MLS-C01 Practice Test 10 — 10 Questions

Question 1

A data scientist is using Amazon Athena to query a CSV file stored in S3. The query fails with the error: 'HIVE_CANNOT_OPEN_SPLIT: Number of fields in line 1502 does not match number of fields in the first line.' What is the most likely cause?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

The CSV file has inconsistent number of columns in some rows.. Athena uses a schema-on-read model where the table definition specifies a fixed number of columns. When a CSV row contains more or fewer comma-separated fields than the header/first line, the Hive SerDe cannot map the fields to columns and throws HIVE_CANNOT_OPEN_SPLIT with the line number. The fix is to clean the CSV (ensure every row has the same column count) or use OpenCSVSerDe with tolerance settings.

Answer

The CSV file uses a different delimiter than comma.

Answer

The CSV file is missing a header row.

Answer

The CSV file is too large for Athena to process.

Question 2

A company is building a data lake on Amazon S3. Data arrives from multiple sources in different formats (CSV, JSON, Parquet). The engineering team wants to query this data using Amazon Athena with minimal transformation. Which approach minimizes query cost and improves performance?

Accepted Answer

Use AWS Glue to convert all data to Parquet format, partition by date, and store in a separate S3 bucket. Converting data to Parquet format (a columnar storage format) significantly reduces the amount of data scanned by Athena, which directly lowers query cost (Athena charges per TB scanned). Partitioning by date further limits scanned data by pruning irrelevant partitions. AWS Glue provides a serverless ETL service to perform this conversion efficiently, and storing the output in a separate S3 bucket avoids polluting the raw data lake.

Answer

Use Amazon Redshift Spectrum to query the data directly without transformation

Answer

Use Amazon EMR to convert data to CSV format and repartition

Answer

Store data as-is in S3 and create external tables in Athena for each format

Question 3

A data scientist is using Amazon SageMaker to train a model. The training dataset contains missing values in several features. The data scientist wants to impute missing values using the median of each feature. Which approach is most appropriate?

Accepted Answer

Compute the median of each feature on the training set only, then impute both training and test sets using that median. Computing the median on the training set only avoids data leakage, and applying that median to both training and test sets ensures consistent imputation without using test set information. Option A is incorrect because dropping rows with missing values discards potentially useful data and is not imputation. Option B is incorrect because computing the median on the entire dataset before splitting introduces data leakage, as the test set influences the imputation values. Option C is incorrect because imputing with zero is arbitrary and does not use the median; also, doing it before splitting would use the entire dataset, causing leakage.

Answer

Drop all rows that contain missing values

Answer

Compute the median on the entire dataset, then split into training and test sets

Answer

Impute missing values with zero for all features before splitting

Question 4

A data engineer needs to transform raw clickstream data (JSON files) stored in S3 into a partitioned Parquet dataset for querying with Athena. The transformation includes cleaning, deduplication, and enrichment. The pipeline should run daily. Which solution is MOST cost-effective and requires the least operational overhead?

Accepted Answer

Use an AWS Glue ETL job with a schedule trigger to perform the transformation and write to S3.. AWS Glue ETL jobs are serverless, require no cluster management, and can be easily scheduled for daily runs. Glue also integrates with the Data Catalog for partitioning. Option A (Amazon EMR) is not the most cost-effective or least operational overhead because it requires managing a cluster, even if it can be terminated after completion. Option C (AWS Lambda) is not suitable for large-scale clickstream data due to execution time limits and lack of built-in support for complex transformations like deduplication. Option D (Amazon Athena CTAS) is not appropriate because Athena is primarily a query engine, not a transformation tool; CTAS queries are good for converting data formats but lack the flexibility for cleaning and enrichment.

Answer

Launch an Amazon EMR cluster with Spark, transform the data, and terminate the cluster after completion.

Answer

Use AWS Lambda functions triggered by S3 events to transform each file incrementally.

Answer

Use Amazon Athena to run CTAS queries to convert and partition the data daily.

Question 5

A data analyst is using Amazon SageMaker Studio to perform exploratory data analysis on a dataset stored in S3. The analyst wants to generate summary statistics and visualizations quickly. Which built-in feature of SageMaker Studio should the analyst use?

Accepted Answer

SageMaker Data Wrangler. SageMaker Data Wrangler is the built-in Studio feature for exploratory data analysis, offering 300+ prebuilt transforms, quick summary statistics, histograms, scatter plots, and a visual data flow — all without writing code. It imports directly from S3, Athena, Redshift, and Snowflake, making it the correct tool for fast EDA on an S3 dataset. The other services serve labeling, AutoML, and bias detection, not exploratory visualization.

