Exam AWS Certified Machine Learning - Specialty topic 1 question 286 discussion

B 1.Data Cleaning: SageMaker Data Wrangler is designed for data preparation tasks, including handling missing values, duplicates, and rare values. It provides a visual interface to clean and transform tabular data efficiently. This addresses the data cleaning requirements mentioned in the question. 2.Model Training: Using the built-in SageMaker XGBoost algorithm is a common and effective choice for classification tasks like customer churn prediction. XGBoost is a powerful and widely used algorithm for binary classification problems.

B. Use SageMaker Data Wrangler to clean the data. Use the built-in SageMaker XGBoost algorithm to train a classification model. Explanation: SageMaker Data Wrangler: SageMaker Data Wrangler is designed for efficient data cleaning and preparation. It provides a visual interface that simplifies the process of cleaning tabular data, handling missing values, and addressing duplicate or rare values. Data Wrangler can generate the necessary preprocessing code automatically, reducing the development effort. SageMaker XGBoost (for Classification): XGBoost is a popular and powerful algorithm for classification tasks, including customer churn prediction. SageMaker provides a built-in XGBoost algorithm, making it easy to train a classification model without the need for extensive coding.

Option B involves using SageMaker Data Wrangler to clean the data and the built-in SageMaker XGBoost algorithm to train a classification model. While this is a valid approach, it requires more manual intervention and development effort compared to using SageMaker Canvas.

A is correct

SageMaker Canvas is an excellent tool for those without ML expertise to build models, but it may not provide the detailed control needed for data cleaning and may not be as robust as Data Wrangler for complex cleaning tasks.

https://aws.amazon.com/tw/about-aws/whats-new/2022/05/amazon-sagemaker-canvas-adds-new-data-capabilities-usability-updates/

Answer A- Sagemaker Canvas + categorical model Reason : SageMaker Canvas: SageMaker Canvas is a no-code machine learning tool that allows users to perform data preparation, feature engineering, and model training with minimal technical expertise. It automatically handles tasks like data cleaning, including the removal of duplicates, filling missing values, and managing rare categories. Categorical Model: A categorical (classification) model is the correct type for churn prediction, as it aims to classify whether a customer will stop using the service (churn) or not. SageMaker Canvas provides user-friendly tools to build and evaluate this type of model.

While Amazon SageMaker Canvas can perform automatic data cleaning and preparation, it has certain limitations when it comes to handling complex data cleaning tasks. SageMaker Canvas is designed for building machine learning models with minimal code and effort, primarily targeting business analysts and non-technical users. It provides a guided user interface and automates many steps in the machine learning pipeline, including data cleaning and preparation. However, SageMaker Canvas has a set of built-in data cleaning and preparation operations, which may not be sufficient for handling all types of data quality issues or complex data transformations. If the data requires more advanced cleaning techniques or custom transformations, SageMaker Data Wrangler (option B) would be a better choice.

A is correct Canvas can do without writing single line of code

This can be done without code using SageMaker Canvas: https://aws.amazon.com/blogs/machine-learning/predicting-customer-churn-with-no-code-machine-learning-using-amazon-sagemaker-canvas/ Hence, A is right.

The best solution, meeting the requirements with the least development effort and correctly addressing the problem nature, is: A. Use SageMaker Canvas to automatically clean the data and to prepare a categorical model. This option leverages the simplicity and automatic features of SageMaker Canvas, ensuring minimal development effort while accurately targeting the need for a classification model in customer churn prediction.

See: https://aws.amazon.com/blogs/machine-learning/predicting-customer-churn-with-no-code-machine-learning-using-amazon-sagemaker-canvas/ Canvas also does no-code data cleaning and preparation. So, least development effort is Canvas.