MLA-C01 Data Preparation for Machine Learning Practice Question

A data engineer is using AWS Glue to prepare a dataset for machine learning. The dataset has several columns with outliers. The engineer wants to detect and handle outliers in a scalable manner. Which TWO approaches should the engineer consider? (Select TWO.)

⚠ Common exam trap

It's easy for candidates to assume that only a single AWS service can handle outlier detection at scale, but the question requires selecting two approaches, and both Glue DynamicFrames and SageMaker Data Wrangler are valid, scalable, and managed AWS solutions for this task.

Answer choices

Why each option matters

Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.

Correct answer & explanation

✓

Use AWS Glue DynamicFrame with Apache Spark to compute interquartile range (IQR) and filter outliers.

Option D is correct because AWS Glue DynamicFrames run on Apache Spark, which scales horizontally across a cluster, and computing the interquartile range (IQR) with Spark SQL/DataFrame operations lets the engineer detect and filter outliers (e.g., values below Q1 - 1.5*IQR or above Q3 + 1.5*IQR) across large datasets efficiently. Option E is correct because Amazon SageMaker Data Wrangler provides built-in outlier detection transforms (such as Robust Standard Deviation, Standard Deviation, and Quantile-based detection) that can be applied to a dataset at scale and exported into a processing or training pipeline, fitting the ML preparation workflow. Option A is not appropriate because manually inspecting data in Amazon S3 does not scale and is error-prone for large datasets. Option B is not appropriate because training a neural network just to remove outliers is overkill, costly, and unnecessary when statistical methods suffice. Option C is not appropriate because pandas in a SageMaker notebook operates in memory on a single instance, so it does not scale to large datasets the way Spark or Data Wrangler do.

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

✗
Manually remove outliers by inspecting the data in Amazon S3.
Why it's wrong here
Inspecting Amazon S3 manually cannot scale to large datasets and provides no statistical outlier detection. It is tempting because visual review suits tiny samples, but the requirement is scalable handling, which Glue's built-in transforms or distributed Spark jobs provide.
✗
Train a neural network to identify anomalies and remove them.
Why it's wrong here
Training a neural network to flag anomalies adds substantial labelling, tuning and compute overhead for straightforward outlier detection. It suits complex high-dimensional anomaly-detection problems, yet the scenario needs scalable statistical handling, which Glue transforms or Spark provide directly.
✗
Use pandas in a SageMaker notebook to calculate z-scores and filter outliers.
Why it's wrong here
Pandas runs single-node in a SageMaker notebook, so it cannot scale across a large dataset and sits outside the Glue job. It is tempting for its familiar z-score filtering, but the requirement is scalable handling within Glue, which distributed Spark or built-in transforms satisfy.
✓
Use AWS Glue DynamicFrame with Apache Spark to compute interquartile range (IQR) and filter outliers.
Why this is correct
Computing IQR per column on a DynamicFrame lets Spark calculate quartiles and filter rows outside the fences in a distributed, scalable manner. This satisfies the outlier-detection constraint across the full dataset without collecting data to a single node.
✓
Use Amazon SageMaker Data Wrangler to apply an outlier detection transform.
Why this is correct
Data Wrangler provides built-in outlier detection transforms that run as part of a scalable SageMaker data flow, letting the engineer identify and handle outliers without writing custom Spark code. This satisfies the requirement for scalable outlier handling within AWS Glue-adjacent ML preparation.

Quick reference

AWS S3 Storage Class Comparison

Storage Class	Min Duration	Retrieval	Use Case
S3 Standard	None	Immediate	Frequently accessed data
S3 Standard-IA	30 days	Immediate	Infrequent access, rapid retrieval
S3 One Zone-IA	30 days	Immediate	Non-critical infrequent data
S3 Intelligent-Tiering	None	Immediate–hours	Unknown or changing access patterns
S3 Glacier Instant	90 days	Milliseconds	Archive with instant retrieval
S3 Glacier Flexible	90 days	Minutes–hours	Archive, flexible retrieval
S3 Glacier Deep Archive	180 days	Hours	Long-term compliance archive

Go deeper

MLA-C01 Data Preparation for Machine Learning Practice Question

Why each option matters

Option-by-option breakdown

AWS S3 Storage Class Comparison

Related to this question