Data Insights NumPy Masterclass — Part 1

NumPy Fundamentals — Arrays, Creation & Indexing

Data Science ka backbone — NumPy. Arrays banane se lekar slicing tak, reshape se lekar broadcasting tak — sab kuch real-world examples aur interview questions ke saath Data Insights par.

📑 Is Part 1 Mein Aap Kya Sikhenge:

• Topic 1: NumPy Kya Hai — Introduction & Why Use It
• Topic 2: np.array() — Array Creation from Lists
• Topic 3: np.zeros(), np.ones(), np.full() — Special Arrays
• Topic 4: np.arange() & np.linspace() — Range Arrays
• Topic 5: Array Indexing & Slicing — 1D & 2D
• Topic 6: Array Shape — reshape(), flatten(), ravel()

text

🔍 Definition: NumPy (Numerical Python) is a fundamental Python library for numerical computing. It provides a powerful N-dimensional array object (ndarray), mathematical functions, linear algebra operations, and random number generation — all implemented in C for maximum performance.

🎯 Samjho Simple Bhasha Mein: Socho tumhare paas 1 lakh numbers hain aur tumhe sabka average nikalna hai. Python list se yeh karo toh loop lagana padega — slow hoga. NumPy array se ek line mein hoga aur 50x fast hoga! NumPy data science ka foundation hai — Pandas, Scikit-learn, TensorFlow sab NumPy ke upar bane hain.

💡 NumPy vs Python List — Key Differences:

Speed: NumPy C mein likha hai — Python list se 50-100x fast
Memory: NumPy less memory use karta hai — contiguous memory blocks
Operations: NumPy mein vectorized operations — loop ki zaroorat nahi
Data Type: NumPy array mein ek hi type — list mein mixed types

💻 Real-World Code Examples:

Example 1: NumPy install aur import karo.

# Install (terminal mein)
# pip install numpy

# Import karo
import numpy as np

# Version check
print(np.version)
print("NumPy ready!")

Example 2: NumPy vs Python List — Speed comparison.

import numpy as np
import time

# Python List approach
py_list = list(range(1000000))
start = time.time()
result = [x * 2 for x in py_list]
print(f"List time: {time.time() - start:.4f} sec")

# NumPy approach
np_arr = np.arange(1000000)
start = time.time()
result = np_arr * 2
print(f"NumPy time: {time.time() - start:.4f} sec")

📊 Expected Output:

1.26.0
NumPy ready!
List time: 0.1245 sec
NumPy time: 0.0021 sec

NumPy ~60x faster!

⚠️ Common Mistakes:

• Mistake: import numpy likhna bina alias ke → Har jagah numpy.array() likhna padega.
  Fix: Hamesha import numpy as np use karo — industry standard hai.
• Mistake: NumPy array mein mixed types dalna → np.array([1, 'hello', 3.5]) — sab string mein convert ho jayega!
  Fix: Ek hi type ka data rakho NumPy array mein.

💬 Interview Questions:

text

🔍 Definition: np.array() creates an ndarray from a Python list, tuple, or nested list. It automatically detects the data type or you can specify it using the dtype parameter. The result is a NumPy array with powerful mathematical capabilities.

🎯 Samjho Simple Bhasha Mein: np.array() Python ki list ko NumPy array mein convert karta hai. Ek list diya — 1D array bana. List of lists diya — 2D array (matrix) bana. Isi tarah 3D, 4D bhi bana sakte hain. Array ka har element same type ka hota hai — yahi iska power hai.

