NumPy & Pandas: Python Notes for Coding & Web Development

📓 Study Notes – NumPy & Pandas

Python – Coding & Web Development

1️⃣ Simple Explanation

Library	Purpose	Key Idea
NumPy	Handles large, multi‑dimensional numerical arrays & matrices efficiently.	Vectorised operations → no explicit loops.
Pandas	Provides high‑level data structures (`Series`, `DataFrame`) for labelled & heterogeneous data.	Built on NumPy; adds indexing, group‑by, I/O, time‑series support.

Why use them?

NumPy gives speed for raw calculations.

Pandas gives convenience for real‑world data analysis (CSV, SQL, etc.).

2️⃣ Key Points

2.1 NumPy

Feature	What It Is	Why It Matters
`np.array`	N‑dimensional container	Core of NumPy; stores homogeneous data.
`dtype`	Data type of array elements	Controls memory usage & precision.
Broadcasting	Implicit expansion of arrays	Lets you add a vector to a matrix without loops.
Universal Functions (`ufuncs`)	Element‑wise math	e.g., `np.sin`, `np.exp`, `np.sqrt`.
Linear Algebra	`np.linalg`	Solve systems, eigenvalues, SVD, etc.
Random module	`np.random`	Generate synthetic data.

2.2 Pandas

Feature	What It Is	Why It Matters
`Series`	1‑D labelled array	Handles time‑series, single column data.
`DataFrame`	2‑D labelled table	Core data structure for tabular data.
Indexing (`loc`, `iloc`, `[]`)	Access by label or integer	Flexible selection.
`groupby`	Aggregate by key(s)	Summarise, pivot, transform.
`merge`, `concat`, `join`	Relational operations	Combine multiple tables.
Missing data handling (`NaN`)	`fillna`, `dropna`	Clean & impute.
I/O (`read_csv`, `to_excel`, etc.)	Import/Export	Connect to databases, files.
Time‑Series tools	`pd.date_range`, resampling	Finance, sensor data.

3️⃣ Worked Example

We'll walk through a mini‑project: sales data analysis.

3.1 Import Libraries

import numpy as np
import pandas as pd

3.2 Create a NumPy array

# 3 rows, 4 columns – sales in units
sales = np.array([[10, 12, 9,  11],
                  [7,  14, 8,  9],
                  [12, 13, 11, 15]])

print(sales)

Output

[[10 12  9 11]
 [ 7 14  8  9]
 [12 13 11 15]]

3.3 Basic NumPy ops

# Total sales per product (column sum)
total_per_product = sales.sum(axis=0)      # axis=0 => sum rows

# Average daily sales
avg_daily = sales.mean(axis=1)             # axis=1 => sum columns

print("Total per product:", total_per_product)
print("Avg daily sales:", avg_daily)

Output

Total per product: [29 39 28 35]
Avg daily sales: [10.5 10.  13. ]

3.4 Convert to Pandas DataFrame

# Column names: P1, P2, P3, P4
df = pd.DataFrame(sales,
                  columns=['P1', 'P2', 'P3', 'P4'],
                  index=['Day1', 'Day2', 'Day3'])

print(df)

      P1  P2  P3  P4
Day1  10  12   9  11
Day2   7  14   8   9
Day3  12  13  11  15

3.5 Add computed columns

df['Total'] = df.sum(axis=1)
df['Avg']   = df.mean(axis=1)
print(df)

      P1  P2  P3  P4  Total   Avg
Day1  10  12   9  11     42  10.5
Day2   7  14   8   9     38  10.0
Day3  12  13  11  15     51  13.0

3.6 Grouping & Aggregation

Suppose we add a Region column.

df['Region'] = ['East', 'West', 'East']
print(df)

      P1  P2  P3  P4  Total   Avg Region
Day1  10  12   9  11     42  10.5   East
Day2   7  14   8   9     38  10.0   West
Day3  12  13  11  15     51  13.0   East

Group by region and compute mean sales per product.

grouped = df.groupby('Region')[['P1', 'P2', 'P3', 'P4']].mean()
print(grouped)

        P1   P2   P3   P4
Region                      
East    11.0 12.5 10.0 13.0
West     7.0 14.0  8.0  9.0

3.7 Merging DataFrames

Create a second DataFrame of Discounts per region.

discounts = pd.DataFrame({
    'Region': ['East', 'West'],
    'Disc%': [5, 10]
}).set_index('Region')

merged = df.join(discounts, on='Region')
print(merged)