Answer

SageMaker Ground Truth

Answer

SageMaker Autopilot

Answer

SageMaker Clarify

Question 6

A data scientist is performing EDA on a high-dimensional dataset with 500 features. They want to visualize the data in 2D to check for clusters. They first apply PCA and get a 2D projection that shows no clear structure. They suspect that the data lies on a non-linear manifold. Which of the following techniques should they try next?

Accepted Answer

Use t-distributed Stochastic Neighbor Embedding (t-SNE).. T-SNE is a non-linear dimensionality reduction technique specifically designed for visualization. Option A (ICA) is a linear method for separating independent components, not for capturing non-linear manifolds. Option B (LDA) is a supervised linear method that maximizes class separation, unsuitable for unsupervised non-linear structure. Option C (PCA with more components) remains linear, so adding components does not help with non-linear manifolds.

Answer

Use Independent Component Analysis (ICA).

Answer

Use Linear Discriminant Analysis (LDA).

Answer

Apply PCA again with more components.

Question 7

A financial services company ingests trade events into an Amazon Kinesis Data Stream with four shards. Downstream consumers must process each trade exactly once and write results to Amazon S3, and the pipeline must tolerate consumer restarts without duplicating records or skipping data. The data engineer wants to use the Kinesis Client Library (KCL) to manage shard processing. Which combination of configuration and behavior will meet these requirements?

Accepted Answer

Use KCL with a unique application name, enable checkpointing after each record is fully processed, and rely on the shard-to-worker lease mechanism so restarts resume from the last checkpoint.. The Kinesis Client Library coordinates shard processing through leases stored in a DynamoDB table named after the application, and checkpoints record progress per shard. Checkpointing only after a record is fully processed means a restart resumes at the last committed point, avoiding both duplicates and skipped records. The lease mechanism handles worker failures by redistributing shards.

Answer

Use KCL but disable checkpointing entirely and instead rely on the Kinesis stream's 24-hour retention to replay all records after a restart.

Answer

Use the Kinesis Data Streams API directly with GetRecords and store the shard iterator sequence number in Amazon S3 after each batch, resuming from S3 on restart.

Answer

Use Kinesis Data Firehose with a Lambda processor to write to S3, since Firehose guarantees exactly-once delivery and handles consumer restarts automatically.

Question 8

A data scientist is exploring a dataset and finds that the variance of a feature is 0. What should be done with this feature?

Accepted Answer

Remove the feature from the dataset. A feature with zero variance has the same value in every row, so it carries no information for distinguishing between observations. It cannot contribute to a model's predictive power and can even cause issues in algorithms that divide by standard deviation (e.g., standardization produces division by zero). The correct EDA action is to identify and remove such constant features.

Answer

Create interaction terms with other features

Answer

Apply Min-Max scaling to normalize the feature

Answer

Impute missing values using the mean

Question 9

A machine learning engineer is performing exploratory data analysis on a dataset containing customer transactions. They notice that the target variable is highly imbalanced: 99% of samples belong to class 0 and 1% to class 1. Which technique should they use to address this imbalance before training a classification model?

Accepted Answer

Apply SMOTE to generate synthetic samples for the minority class.. SMOTE (Synthetic Minority Over-sampling Technique) generates synthetic samples for the minority class, which helps balance the dataset and improves model performance on the minority class without losing information from the majority class. Option A is wrong: training on raw data without addressing imbalance will cause the model to be biased toward the majority class and perform poorly on the minority class. Option C is wrong: accuracy is not a suitable evaluation metric for imbalanced datasets because a model that always predicts the majority class will achieve 99% accuracy, masking poor performance on the minority class; instead, metrics like precision, recall, F1-score, or AUC should be used. Option D is wrong: under-sampling the majority class to match the minority class size discards a large amount of data, potentially losing valuable patterns and reducing model performance.

Answer

Train the model on the raw data without any modification.

Answer

Use accuracy as the evaluation metric and train on the raw data.

Answer

Under-sample the majority class to match the minority class size.

Question 10

A data scientist is analyzing a dataset with 10,000 rows and 50 columns. The target variable is binary. Which technique is most appropriate for identifying the most important features for predicting the target?

Accepted Answer

Train a Random Forest classifier and use feature_importances_. A Random Forest classifier provides a direct, model-based measure of feature importance via the feature_importances_ attribute, which quantifies each feature's contribution to reducing impurity across all trees. Because the target is binary, a classifier is appropriate, and Random Forest handles mixed feature types and non-linear relationships well. This gives an interpretable ranking of the 50 features for predicting the target.

Answer

Use t-SNE to reduce dimensionality and inspect clusters

Answer

Run K-means clustering and examine cluster centroids

Answer

Apply PCA and select components with highest variance