💡 Important Array Properties:
arr.shape → Dimensions ka size (rows, cols)
arr.ndim → Number of dimensions
arr.dtype → Data type (int64, float64, etc.)
arr.size → Total number of elements
arr.itemsize → Each element ka size in bytes

💻 Real-World Code Examples:

Example 1: 1D, 2D aur 3D arrays banana.

import numpy as np
# 1D Array (Vector)
arr_1d = np.array([10, 20, 30, 40, 50])
print("1D Array:", arr_1d)
print("Shape:", arr_1d.shape)
print("Dimensions:", arr_1d.ndim)

# 2D Array (Matrix) — Student marks
marks = np.array([
[85, 90, 78], # Rahul: Math, Science, English
[92, 88, 95], # Priya
[70, 75, 80] # Amit
])
print("\n2D Array (marks):\n", marks)
print("Shape:", marks.shape) # (3 rows, 3 cols)
print("Total elements:", marks.size)

Example 2: dtype specify karna — data type control.

# Default dtype (auto-detect)
arr_int = np.array([1, 2, 3, 4])
print("Default dtype:", arr_int.dtype) # int64

# Specify float dtype
arr_float = np.array([1, 2, 3, 4], dtype=np.float64)
print("Float dtype:", arr_float.dtype)

# Boolean array
arr_bool = np.array([True, False, True], dtype=bool)
print("Bool array:", arr_bool)

# Type conversion
arr_str = np.array(['1', '2', '3'])
arr_converted = arr_str.astype(np.int32)
print("Converted:", arr_converted, "dtype:", arr_converted.dtype)

Example 3: Real-world — Employee salary data array.

# Employee salary data
salaries = np.array([75000, 85000, 92000, 68000,
55000, 110000, 48000, 95000],
dtype=np.float64)

print("Salaries:", salaries)
print("Shape:", salaries.shape)
print("Dtype:", salaries.dtype)
print("Memory (bytes):", salaries.nbytes)

📊 Expected Output:

1D Array: [10 20 30 40 50]
Shape: (5,)
Dimensions: 1

2D Array (marks):
[[85 90 78]
[92 88 95]
[70 75 80]]
Shape: (3, 3)
Total elements: 9

Default dtype: int64
Float dtype: float64
Bool array: [ True False True]
Converted: [1 2 3] dtype: int32

Salaries: [ 75000. 85000. 92000. 68000. 55000. 110000. 48000. 95000.]
Shape: (8,)
Dtype: float64
Memory (bytes): 64

text

⚠️ Common Mistakes:

• Mistake: Unequal rows dena 2D array mein → np.array([[1,2,3],[4,5]]) — ragged array ban jaati hai, zyaadatar operations fail honge.
  Fix: Saari rows mein same number of elements hone chahiye.
• Mistake: arr.shape ko arr.shape() call karna → shape property hai, method nahi — parentheses mat lagao.
  Fix: arr.shape ✅ — arr.shape() ❌
• Mistake: Integer array mein float result expect karna → np.array([1,2,3]) / 2 → float result aayega Python 3 mein but dtype track karo.
  Fix: Explicit dtype=float64 specify karo financial calculations mein.

💬 Interview Questions:

text

🔍 Definition: NumPy provides functions to create pre-filled arrays without manually listing values. np.zeros() creates arrays filled with 0s, np.ones() with 1s, np.full() with any specified value, np.eye() creates identity matrix, and np.empty() creates uninitialized array (random garbage values).

🎯 Samjho Simple Bhasha Mein: Kabhi kabhi tumhe ek placeholder array chahiye hoti hai — jaise Machine Learning mein weights initialize karna ho, ya ek empty grid banana ho. Manually sab zero likhna tedious hai. np.zeros() ek command mein koi bhi size ka zero-filled array de deta hai. Yeh initialization ke liye bahut use hota hai.

💡 Quick Reference:
np.zeros(shape) → All zeros
np.ones(shape) → All ones
np.full(shape, value) → All same value
np.eye(n) → Identity matrix (diagonal=1)
np.empty(shape) → Uninitialized (fastest)
np.zeros_like(arr) → Same shape as arr, all zeros

💻 Real-World Code Examples:

Example 1: Basic special arrays banana.

import numpy as np
# Zeros array
zeros_1d = np.zeros(5)
zeros_2d = np.zeros((3, 4)) # 3 rows, 4 cols
print("Zeros 1D:", zeros_1d)
print("Zeros 2D:\n", zeros_2d)

# Ones array
ones_2d = np.ones((2, 3), dtype=np.int32)
print("\nOnes 2D:\n", ones_2d)

# Full array — fill with specific value
full_arr = np.full((3, 3), 7)
print("\nFull array (7s):\n", full_arr)

# Identity matrix
identity = np.eye(4)
print("\nIdentity Matrix:\n", identity)

Example 2: Real-world — Sales tracking grid initialize karna.

