Explore richer structures

A MultiIndex stores more than one label per row. Here Campus and Student_ID jointly locate a student. sort_index arranges the levels for predictable selection.

df = students().set_index(["Campus", "Student_ID"]).sort_index()
result = df.loc[("RAK", "S003"), ["Name", "Score"]]
print(result)

Name     Chloe
Score       91
Name: (RAK, S003), dtype: object

Watch for this. A tuple used as a MultiIndex key is different from a list of separate row keys.

df = students().set_index(["Campus", "Student_ID"]).sort_index()
result = df.xs("Dubai", level="Campus")[["Name", "Score"]]
print(result)

What you should be able to explain. Name index levels so selections are understandable.

stack moves column labels into an additional row-index level. unstack moves an index level back into columns. The explicit future_stack=True selects the modern stack implementation on pandas 2.2 and 3.x.

wide = students().head(2).set_index("Student_ID")[["Score", "Attendance"]]
result = wide.stack(future_stack=True)
print(result)

Student_ID            
S001        Score         88
            Attendance    95
S002        Score         72
            Attendance    85
dtype: int64

Watch for this. Missing combinations can appear when reshaping sparse data; inspect the resulting shape.

wide = students().head(3).set_index("Student_ID")[["Score", "Attendance"]]
stacked = wide.stack(future_stack=True)
result = stacked.unstack().reindex(columns=wide.columns)
print(result)
print("Values restored:", result.equals(wide))

What you should be able to explain. Stack and unstack move labels between axes.

Sometimes one cell contains several items, such as a list of skills. explode creates a separate row for each item and repeats the other columns. This changes the unit of analysis from one row per student to one row per student-skill.

df = pd.DataFrame({"Student_ID": ["S001", "S002"], "Skills": [["Python", "SQL"], ["Excel"]]})
result = df.explode("Skills", ignore_index=True)
print(result)

  Student_ID  Skills
0       S001  Python
1       S001     SQL
2       S002   Excel

Watch for this. An empty list becomes a row with a missing item; a string that looks like a list is still a string.

df = pd.DataFrame({"Student_ID": ["S001", "S002"], "Skills": [["Python", "SQL"], ["Excel", "SQL"]]})
long = df.explode("Skills", ignore_index=True)
result = long["Skills"].value_counts()
print(result)

What you should be able to explain. Check what a row represents after changing the layout.

Categorical values can follow a chosen order instead of alphabetical order. get_dummies makes one indicator column per category. These indicator values describe membership, not a numerical ranking.

df = students().head(4).copy()
df["Program"] = pd.Categorical(df["Program"], categories=["IT", "AI"], ordered=True)
ordered = df.sort_values("Program", kind="stable")
result = pd.get_dummies(ordered["Program"], prefix="Program", dtype=int)
print(ordered[["Student_ID", "Program"]])
print(result)

  Student_ID Program
1       S002      IT
3       S004      IT
0       S001      AI
2       S003      AI
   Program_IT  Program_AI
1           1           0
3           1           0
0           0           1
2           0           1

Watch for this. Values outside a specified category list become missing; validate allowed categories first.

df = students().head(4).copy()
df["Program"] = pd.Categorical(df["Program"], categories=["AI", "IT"], ordered=True)
ordered = df.sort_values("Program", kind="stable")
result = pd.get_dummies(ordered["Program"], prefix="Program", dtype=int)
print(ordered[["Student_ID", "Program"]])
print(result)

What you should be able to explain. Category order and indicator encoding are separate choices.