Q12Deployment and Orchestration of ML Workflows
Case study - An ML engineer is developing a fraud detection model on AWS. The training dataset includes transaction logs, customer profiles, and tables from an on-premises MySQL database. The transaction logs and customer profiles are stored in Amazon S3. The dataset has a class imbalance that affects the learning of the model's algorithm. Additionally, many of the features have interdependencies. The algorithm is not capturing all the desired underlying patterns in the data. The training dataset includes categorical data and numerical data. The ML engineer must prepare the training dataset to maximize the accuracy of the model. Which action will meet this requirement with the LEAST operational overhead?
← → navigate · a answer
Community votes
Discussion · 3
C 3
Selected Answer: C
https://docs.aws.amazon.com/sagemaker/latest/dg/data-wrangler-transform.html
C 2
Selected Answer: C
You need to convert category data to numeric since ML models work with numbers, so it is either A or C. Data Wrangler offers a built-in transformation to encode categorical data - https://docs.aws.amazon.com/sagemaker/latest/dg/data-wrangler-transform.html#data-wrangler-transform-cat-encode while Glue doesn't provide a managed transformation for encoding data - https://docs.aws.amazon.com/glue/latest/dg/edit-jobs-transforms.html
C 1
Selected Answer: C
Data Wrangler can be used to encode categorical data, i.e. the process of creating a numerical representation for categories. Categorical encoding converts categorical data that is in string format into arrays of integers. Data Wrangler supports ordinal and one-hot encoding, as well as similarity encoding (more advanced).
https://docs.aws.amazon.com/sagemaker/latest/dg/data-wrangler-transform.html#data-wrangler-transform-cat-encode
AWS Glue also has Data science recipe steps for One Hot Encoding and Categorical Mapping.
https://docs.aws.amazon.com/databrew/latest/dg/recipe-actions.data-science.html
However, Data Wrangler is more user-friendly with visual and natural language interfaces, so there is less operational overhead