Batch prediction vs real-time serving in MLOps - Hands-On Comparison

Batch Prediction vs Real-Time Serving in MLOps

📖 Scenario: You work as a machine learning engineer in a company that predicts customer churn. You want to compare two ways to get predictions from your model: batch prediction and real-time serving.Batch prediction means you run the model on many customers at once, like a nightly job. Real-time serving means you get predictions instantly when a customer interacts with your app.

🎯 Goal: Build a simple Python program that simulates batch prediction and real-time serving using a dummy model. You will create data, configure a threshold, apply prediction logic, and print the results.

📋 What You'll Learn

Create a list of customer IDs and their feature values

Set a prediction threshold variable

Write a function to simulate model prediction

Use batch prediction to predict for all customers

Use real-time serving to predict for one customer

Print both batch and real-time prediction results

💡 Why This Matters

🌍 Real World

Companies use batch prediction to process large amounts of data overnight, saving resources. Real-time serving is used when instant decisions are needed, like fraud detection or personalized recommendations.

💼 Career

Understanding batch vs real-time prediction is key for MLOps engineers to design efficient and responsive machine learning systems.

1

Create customer data list

Create a list called customers with these exact entries: (101, 0.2), (102, 0.8), (103, 0.5), (104, 0.9). Each tuple has a customer ID and a feature value.

MLOps

# Create the list called customers with the given tuples
# Your code here

Need a hint?

Use a list of tuples with the exact customer IDs and feature values.

2

Set prediction threshold

Create a variable called threshold and set it to 0.6.

MLOps

customers = [(101, 0.2), (102, 0.8), (103, 0.5), (104, 0.9)]
# Set the threshold variable to 0.6
# Your code here

Need a hint?

Just assign 0.6 to the variable named threshold.

3

Write prediction function and batch prediction

Define a function called predict that takes a feature value and returns True if the feature is greater than threshold, otherwise False. Then create a list called batch_results using a list comprehension that applies predict to each customer's feature in customers.

MLOps

customers = [(101, 0.2), (102, 0.8), (103, 0.5), (104, 0.9)]
threshold = 0.6
# Define predict(feature) function
# Create batch_results list with predictions for all customers
# Your code here

Need a hint?

Use a function with a simple if check and a list comprehension to apply it to all customers.

4

Print batch and real-time prediction results

Print the batch_results list. Then print the prediction result for customer ID 105 with feature value 0.7 using the predict function (simulate real-time serving).

MLOps

customers = [(101, 0.2), (102, 0.8), (103, 0.5), (104, 0.9)]
threshold = 0.6

def predict(feature):
    return feature > threshold

batch_results = [predict(feature) for _, feature in customers]
# Print batch_results
# Print prediction for customer 105 with feature 0.7
# Your code here

Need a hint?

Use print statements to show the batch list and the single prediction result.