# 4 products, 12 months ka sales grid initialize karo
# Pehle zero se shuru, baad mein data fill karenge
sales_grid = np.zeros((4, 12), dtype=np.float64)
print("Sales Grid (4 products x 12 months):")
print(sales_grid)
print("Shape:", sales_grid.shape)

# zeros_like — same shape as existing array
existing_data = np.array([[100, 200], [300, 400]])
reset_data = np.zeros_like(existing_data)
print("\nReset to zeros:", reset_data)

# ones_like for multiplication mask
multiplier = np.ones_like(existing_data, dtype=np.float64)
multiplier = multiplier * 1.1 # 10% increase
result = existing_data * multiplier
print("10% increased:", result)

📊 Expected Output:

Zeros 1D: [0. 0. 0. 0. 0.]
Zeros 2D:
[[0. 0. 0. 0.]
[0. 0. 0. 0.]
[0. 0. 0. 0.]]

Ones 2D:
[[1 1 1]
[1 1 1]]

Full array (7s):
[[7 7 7]
[7 7 7]
[7 7 7]]

Identity Matrix:
[[1. 0. 0. 0.]
[0. 1. 0. 0.]
[0. 0. 1. 0.]
[0. 0. 0. 1.]]

Reset to zeros: [[0 0] [0 0]]
10% increased: [[110. 220.] [330. 440.]]

text

⚠️ Common Mistakes:

• Mistake: np.zeros(3, 4) likhna → Error! Shape tuple mein deni hai.
  Fix: np.zeros((3, 4)) — double parentheses zaroori hain 2D ke liye.
• Mistake: np.empty() mein values 0 assume karna → Empty array mein garbage/random values hoti hain — use karne se pehle fill karo.
  Fix: Agar zeros chahiye toh np.zeros() use karo — empty performance ke liye hai jab tum immediately overwrite karoge.

💬 Interview Questions:

text

🔍 Definition: np.arange(start, stop, step) creates evenly spaced values within a given interval using a step size — similar to Python's range() but returns an array. np.linspace(start, stop, num) creates exactly num evenly spaced values between start and stop (both inclusive by default).

🎯 Samjho Simple Bhasha Mein: np.arange() step size bata ke values banata hai — "0 se 10 tak, har 2 step pe." np.linspace() count bata ke values banata hai — "0 se 1 ke beech exactly 5 values chahiye." Plotting ke liye linspace zyada use hota hai — exact count chahiye hota hai. Data engineering mein arange zyada use hota hai.

💡 arange vs linspace — Key Difference:
arange: Step size specify karo → count auto-determine
linspace: Count specify karo → step size auto-calculate

np.arange(0, 1, 0.1) → [0, 0.1, 0.2...] (10 values, stop excluded)
np.linspace(0, 1, 11) → [0, 0.1, 0.2...1.0] (11 values, stop included)

💻 Real-World Code Examples:

Example 1: np.arange() — step-based sequences.

import numpy as np
# Basic arange
arr1 = np.arange(10) # 0 to 9
arr2 = np.arange(1, 11) # 1 to 10
arr3 = np.arange(0, 20, 2) # 0 to 18, step 2 (even numbers)
arr4 = np.arange(10, 0, -1) # 10 to 1 (reverse)

print("0 to 9:", arr1)
print("1 to 10:", arr2)
print("Even numbers:", arr3)
print("Reverse:", arr4)

# Float step
float_arr = np.arange(0.0, 1.0, 0.25)
print("Float step:", float_arr)

Example 2: np.linspace() — count-based sequences.

