A data set can be an array of numbers or a table with rows and columns. Every ML model needs data to learn from.

Think of a data set like a filing cabinet. Each drawer is a table, each folder inside is a row, and each label on the folder is a column. The filing cabinet itself is your data set.

Two forms of data sets:

1. Arrays (simple lists):

speeds = [99, 86, 87, 88, 111, 86, 103, 87, 94, 78, 77, 85, 86]
print(len(speeds))
# Output: 13
print(max(speeds))
# Output: 111
print(min(speeds))
# Output: 77

2. Tabular data (rows and columns):

💡 Key Insight: In ML, the size and quality of your data set matters more than the cleverness of your algorithm. A simple algorithm on great data will beat a fancy algorithm on bad data every time.

A data set is like an unread diary — the stories are in there, but you have to ask the right questions. Can we predict a car''s AutoPass status from its age and speed? Only the data knows.

Asking questions with Python:

speeds = [99, 86, 87, 88, 111, 86, 103, 87, 94, 78, 77, 85, 86]
# Q: What is the average speed?
avg = sum(speeds) / len(speeds)
print(round(avg, 2))
# Output: 89.77
# Q: How many cars are faster than 90?
fast = [s for s in speeds if s > 90]
print(len(fast))
# Output: 4
# Q: What percentage are above average?
above = [s for s in speeds if s > avg]
print(f"{len(above)}/{len(speeds)}")
# Output: 5/13

With Pandas — even more powerful:

import pandas as pd
df = pd.DataFrame({"Age": [5,7,8,2,17], "Speed": [99,86,87,111,86]})
print(df.describe())
# Output:
#             Age      Speed
# count   5.0000    5.00000
# mean    7.8000   93.80000
# std     5.4037   10.56882
# min     2.0000   86.00000
# max    17.0000  111.00000

💡 Key Insight: Before building any ML model, always explore your data first. Use .describe(), check for missing values with .isnull().sum(), and plot it visually. Skipping this step is how you build models that look great on paper and fail in production.

Think of Python as a multilingual translator for data. It can read arrays, CSVs, databases, JSON — you name it. The two most common tools are lists (built-in) and Pandas DataFrames (for tabular data).

Working with lists:

speeds = [99, 86, 87, 88, 111, 86, 103, 87, 94, 78, 77, 85, 86]
# Slice the first 5 values
print(speeds[:5])
# Output: [99, 86, 87, 88, 111]
# Check if a value exists
print(111 in speeds)
# Output: True

Working with Pandas DataFrames:

import pandas as pd
# Create a DataFrame from a dictionary
data = {
    "Car": ["BMW", "Volvo", "VW"],
    "Age": [5, 7, 8],
    "Speed": [99, 86, 87]
}
df = pd.DataFrame(data)
print(df)
# Output:
#      Car  Age  Speed
# 0    BMW    5     99
# 1  Volvo    7     86
# 2     VW    8     87

💡 Key Insight: Pandas DataFrames are the lingua franca of ML in Python. Almost every ML library (scikit-learn, TensorFlow, etc.) expects your data in this format. Learning Pandas is not optional — it is step one.