beginnertutorial

Binary Encoding: Predict Auto Insurance Premiums

Part 3 of the ML Regression Lab. Models only understand numbers — learn to convert a yes/no feature into 1/0 with binary encoding, then predict insurance premiums.

@shvinn

Machine Learning Engineer

1 min readJan 24, 2026

View source repository

Linear models do arithmetic — they can't multiply a weight by the word "yes". Part 3 introduces feature encoding: the step that turns categorical data into numbers a model can use. We start with the simplest case, a two-value (binary) feature.

Follow along: 03_auto_insurance_premium_prediction in github.com/shvinn/ai-lab-99.

The goal

Predict an annual insurance premium from a driver's age (numeric) and whether they caused an accident in the past two years (yes/no).

Step 1 — Inspect the categorical column

import pandas as pd
 
df = pd.read_csv("auto_insurance_data.csv")
df["caused_an_accidient_in_past_two_years"].value_counts()

Step 2 — Binary-encode yes/no → 1/0

A single column with exactly two categories maps cleanly to 1/0. pandas .map() does it in one line:

df["caused_an_accidient_in_past_two_years"] = (
    df["caused_an_accidient_in_past_two_years"].map({"yes": 1, "no": 0})
)

Now both features are numeric and the column is ready for the model.

Step 3 — Train

from sklearn.model_selection import train_test_split
from sklearn.linear_model import LinearRegression
 
X = df[["age", "caused_an_accidient_in_past_two_years"]]
y = df["annual_premium_usd"]
 
X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=42
)
 
model = LinearRegression()
model.fit(X_train, y_train)

Step 4 — Interpret

model.coef_        # → [-9.95, 499.19]
model.intercept_   # → 1997.92

The encoded feature's coefficient has a clean reading: having caused an accident (1 instead of 0) adds about ** $499** to the predicted premium. Each extra year of age subtracts ~$ 10. This interpretability is exactly why the encoding matters — the number means something.

Step 5 — Evaluate

from sklearn.metrics import mean_squared_error
 
rmse_test = mean_squared_error(y_test, model.predict(X_test)) ** 0.5
print(f"RMSE Test: {rmse_test:.2f}")

Step 6 — Deploy

The app collects the raw yes/no choice and applies the same encoding before prediction — your serving code must mirror your training preprocessing exactly.

app.py

import streamlit as st
 
 
def predict_auto_insurance_premium(age, accident_history):
    return -9.95088805 * age + 499.18950284 * accident_history + 1997.9182518985917
 
 
st.title("🚗 Auto Insurance Premium Predictor")
age = st.number_input("Driver Age", 16, 100, 30, 1)
accident_option = st.selectbox("Accident in the past 2 years?", ["yes", "no"])
accident_history = 1 if accident_option == "yes" else 0   # same map as training
 
if st.button("Predict Premium"):
    premium = predict_auto_insurance_premium(age, accident_history)
    st.success(f"💵 Estimated Annual Premium: ${premium:.2f}")

What's next

Binary encoding works for two categories. But what about three or more, like a subscription tier? Using 1, 2, 3 would falsely imply an ordering. Part 4 solves it with one-hot encoding.

Recommended next

One-Hot Encoding: Predict Streaming Revenue

Part 4 of the ML Regression Lab. Handle a multi-category feature the right way — one-hot encode subscription tiers and predict a streaming user's monthly revenue.

Multiple Linear Regression: Predict Energy Consumption

Part 2 of the ML Regression Lab. Extend regression to several features — predict energy use from temperature and humidity — and learn to read multi-feature coefficients.

1 minRead

Machine Learning

intermediatetutorial

One-Hot Encoding: Predict Streaming Revenue

Part 4 of the ML Regression Lab. Handle a multi-category feature the right way — one-hot encode subscription tiers and predict a streaming user's monthly revenue.

1 minRead