# Basic linspace
arr5 = np.linspace(0, 10, 5) # 5 values from 0 to 10
arr6 = np.linspace(0, 1, 11) # 11 values from 0 to 1
print("5 values 0-10:", arr5)
print("11 values 0-1:", arr6)

# endpoint=False (exclude end)
arr7 = np.linspace(0, 10, 5, endpoint=False)
print("Exclude end:", arr7)

# Real-world: Temperature analysis points
temps = np.linspace(0, 100, 9) # 9 evenly spaced temperature points
print("Temperature points:", temps)

# Real-world: Time series (0 to 2π for sine wave)
import numpy as np
time_points = np.linspace(0, 2 * np.pi, 100)
sine_wave = np.sin(time_points)
print("Sine wave first 5:", np.round(sine_wave[:5], 3))

📊 Expected Output:

0 to 9: [0 1 2 3 4 5 6 7 8 9]
1 to 10: [ 1 2 3 4 5 6 7 8 9 10]
Even numbers: [ 0 2 4 6 8 10 12 14 16 18]
Reverse: [10 9 8 7 6 5 4 3 2 1]
Float step: [0. 0.25 0.5 0.75]

5
values 0-10: [ 0. 2.5 5. 7.5 10. ]
11
values 0-1: [0. 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1. ]
Exclude
end: [0. 2. 4. 6. 8.]
Temperature points: [ 0. 12.5 25. 37.5 50. 62.5 75. 87.5 100. ]
Sine wave first 5: [0. 0.063 0.125 0.187 0.248]

⚠️ Common Mistakes:

• Mistake: arange mein float step dena → Floating point precision errors aa sakte hain — count unpredictable ho sakti hai.
  Fix: Float ranges ke liye np.linspace() prefer karo — exact count guaranteed.
• Mistake: arange ki stop value included samajhna → arange stop value EXCLUDE karta hai (like Python range).
  Fix: np.arange(1, 11) gives 1 to 10 — 11 included nahi. linspace mein stop included hai by default.

💬 Interview Questions:

text

🔍 Definition: Array indexing accesses individual elements using their position (index starts at 0). Slicing extracts a subset using start:stop:step notation. NumPy supports standard Python slicing plus advanced indexing methods: boolean indexing and fancy indexing (using arrays as indices).

🎯 Samjho Simple Bhasha Mein: Indexing matlab "mujhe exactly yeh element chahiye." Slicing matlab "mujhe yahan se wahan tak ke elements chahiye." 2D array mein dono dimensions specify karne padte hain — row aur column. NumPy mein slicing original array ka VIEW deta hai — copy nahi! Yeh bahut important hai — slice modify karo toh original bhi change hoga.

💡 Indexing Rules:
Positive index: 0 se shuru (left to right)
Negative index: -1 se shuru (right to left)
Slice: arr[start:stop:step] — stop excluded
2D: arr[row, col] or arr[row][col]
View vs Copy: Slice → View. arr.copy() → New copy

💻 Real-World Code Examples:

Example 1: 1D Array indexing aur slicing.

import numpy as np
salaries = np.array([75000, 85000, 92000, 68000,
55000, 110000, 48000, 95000])

# Basic indexing
print("First salary:", salaries[0]) # 75000
print("Last salary:", salaries[-1]) # 95000
print("Third salary:", salaries[2]) # 92000

# Slicing
print("First 3:", salaries[:3])
print("Last 3:", salaries[-3:])
print("Middle (2-5):", salaries[2:5])
print("Every 2nd:", salaries[::2])
print("Reversed:", salaries[::-1])

Example 2: 2D Array indexing — marks matrix.

