🚀 Stage 1 of 13

Meet the dataset — 4 questions you always ask

head(), shape, info(), describe(). The 4 questions you ask of ANY dataset before touching ML.

🧠 The intuition Before you train anything, you must answer 4 questions about the data. These 4 questions are non-negotiable — skipping them is the #1 cause of broken ML projects. 1. head() -> what does a row LOOK like? (columns, value shapes) 2. shape -> how many rows x columns? (is there enough data?) 3. info() -> what type is each column? are there nulls? 4. describe() -> what's the range/mean/std of each numeric column? If you can answer all 4 in one sentence each, you understand the dataset. If you can't, you're not ready to model it.

✏️ Tiny task — write this yourself Run these 4 commands on the students.csv below and write down ONE sentence you learn from each. import pandas as pd df = pd.read_csv('students.csv') df.head() df.shape df.info() df.describe()

✅ Check — what you should see You should learn something like: head() -> each row is 1 student, columns are study_hours, attendance, ... shape -> 1000 rows x 8 columns (enough for ML) info() -> all numeric, no nulls (lucky us) describe() -> final_score ranges 0-100, mean 67, std 14

🧮 Code files for this stage

📄 data/students.csv csv

student_id,study_hours,previous_score,attendance,assignments_completed,sleep_hours,extracurricular_hours,parental_support,final_score
1,4.5,72,85,8,7,2,3,68
2,6.0,81,92,10,7,1,4,79
3,2.0,55,60,4,6,4,2,51
4,7.5,88,95,10,8,0,5,91
5,3.0,65,72,6,5,3,3,62
6,5.5,78,88,9,7,2,4,76
7,1.5,45,55,3,5,5,1,42
8,8.0,92,98,10,8,1,5,95
9,4.0,68,80,7,6,2,3,66
10,6.5,85,90,10,7,0,4,83
... (1000 rows total)

📄 explore.py python

import pandas as pd

df = pd.read_csv('data/students.csv')

# Question 1: what does a row look like?
print(df.head())

# Question 2: how big is the dataset?
print(df.shape)        # (1000, 8)

# Question 3: types + nulls?
print(df.info())

# Question 4: spread of numeric columns?
print(df.describe())

How it works

head() shows the first 5 rows so you can see what a 'record' actually looks like — is it a student? a transaction? an image? shape tells you if you have enough data (100 rows = too few, 1M rows = too slow for this project). info() reveals nulls and type mismatches — study_hours as a string instead of a float is a classic bug. describe() gives you the statistical sanity check — if final_score's max is 250, something is wrong with the data.

🚧 Routes — what's wired up at this stage Still no routes. This stage is pure pandas — runs in a Jupyter cell, a Python REPL, or our /notebook. No web server needed yet.

🧮 Is a response model necessary here? NO — use response_class=HTMLResponse

Pandas doesn't need a Pydantic response model — it's a data-analysis library, not a web framework. The response model question only becomes real when FastAPI enters at Stage 8.

⚠️ Common mistakes to avoid

Skipping info() and being surprised by nulls at training time.
Not noticing a column is a string when it should be a number — pandas will keep it as object dtype and sklearn will crash later.
Looking at describe() but not asking 'does this make sense?' — if sleep_hours has a max of 24, fine; if it has a max of 50, the data is broken.

👀 What's next Next: Stage 2 visualises 3 relationships — does study time matter? does attendance matter? does previous score correlate with final?

Next — Stage 2

3 plots that tell you if ML is even possible

→