Data Science | VoidX Academy

Build With Me: The Feature Forge

Build With Me 02

Cleaning the Mess

Models hate text. Models crash on missing data. Models get confused by wildly different scales. In this build, we are going to write a professional transformation pipeline. Watch the Feature Forge on the left to see exactly how your code mutates the raw data into model-ready features.

Step 1: Handling the Voids

Look at the Raw Data tab on the left. We have missing ages (NaN). If we feed this to a model, it will crash. Let's fix it by imputing the mean.

import pandas as pd
from sklearn.preprocessing import MinMaxScaler

df = pd.read_csv('raw_customers.csv')

# Impute missing values with the column average
df['Age'] = df['Age'].fillna(df['Age'].mean().round())

Your Turn: Hit Run Pipeline. Watch the red NaN values turn green and populate with numbers!

Step 2: One-Hot Encoding

Now look at the 'Tier' column. It says 'Pro', 'Free', and 'Max'. A math equation can't multiply the word 'Pro'. We need to split this into binary columns (0s and 1s).

# Convert categorical text into binary columns
df = pd.get_dummies(df, columns=['Tier'])

Your Turn: Hit Run. Watch how the single 'Tier' column explodes into multiple columns where text becomes machine-readable numbers.

Step 3: Feature Scaling

Finally, look at Income vs Score. Income is in the hundreds of thousands, while Score is 1 to 10. The model will think Income is 10,000x more important just because the numbers are bigger. We must scale it.

# Scale Income to a 0.0 - 1.0 range
scaler = MinMaxScaler()
df[['Income']] = scaler.fit_transform(df[['Income']])

print("Pipeline complete. Data is ready for modeling.")

Your Turn: Hit Run. The massive income numbers are now compressed into neat decimals. You just built a production ETL pipeline!

Data Science: Feature Forge

ID	Age	Tier	Income	Score
101	28	Pro	$85,000	7.2
102	NaN	Free	$42,000	4.1
103	45	Max	$150,000	9.8
104	NaN	Pro	$92,000	8
105	22	Free	$35,000	5.5

pipeline.py

Python 3

CONSOLE OUTPUT

Awaiting transformation pipeline...

Build With Me: The Feature Forge

Build With Me 02

Cleaning the Mess

Step 1: Handling the Voids

Look at the Raw Data tab on the left. We have missing ages (NaN). If we feed this to a model, it will crash. Let's fix it by imputing the mean.

import pandas as pd
from sklearn.preprocessing import MinMaxScaler

df = pd.read_csv('raw_customers.csv')

# Impute missing values with the column average
df['Age'] = df['Age'].fillna(df['Age'].mean().round())

Your Turn: Hit Run Pipeline. Watch the red NaN values turn green and populate with numbers!

Step 2: One-Hot Encoding

Now look at the 'Tier' column. It says 'Pro', 'Free', and 'Max'. A math equation can't multiply the word 'Pro'. We need to split this into binary columns (0s and 1s).

# Convert categorical text into binary columns
df = pd.get_dummies(df, columns=['Tier'])

Your Turn: Hit Run. Watch how the single 'Tier' column explodes into multiple columns where text becomes machine-readable numbers.

Step 3: Feature Scaling

# Scale Income to a 0.0 - 1.0 range
scaler = MinMaxScaler()
df[['Income']] = scaler.fit_transform(df[['Income']])

print("Pipeline complete. Data is ready for modeling.")

Your Turn: Hit Run. The massive income numbers are now compressed into neat decimals. You just built a production ETL pipeline!

Data Science: Feature Forge

ID	Age	Tier	Income	Score
101	28	Pro	$85,000	7.2
102	NaN	Free	$42,000	4.1
103	45	Max	$150,000	9.8
104	NaN	Pro	$92,000	8
105	22	Free	$35,000	5.5

pipeline.py

Python 3

CONSOLE OUTPUT

Awaiting transformation pipeline...

Build With Me: The Feature Forge

Cleaning the Mess

Step 1: Handling the Voids

Step 2: One-Hot Encoding

Step 3: Feature Scaling

Knowledge Check

Build With Me: The Feature Forge

Cleaning the Mess

Step 1: Handling the Voids

Step 2: One-Hot Encoding

Step 3: Feature Scaling

Knowledge Check