Little Bits of Artificial Intelligence

1. In Simple Words:

• Regular regression gives us fixed numbers (e.g., “Our house is worth $350,000”).
• Bayesian regression says: “There’s a 75% chance our house is worth between $330,000 and $370,000, based on what we know now.”
• It combines our prior belief (what we already knew or assumed) with new data to make a better prediction.

It uses Bayes’ Theorem:

New Belief = (What Data Says × What We Believed Before) / How Surprising the Data Is

2. Real-Life Story

1: Predicting Used Car Prices

Context: Imagine Maya, opening a used car shop. Maya already have some knowledge about car prices based on:

• Age of the car
• Mileage
• Brand

Maya believes a 5-year-old Toyota with 60,000 miles usually sells for around ₹6 lakhs.Now, Maya gathers more data from your first 50 sales.

Maya applies Bayesian regression. Now instead of saying:

“Every 5-year-old Toyota = ₹6,00,000.”

Maya says: “Based on past data, there’s a 70% probability that the price falls between ₹5.7L and ₹6.3L.”

Maya also notices that:

• Low mileage makes the price go up
• A certain brand (say, Maruti) consistently sells better than others

So, Maya’s prediction becomes personalized and adaptive.

Why Bayesian? Because it keeps updating itself as more cars are sold. The more data one have, the less uncertain the model becomes.

2: Medical Diagnosis for Blood Pressure

Context: A doctor Raj treats patients for high blood pressure.

He already has a prior belief: People over 50 with a BMI over 30 tend to have high BP.

Now he starts collecting data from his own patients.

He feeds this into a Bayesian regression model to predict blood pressure based on:

• Age
• BMI
• Exercise habits
• Stress levels

Instead of saying: “Your BP will be 145/90.”

He now says: “There’s a 60% probability your BP is between 140–150 based on current data, and 20% chance it’s above 150 if your stress levels stay high.”

As more patient records come in, his model starts adjusting:

• Maybe BMI is not as important as stress
• Maybe people who walk daily have better BP regardless of age

Why Bayesian? Because medicine isn’t absolute. Confidence levels are important, not just yes/no answers.