Good features often beat fancy models.
Direct answer
Feature engineering means turning raw data into inputs a model can use well: encoding categories, scaling numbers, extracting parts of dates, handling missing values and creating new combinations. Apply the same transformations to training and test data, and fit them using the training data only.
Common techniques
| Technique | Example |
|---|---|
| One-hot encoding | colour into red, blue, green columns |
| Scaling | put values between 0 and 1 |
| Date features | day of week, month |
| Ratios | price per square metre |
| Binning | age into groups |
Example
import pandas as pd
df = pd.DataFrame({"city": ["Lagos", "Accra", "Lagos"], "price": [100, 80, 120]})
df = pd.get_dummies(df, columns=["city"])
Tips
- Think about what the model needs to know.
- Avoid leaking the target into features.
- Keep transformations in a pipeline.
Where Teyro fits
Teyro's Machine Learning for Beginners course turns this into short lessons you can play in a few minutes a day. It has 200 lessons, and you can start free.
The bottom line
Spend time on features before spending time on algorithms.
