop.

Answer: df.groupby('col').agg(function)

Use df.groupby('col').agg(func). Can use built-in (mean, sum) or custom functions. Pass multiple functions or dict for different columns.

Answer: Joins DataFrames like SQL join

pd.merge(df1, df2) performs database-style joins. Types: inner, outer, left, right. Join on columns or indices. Similar to SQL JOIN operations.

Answer: One-dimensional labeled array

Series is a one-dimensional labeled array that can hold any data type. Like a column in a spreadsheet or dictionary with index labels.

Answer: Memory-efficient storage for categorical data

Category dtype stores categorical data efficiently. Reduces memory for repeated strings. Ordered/unordered. Much faster operations. Convert with astype('category').

Answer: merge joins on columns, join on indices

merge joins on columns (more flexible), join primarily on indices. Both achieve similar results. merge is more explicit and commonly used.

Answer: Two-dimensional labeled data structure with columns

DataFrame is a 2D labeled data structure with columns of potentially different types. Like a spreadsheet or SQL table. Most commonly used Pandas structure.

Answer: Efficient storage for data with many missing values

Sparse data structures efficiently store data with many missing/zero values. Use pd.SparseDtype. Saves memory significantly for sparse datasets.

Answer: All of the above

All methods work, but import pandas as pd is the standard convention. Provides shorter alias while being clear and avoiding namespace pollution.

Answer: Processing large files in chunks

Chunking processes large files in chunks: pd.read_csv(chunksize=10000). Iterate over chunks. Prevents memory errors. Essential for big data in Pandas.

Answer: Concatenates DataFrames along axis

pd.concat([df1, df2]) concatenates along axis. axis=0 (default) stacks vertically, axis=1 horizontally. Use for combining DataFrames without keys.

Answer: pd.DataFrame(dict)

Use pd.DataFrame(dictionary). Keys become column names, values become column data. Most common way to create DataFrame from Python data.

Answer: Reshapes data from long to wide format

pivot() reshapes data: unique values of column become new columns. Creates pivot table. Use pivot_table() for aggregation with duplicate entries.

Answer: Chains operations for readable code

pipe() enables method chaining with functions. df.pipe(func1).pipe(func2). Creates readable data pipelines. Alternative to nested function calls.

Answer: Creates spreadsheet-style pivot table with aggregation

pivot_table() creates pivot table with aggregation. Handles duplicate entries. Specify values, index, columns, aggfunc. More flexible than pivot().

Answer: pd.read_csv('file.csv')

pd.read_csv('filename.csv') reads CSV file into DataFrame. Most common data import method. Supports many parameters for customization (sep, header, etc.).

Answer: JIT compilation for fast custom operations

Numba JIT-compiles Python code to machine code. Use @jit with apply: df.apply(numba_func, engine='numba'). Dramatic speedups for custom operations.

Answer: Parallel computing library for larger-than-memory datasets

Dask provides parallel computing for datasets larger than memory. Similar API to Pandas. Lazy evaluation, task graphs. Use when Pandas hits memory limits.

Answer: df.head()

df.head() displays first 5 rows by default. df.head(n) shows first n rows. Quick way to preview data. Similar: df.tail() for last rows.

Answer: Reshapes data from wide to long format

melt() unpivots DataFrame from wide to long format. Opposite of pivot. Specify id_vars (identifier columns) and value_vars (columns to unpivot).