Pandas — Loading & Exploring DataFrames

• Install pandas and import it as pd.
• Load a CSV with pd.read_csv.
• Get a fast feel for the data with head, info, describe, shape, columns.
• Inspect dtypes — and why pandas guesses sometimes wrongly.

pip install pandas

import pandas as pd

df = pd.read_csv("students.csv")
print(df)

      name  age  score
0   Mia   13     88
1  Leo   14     75
2   Victor   13     92
3      Mei   14     80

Three things to spot: column names came from the header, the index started at 0, every dtype was inferred from the values. That's pandas in one breath.

1. shape — "how big?"

print(df.shape)   # → (4, 3)  -- 4 rows, 3 columns

2. columns — "what columns?"

print(df.columns.tolist())   # → ['name', 'age', 'score']

3. head / tail — "what does it look like?"

df.head()       # first 5 rows
df.head(3)      # first 3
df.tail(2)      # last 2

4. info — "dtypes and missing?"

df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 4 entries, 0 to 3
Data columns (total 3 columns):
 #   Column  Non-Null Count  Dtype
---  ------  --------------  -----
 0   name    4 non-null      object
 1   age     4 non-null      int64
 2   score   4 non-null      int64
dtypes: int64(2), object(1)

Three columns, no missing values, two int64s and one object (string). If a column has the wrong dtype, you spot it here.

5. describe — "what do the numbers look like?"

df.describe()

             age      score
count   4.000000   4.000000
mean   13.500000  83.750000
std     0.577350   7.804913
min    13.000000  75.000000
25%    13.000000  78.750000
50%    13.500000  84.000000
75%    14.000000  89.000000
max    14.000000  92.000000

Counts, mean, std, min, max, and three quartiles for every numeric column. One method, half a stats course.

Common kwargs to read_csv

sep=";"            different delimiter
header=None        no header row
names=[...]        supply your own column names
parse_dates=["d"]  parse a column as datetime on load
na_values=["-"]    treat extra strings as missing
dtype={"id": str}  force a column's dtype

Save the L4-08 cleaned orders as clean.csv. Run this in a notebook or script:

import pandas as pd

df = pd.read_csv("clean.csv", parse_dates=["date"])

print("📁 file shape:", df.shape)
print("📋 columns   :", df.columns.tolist())
print("📅 date range:", df["date"].min(), "→", df["date"].max())
print()
print("first 5 rows:")
print(df.head())
print()
print("summary:")
print(df.describe(include="all"))

Sample output

📁 file shape: (5, 6)
📋 columns   : ['order_id', 'customer', 'product', 'quantity', 'price', 'date']
📅 date range: 2026-05-01 → 2026-05-03

first 5 rows:
   order_id  customer product  quantity  price       date
0      1001     Ahmad    Pancake         3   1.50 2026-05-01
1      1002       Mei    Cocoa         2   3.00 2026-05-01
2      1003    Victor    Rice         3   8.00 2026-05-02
3      1005      Maya    Rice         2   8.00 2026-05-03
4      1006       Alex    Cocoa         4   3.00 2026-05-03

summary:
       order_id customer product  quantity     price                 date
count        5        5       5  5.000000  5.000000                    5
unique     NaN        5       3       NaN       NaN                  NaN
top        NaN     Ahmad    Rice       NaN       NaN                  NaN
freq       NaN        1       2       NaN       NaN                  NaN
mean    1003.4      NaN     NaN  2.800000  4.700000  2026-05-02 04:48:00
...

Read the diff

The parse_dates= kwarg turned date into real datetime64. describe(include="all") shows stats for non-numeric columns too — counts and modes. Five lines tell you everything you need before starting analysis.

pd.read_csv loads, then 5 methods (shape, columns, head, info, describe) tell you what you've got. parse_dates and dtype kwargs save you from later astype calls. Run this 5-line ritual on every new dataset.