Interpolation Techniques

What is Interpolation?

Interpolation guesses missing values based on values around it.

Think of it like this:

Monday: 10°C
Tuesday: ? (missing)
Wednesday: 20°C

Interpolation guesses Tuesday is about 15°C (middle of 10 and 20).

Linear Interpolation

Draws a straight line between known points.

code.py

import pandas as pd

df = pd.DataFrame({
    'Day': [1, 2, 3, 4, 5],
    'Temp': [10, None, None, 20, 25]
})

print("Before:")
print(df)

df['Temp'] = df['Temp'].interpolate()

print("\nAfter interpolation:")
print(df)

Output:

Before:
   Day  Temp
0    1  10.0
1    2   NaN
2    3   NaN
3    4  20.0
4    5  25.0

After interpolation:
   Day  Temp
0    1  10.00
1    2  13.33  <- estimated
2    3  16.67  <- estimated
3    4  20.00
4    5  25.00

Values go smoothly from 10 to 20.

Interpolation vs Forward Fill

code.py

df = pd.DataFrame({
    'Value': [10, None, None, 40]
})

# Forward fill: copies previous value
print("Forward fill:")
print(df['Value'].ffill())

# Interpolate: estimates based on pattern
print("\nInterpolate:")
print(df['Value'].interpolate())

Output:

Forward fill:
0    10.0
1    10.0  <- same as before
2    10.0  <- same as before
3    40.0

Interpolate:
0    10.0
1    20.0  <- goes up smoothly
2    30.0  <- goes up smoothly
3    40.0

When to Use Each

Method	Use When
Forward Fill	Value stays same until change
Interpolate	Value changes gradually

Examples:

Forward fill: Stock price (stays same until trade)
Interpolate: Temperature (changes gradually)

Limit Interpolation

Don't fill too many missing values in a row:

code.py

df = pd.DataFrame({
    'Value': [10, None, None, None, None, 60]
})

# Only fill up to 2 missing values
df['Value'] = df['Value'].interpolate(limit=2)
print(df)

Output:

   Value
0   10.0
1   20.0  <- filled
2   30.0  <- filled
3    NaN  <- not filled (limit reached)
4    NaN  <- not filled
5   60.0

Key Points

interpolate() estimates missing values smoothly
Works best for data that changes gradually
Default is linear (straight line)
Use limit=n to control how many to fill
Better than forward fill for trending data

What's Next?

Learn to convert data types - change text to numbers, numbers to dates, and more.