machine_learning.md

SQLY Machine Learning Queries

📖 Introduction

Machine Learning (ML) queries in SQLY allow users to apply predictive models, classify data, and generate insights directly within queries.

---

🔍 Applying Pre-Trained Models

SQLY allows integrating pre-trained ML models for predictions.

✅ Example 1: Predict Customer Churn

query:
  select: [customer_id, name, churn_probability]
  from: customers
  where:
    churn_probability:
      predict:
        model: customer_churn_model

This retrieves customer churn probabilities using a pre-trained model.

---

📊 Classifying Data

ML models can classify records into predefined categories.

✅ Example 2: Categorize Support Tickets

query:
  select: [ticket_id, description, category]
  from: support_tickets
  where:
    category:
      predict:
        model: ticket_classifier

This assigns categories to support tickets using an ML classifier.

---

📈 Forecasting Trends

SQLY supports time-series forecasting for predictive analytics.

✅ Example 3: Forecast Future Sales

query:
  select: [date, predicted_sales]
  from:
    forecast:
      model: sales_forecast_model
      input:
        past_sales:
          select: [date, total_sales]
          from: sales

This predicts future sales based on historical sales data.

---

🤖 Anomaly Detection

ML models can detect outliers and unusual patterns.

✅ Example 4: Detect Fraudulent Transactions

query:
  select: [transaction_id, amount, is_fraud]
  from: transactions
  where:
    is_fraud:
      predict:
        model: fraud_detection_model

This flags potentially fraudulent transactions.

---

📌 Summary

• predict applies ML models to classify and analyze data.
• Forecasting models enable trend prediction.
• Anomaly detection helps identify irregularities.