Manage Data Transformation to Conduct Data Preparation PMI-CPMAI Practice Questions (Page 5)

Questions 21–25

1. 21foundation · easy

   Which activity is a core part of a data cleaning workflow?

   • AGenerating new training samples from existing data
   • BDetecting and correcting missing values, duplicates, and outliers
   • CSelecting the most relevant features for model training
   • DScaling features to a comparable range
   Reveal Solution Discussion Select an answer first
2. 22foundation · easy

   Which component is typically part of an end-to-end preprocessing pipeline?

   • AData cleaning and feature scaling
   • BModel deployment and monitoring
   • CHyperparameter tuning and model selection
   • DData visualization and reporting
   Reveal Solution Discussion Select an answer first
4. 23expert · hard

   A data science team is working on a project that requires compliance with internal audit standards. They need to document the data transformation process, including the rationale for each step, the specific parameters used, and the outputs. They also need to ensure that the documentation is accessible to auditors and can be updated as the pipeline evolves. Which approach should they take?

   • AKeep documentation in a shared drive with a single file that is overwritten each time
   • BUse code comments as the only documentation
   • CDocument only the final transformed dataset and not the steps
   • DMaintain a versioned documentation repository that includes transformation logs, parameter files, and output samples
   Reveal Solution Discussion Select an answer first
5. 24application · medium

   A computer vision team is training a convolutional neural network to classify images of industrial parts. They have only 1,000 labeled images per class, which is insufficient for the model to generalize well. They want to increase the diversity of the training data without collecting new images. Which approach should they use?

   • AApply random rotations, flips, and brightness adjustments to existing images
   • BDuplicate the existing images to increase dataset size
   • CApply principal component analysis to the images
   • DUse SMOTE to generate synthetic images
   Reveal Solution Discussion Select an answer first
6. 25application · medium

   A healthcare startup is developing a model to detect a rare disease from patient records. The dataset has only 200 positive cases and 10,000 negative cases. The team wants to improve model performance without collecting more real data, and they are concerned about patient privacy. They decide to generate synthetic positive samples that mimic the real data distribution. Which method should they use?

   • ARandom oversampling with replacement
   • BData augmentation using rotation and flipping
   • CSMOTE (Synthetic Minority Over-sampling Technique)
   • DUndersampling the majority class
   Reveal Solution Discussion Select an answer first