# 2D: Students x Subjects marks
# Subjects: Math, Science, English, History
marks = np.array([
[85, 90, 78, 88], # Rahul
[92, 88, 95, 70], # Priya
[70, 75, 80, 65], # Amit
[88, 92, 85, 90] # Sneha
])

# Single element
print("Priya's Science marks:", marks[1, 1]) # Row 1, Col 1

# Entire row (Rahul's all marks)
print("Rahul's marks:", marks[0, :])

# Entire column (all Math marks)
print("Math marks:", marks[:, 0])

# Sub-matrix (first 2 students, first 2 subjects)
print("Sub-matrix:\n", marks[:2, :2])

# Last row, last column
print("Sneha's History:", marks[-1, -1])

Example 3: Boolean Indexing — condition-based filtering.

# Boolean Indexing — condition se filter
salaries = np.array([75000, 85000, 92000, 68000,
55000, 110000, 48000, 95000])

# Create boolean mask
high_salary_mask = salaries > 80000
print("Mask:", high_salary_mask)

# Apply mask
high_earners = salaries[high_salary_mask]
print("High earners (>80K):", high_earners)

# One-liner
mid_range = salaries[(salaries >= 60000) & (salaries 90000)]
print("Mid range (60K-90K):", mid_range)

# Fancy Indexing — array of indices
selected = salaries[[0, 2, 5]] # 1st, 3rd, 6th element
print("Selected (index 0,2,5):", selected)

📊 Expected Output:

First salary: 75000
Last salary: 95000
Third salary: 92000
First 3: [75000 85000 92000]
Last 3: [110000 48000 95000]
Middle (2-5): [92000 68000 55000]
Every 2nd: [75000 92000 55000 48000]
Reversed: [ 95000 48000 110000 55000 68000 92000 85000 75000]

Priya's Science marks: 88
Rahul's marks: [85 90 78 88]
Math marks: [85 92 70 88]
Sub-matrix:
[[85 90]
[92 88]]
Sneha's History: 90

Mask: [False True True False False True False True]
High earners (>80K): [ 85000 92000 110000 95000]
Mid range (60K-90K): [75000 85000 68000]
Selected (index 0,2,5): [ 75000 92000 110000]

text

⚠️ Common Mistakes:

• Mistake: Slice ko copy samajhna → b = a[2:5] — b change karo toh a bhi change hoga! Yeh VIEW hai.
  Fix: Independent copy chahiye toh b = a[2:5].copy() use karo.
• Mistake: Boolean indexing mein and/or use karna → salaries > 80000 and salaries < 90000 — Error!
  Fix: & (and), | (or), ~ (not) use karo NumPy mein. Conditions ko parentheses mein wrap karo.
• Mistake: 2D mein arr[1][2] vs arr[1, 2] → Dono kaam karte hain but arr[1, 2] efficient hai (ek lookup).
  Fix: arr[row, col] prefer karo — faster aur cleaner.

💬 Interview Questions:

text

🔍 Definition: Shape manipulation functions change the structure of an array without changing its data. reshape() gives a new shape to the array. flatten() converts to 1D and returns a copy. ravel() converts to 1D and returns a view. transpose() flips rows and columns.

🎯 Samjho Simple Bhasha Mein: Socho 12 numbers hain — tumhe kabhi 3x4 matrix chahiye, kabhi 4x3, kabhi 2x6, kabhi 1D. Reshape se yeh sab possible hai — data same rehta hai, sirf shape change hoti hai. Machine Learning mein yeh bahut zaroori hai — images 3D hoti hain (pixels x pixels x channels) lekin neural network ko 1D input chahiye — flatten se convert karte hain.

💡 reshape() Rules:
Total elements same rehne chahiye: 3x4=12 → 2x6=12 ✅ → 2x5=10 ❌
-1 use karo ek dimension auto-calculate ke liye
arr.reshape(3, -1) → 3 rows, columns auto
arr.reshape(-1) → Flatten to 1D automatically

💻 Real-World Code Examples:

Example 1: reshape() — array structure change karo.

import numpy as np
# 12 elements ka 1D array
arr = np.arange(1, 13)
print("Original 1D:", arr)

