AWS Certified Machine Learning - Specialty Practice Exam: Test Your Knowledge 2025

A data scientist needs to ingest streaming clickstream data from a web application into Amazon S3 for machine learning model training. The data arrives at a rate of 10,000 records per second and must be transformed before storage. Which solution provides the MOST scalable and managed approach?

A machine learning team is preparing a dataset stored in Amazon S3 for training. The dataset contains 500 GB of CSV files with inconsistent data types and missing values. Which AWS service combination would be MOST efficient for data cleaning and feature engineering at scale?

A company stores training data across multiple AWS accounts in different S3 buckets. A centralized machine learning account needs secure access to this data for model training using Amazon SageMaker. What is the MOST secure and scalable approach?

A data engineer needs to create a data catalog for petabytes of structured and semi-structured data stored in Amazon S3 to enable quick data discovery for ML projects. Which AWS service should be used?

An ML engineer is analyzing a dataset with 50 features and suspects multicollinearity among predictor variables. Which technique should be used to identify and quantify the correlation between features?

A data scientist notices that their dataset has a target variable where 95% of samples belong to one class and 5% to another. Which Amazon SageMaker feature would help visualize and understand this class imbalance during exploratory analysis?

During exploratory data analysis, a data scientist discovers that a numerical feature has values ranging from 0.001 to 10,000 with a highly skewed distribution. Which transformation would be MOST appropriate before training a linear regression model?

A machine learning team is exploring a text dataset for sentiment analysis. They need to understand the most common words and phrases in customer reviews. Which approach using AWS services would be MOST effective for this initial text exploration?

An analyst is performing feature selection on a dataset with 200 features for a classification problem. They want to identify features that have the strongest relationship with the target variable. Which statistical method would be MOST appropriate for this initial screening?

A data scientist notices significant outliers in multiple numerical features of their dataset. Before deciding on an outlier treatment strategy, which visualization in Amazon SageMaker Data Wrangler would provide the BEST understanding of the outlier distribution and impact?

A company needs to build a recommendation system for their e-commerce platform with millions of users and products. Which Amazon SageMaker built-in algorithm would be MOST appropriate?

A data scientist is training a deep learning model using Amazon SageMaker and notices that the training loss decreases steadily but validation loss increases after epoch 15. What is the BEST approach to address this issue?

A machine learning engineer needs to train a computer vision model to classify images into 1,000 categories. The training dataset contains only 500 labeled images per category. Which approach would MOST likely improve model performance?

A data scientist is using Amazon SageMaker's XGBoost algorithm for a binary classification problem. The model achieves 95% accuracy, but the business is primarily concerned with minimizing false negatives. Which metric should be optimized and what hyperparameter adjustment would help?

A company is building a real-time fraud detection system that must make predictions within 100 milliseconds. Which Amazon SageMaker deployment option would BEST meet this latency requirement?

A data scientist needs to perform hyperparameter tuning for a SageMaker training job with 15 hyperparameters. The training job takes 2 hours to complete, and there's a limited budget. Which hyperparameter tuning strategy would be MOST cost-effective?

A machine learning team is building a multi-class classification model with 50 classes. They notice that some classes have only 20 training examples while others have 5,000. Which technique would BEST address this class imbalance during training?

A company deployed a machine learning model using Amazon SageMaker, but predictions are taking longer than expected. Investigation reveals that input data preprocessing is the bottleneck. What is the BEST solution to reduce inference latency?

A data scientist needs to monitor a deployed machine learning model for prediction drift and data quality issues in production. Which AWS service provides built-in capabilities for detecting these issues?

A machine learning team needs to implement a CI/CD pipeline for model training and deployment. They want to automatically retrain models when new data arrives in S3 and deploy only if the model meets performance criteria. Which AWS services combination would BEST accomplish this?