Predict Salaries with Linear Regression in Python

Predict a salary from years of experience with your first real machine learning model — linear regression in Python with scikit-learn, trained and honestly tested. The model learns a line: start at $34,031.77, add $5,880.45 for every extra year. On five people it never saw, it predicts 6 years at $69,314, misses by $3,788.91 on average (MAE), and scores R-squared 0.977. You will learn the best-fit line, slope and intercept in plain words, features (X) versus label (y), the honest train/test split, model.fit and model.predict, and the shape mistake that breaks every beginner's first prediction. Why learn this? Machine learning theory appears in 72.5% of 10,704 machine learning job postings (Huntr 2026 analysis), Python in 81.4%, and scikit-learn in 15.8% — and linear regression is the first model interviewers ask about. FULL CODE — salary_prediction.py (copy, run, put it on GitHub): """Predict Salaries with Linear Regression - Portfolio project #15.""" import pandas as pd from sklearn.model_selection import train_test_split from sklearn.linear_model import LinearRegression from sklearn.metrics import mean_absolute_error, r2_score 20 people: years of experience and their salary in dollars. people = pd.DataFrame({ "years_experience": [0, 1, 1, 2, 2, 3, 3, 4, 5, 5, 6, 7, 7, 8, 9, 10, 11, 12, 13, 15], "salary": [34000, 39000, 36000, 48000, 44000, 53000, 50000, 61000, 58000, 69000, 66000, 79000, 73000, 86000, 82000, 98000, 94000, 112000, 106000, 128000], }) FEATURES (X) = the question. LABEL (y) = the answer. X = people[["years_experience"]] y = people["salary"] print(f"Examples: {len(people)} people") print(f"Feature (X): years_experience") print(f"Label (y): salary") Honest split first (Episode 12): train on 15, test on 5 unseen. X_train, X_test, y_train, y_test = train_test_split( X, y, test_size=0.25, random_state=42) print(f"Train: {len(X_train)} people, Test: {len(X_test)} people") Train: fit finds the best-fit line through the training points. model = LinearRegression() model.fit(X_train, y_train) print(f"Slope: ${model.coef_[0]:,.2f} per extra year") print(f"Intercept: ${model.intercept_:,.2f} at zero years") Predict on the 5 people the model never saw. predictions = model.predict(X_test) print("\nTest predictions vs reality:") for years, real, guess in zip(X_test["years_experience"], y_test, predictions): print(f" {years} years: predicted ${guess:,.0f}, real ${real:,}") Honest scores: average miss in dollars, and R-squared. mae = mean_absolute_error(y_test, predictions) r2 = r2_score(y_test, predictions) print(f"\nMean absolute error: ${mae:,.2f}") print(f"R-squared: {r2:.3f}") Predict a brand-new person: 6 years of experience. Note: predict needs a table shape, not a bare number. new_person = pd.DataFrame({"years_experience": [6]}) print(f"\nPrediction for 6 years: ${model.predict(new_person)[0]:,.0f}") print("Done - your first real model, trained and honestly tested.") CHAPTERS: 0:00 Hook: predict a salary in one line 1:08 Why employers care: ML theory in 72.5% of postings 2:26 Regression: the best-fit line, slope and intercept 3:52 Features, label and the honest split: 20 salaries 5:25 Train, fit and predict with scikit-learn 7:15 Honest scores: MAE and R-squared 8:53 The shape mistake, and the extrapolation warning 10:35 Pause & build: portfolio project #15 12:21 Recap + tomorrow: classification Series: Python from Scratch for AI Engineers (30 days) Watch the full series:    • AI/ML with Python — Beginner Series   Subscribe for daily Python-for-ML tutorials: new episode every day. Question for you: what number would YOU predict — a house price, a salary, an exam score? Tell me your feature and your label in the comments. I read every comment. Job stat cited: Huntr analysis of 10,704 ML/AI job postings (2026) #python #linearregression #machinelearning #scikitlearn #pythonforbeginners