# Reshape to 2D
arr_3x4 = arr.reshape(3, 4) # 3 rows, 4 cols
arr_4x3 = arr.reshape(4, 3) # 4 rows, 3 cols
arr_2x6 = arr.reshape(2, 6) # 2 rows, 6 cols

print("\n3x4 Matrix:\n", arr_3x4)
print("\n4x3 Matrix:\n", arr_4x3)

# Auto-calculate one dimension using -1
arr_auto = arr.reshape(3, -1) # 3 rows, NumPy figures out cols (4)
print("\nAuto reshape (3, -1):\n", arr_auto)

# Reshape to 3D
arr_3d = arr.reshape(2, 2, 3) # 2 blocks, 2 rows, 3 cols
print("\n3D Array (2x2x3):\n", arr_3d)

Example 2: flatten(), ravel() aur transpose().

# 2D array
matrix = np.array([[1, 2, 3],
[4, 5, 6],
[7, 8, 9]])

# flatten() — returns COPY, always 1D
flat_copy = matrix.flatten()
flat_copy[0] = 999 # Change copy
print("Flattened copy:", flat_copy)
print("Original unchanged:\n", matrix)

# ravel() — returns VIEW (if possible)
raveled = matrix.ravel()
raveled[0] = 100 # Change view
print("Raveled:", raveled)
print("Original CHANGED:\n", matrix) # matrix[0,0] = 100!

# Transpose — rows become columns
matrix2 = np.array([[1, 2, 3],
[4, 5, 6]]) # Shape: (2,3)
transposed = matrix2.T # Shape: (3,2)
print("\nOriginal (2x3):\n", matrix2)
print("Transposed (3x2):\n", transposed)

📊 Expected Output:

Original 1D: [ 1 2 3 4 5 6 7 8 9 10 11 12]
3x4 Matrix:
[[ 1 2 3 4]
[ 5 6 7 8]
[ 9 10 11 12]]

4x3 Matrix:
[[ 1 2 3]
[ 4 5 6]
[ 7 8 9]
[10 11 12]]

Auto reshape (3, -1):
[[ 1 2 3 4]
[ 5 6 7 8]
[ 9 10 11 12]]

Flattened copy: [999 2 3 4 5 6 7 8 9]
Original unchanged:
[[1 2 3] [4 5 6] [7 8 9]]

Original CHANGED:
[[100 2 3]
[ 4 5 6]
[ 7 8 9]]

Original (2x3):
[[1 2 3] [4 5 6]]
Transposed (3x2):
[[1 4] [2 5] [3 6]]

text

⚠️ Common Mistakes:

• Mistake: reshape mein incompatible shape dena → arr.reshape(3, 5) for 12 elements → ValueError: cannot reshape array of size 12 into shape (3,5).
  Fix: Total elements same hona chahiye: 3×5=15 ≠ 12. Use -1 to auto-calculate.
• Mistake: flatten() aur ravel() ko same samajhna → flatten() hamesha copy. ravel() hamesha view (agar possible).
  Fix: Original protect karna ho → flatten(). Performance chahiye → ravel().
• Mistake: reshape ke baad original changed samajhna → reshape() ek VIEW return karta hai (mostly). Original data same memory mein.
  Fix: Agar independent reshape chahiye → arr.reshape(...).copy()

💬 Interview Questions:

Function	Purpose	Returns	Key Note
np.array()	List to array	ndarray	dtype auto-detect
np.zeros()	All zeros array	ndarray	Shape as tuple
np.ones()	All ones array	ndarray	Good for masks
np.arange()	Step-based range	ndarray	Stop excluded
np.linspace()	Count-based range	ndarray	Stop included
reshape()	Change shape	View/ndarray	Total elements same
flatten()	To 1D	Copy	Safe, independent
ravel()	To 1D	View	Fast, shares memory
arr.T	Transpose	View	Rows ↔ Columns

Next: Data Insights NumPy Masterclass — Part 2

Part 2 mein hum cover karenge: Array Operations — Arithmetic, Broadcasting, Comparison, np.where(), aur Math Functions with real-world data analysis examples Data Insights par.

Happy Coding & Keep Learning! 🚀