What is a Model?

The core concept: a function learned from data that makes predictions

Knowledge Debt detected

You can study this freely — but your score may plateau if these foundations have gaps. The Mastery badge requires them to be solid.

seaborn vs matplotlib0%

Explanation

A machine learning model is a mathematical function that maps inputs to outputs, learned from data rather than manually programmed.

Traditional programming:

Rules + Data → Output

Machine learning:

Data + Output (labels) → Rules (the model)

Example — predicting house price:

Input (features): size, bedrooms, location, age
Output (label): price
Model: learns the relationship between features and price from historical data
Prediction: given a new house's features, output a price estimate

The workflow:

1. Collect labeled data (examples with known outputs)
2. Choose a model type (linear regression, decision tree, neural network…)
3. Train the model — adjust its internal parameters to fit the data
4. Evaluate — how well does it predict on data it hasn't seen?
5. Predict — apply the trained model to new, unseen inputs

Key insight: The model doesn't "know" anything — it has learned statistical patterns. It will fail on data that looks nothing like its training data.

Examples

The simplest possible model

fit() = train, predict() = apply to new data

from sklearn.linear_model import LinearRegression
import numpy as np

# Training data: hours studied → exam score
X = np.array([[1],[2],[3],[4],[5]])  # features must be 2D
y = np.array([50, 60, 65, 72, 80])   # labels

model = LinearRegression()
model.fit(X, y)        # TRAINING

# Predict for a student who studied 6 hours
pred = model.predict([[6]])
print(f"Predicted score: {pred[0]:.1f}")  # ~86

Next in Machine Learning Basics

Train/Test Split

Continue

Unlocks

Train/Test Split