Your First ML Project: End to End
This is the capstone. Across this course you learned data, cleaning, features, splitting, models, metrics and overfitting as separate ideas. Now we glue every one of them onto a single real project — predicting apartment prices in Bishkek — and walk it from a blank folder to an honest, finished result. When you can do this whole loop yourself, you are a machine learning practitioner, not just a reader.
Learning goals
By the end of this lesson you will be able to:
- Turn a vague wish (“guess apartment prices”) into a precise ML problem: regression with a clear target.
- Inspect a raw dataset and spot missing values, duplicates and absurd outliers.
- Clean the data and engineer useful features (one-hot districts, distance to center, a price-per-m² sanity check).
- Split 80/20 before scaling, so you never leak the test set.
- Build a baseline, then LinearRegression, then RandomForest, and compare them with MAE and RMSE.
- Diagnose over/underfitting from the train-vs-test gap, improve one iteration, and sanity-check real predictions.
- Communicate results honestly, know what deployment and monitoring mean, and reuse a checklist on any future project.