Regression Analysis

Key Features

• ✓ Simple and multiple linear regression
• ✓ Automatic coefficient calculation
• ✓ R-squared and adjusted R-squared
• ✓ Standard errors and p-values
• ✓ Model persistence and predictions
• ✓ Integration with SQL queries

-- Predict house prices based on square footage
REGRESS price FROM houses USING sqft

-- Output:
-- Intercept: 50000.00
-- Coefficient (sqft): 150.25
-- R-squared: 0.85
-- Adjusted R-squared: 0.84
-- Standard Error: 5000.00

-- Predict price with multiple features
REGRESS price FROM real_estate
USING sqft, bedrooms, bathrooms, age, location_score

-- Results show coefficient for each predictor
-- with corresponding p-values and significance levels

-- Create and name a model
CREATE REGRESSION MODEL price_model
AS SELECT sqft, bedrooms, age
FROM housing_data
PREDICT price;

-- Use the model for predictions
SELECT
    address,
    sqft,
    bedrooms,
    age,
    PREDICT(price_model) as predicted_price
FROM new_listings;

-- Calculate prediction errors
SELECT
    address,
    actual_price,
    PREDICT(price_model) as predicted_price,
    actual_price - PREDICT(price_model) as residual,
    ABS(actual_price - PREDICT(price_model)) as absolute_error
FROM validation_set;

Example 1: Sales Forecasting

-- Predict sales based on advertising spend
REGRESS sales FROM marketing_data
USING tv_advertising, radio_advertising, social_media_spend

-- Interpretation:
-- Each $1000 in TV advertising → +$5000 in sales
-- Each $1000 in radio → +$3000 in sales
-- Each $1000 in social media → +$2500 in sales

Example 2: Customer Lifetime Value

-- Predict customer value based on behavior
CREATE REGRESSION MODEL clv_model
AS SELECT
    purchase_frequency,
    average_order_value,
    account_age_months,
    engagement_score
FROM customer_history
PREDICT lifetime_value;

-- Score new customers
SELECT
    customer_id,
    PREDICT(clv_model) as predicted_clv,
    CASE
        WHEN PREDICT(clv_model) > 10000 THEN 'High Value'
        WHEN PREDICT(clv_model) > 5000 THEN 'Medium Value'
        ELSE 'Low Value'
    END as segment
FROM new_customers;

Model Statistics