Your ML Setup: scikit-learn in 10 Minutes

One library, one pattern, everything after

scikit-learn is the standard toolkit for classical ML: datasets, dozens of models, and evaluation metrics behind a single consistent API. Learn its rhythm once — fit to train, predict to guess, score to grade — and every model in this track (and most you'll meet in jobs) works the same way.

Install and run your first model (copy this)

pip install scikit-learn pandas

from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier

X, y = load_iris(return_X_y=True)   # 150 flowers, 4 measurements each
Xtr, Xte, ytr, yte = train_test_split(X, y, test_size=0.2, random_state=7)
model = RandomForestClassifier(n_estimators=50, random_state=7)
model.fit(Xtr, ytr)
print("accuracy:", model.score(Xte, yte))  # expect ~0.95+

That loop — load → split → fit → score — is 90% of this whole track. Run it now, in our playground or your terminal, before reading further. A running model in minute 10 beats a perfect mental model in week 3.

Why toy datasets first

Iris (150 flowers) and California housing are tiny, clean, and documented — mistakes surface as wrong numbers, not as mysterious data bugs. Graduate to messy real data only after the loop feels boring. And random_state=7 everywhere: fixed seeds make every run reproducible, so "it worked yesterday" stays true.

Remember this

• fit trains, predict guesses, score grades. Same verbs, every model.
• Toy data first, seeds fixed, loop running by minute 10.