Objective 2.3 Work with Textual, Numerical, Audio, or Video Data Formats AIP-210 Practice Questions (Page 2)

Questions 6–10

1. 6foundation · easy

   Which textual data format is best suited for storing a simple list of records where each record has the same fields, and the data is primarily used for tabular analysis?

   • ACSV
   • BXML
   • CPlain text
   • DJSON
   Reveal Solution Discussion Select an answer first
2. 7application · medium

   A financial firm is training a fraud-detection model on transaction data. The features include transaction amount (a float), account age in days (an integer), and a one-hot encoded merchant category. The dataset has 50 million rows. The team wants to reduce memory usage during training. Which storage representation is most appropriate?

   • AStore the one-hot columns as float32 to match the other features
   • BStore all features as 64-bit floats to preserve precision
   • CCast the amount to float32, keep account age as int32, and store the one-hot columns as uint8
   • DConvert the amount to an integer by rounding to the nearest cent
   Reveal Solution Discussion Select an answer first
4. 8application · medium

   An ML engineer is building a bird-species classifier from field recordings. The raw audio is stored as 48 kHz WAV files. The team wants to train a CNN on spectrogram images. What is the most appropriate first step in the conversion pipeline?

   • AResample the audio to 8 kHz to reduce computational load, then save as a spectrogram
   • BLoad the WAV directly as a 2D array and treat it as an image without transformation
   • CConvert WAV to MP3 to reduce file size before processing
   • DDecode the WAV into a time-domain amplitude array, then compute a short-time Fourier transform (STFT) to produce a spectrogram
   Reveal Solution Discussion Select an answer first
5. 9application · medium

   A customer-support team has a large corpus of support tickets in plain text. They want to train a model to predict ticket priority. The ML engineer needs to convert the text into a numerical representation. Which approach is most appropriate?

   • AUse TF-IDF vectorization on the raw text
   • BUse the length of each ticket as the only feature
   • CConvert the text to JSON and then to a numerical array
   • DAssign each unique word a random integer ID and use those integers directly as features
   Reveal Solution Discussion Select an answer first
6. 10expert · hard

   A security company is building a model to detect shoplifting from surveillance video. The cameras record at 30 fps, 1080p, in H.264 MP4 format. The model needs to detect quick hand movements, which occur over 2-3 frames. The team has limited GPU memory and must process 24 hours of video per day. They are considering extracting frames at different rates. Which frame-extraction strategy is most appropriate?

   • AExtract frames at 15 fps
   • BExtract frames at 30 fps only during motion events detected by a lightweight motion sensor
   • CExtract every frame at 30 fps to avoid missing any hand movement
   • DExtract frames at 10 fps
   Reveal Solution Discussion Select an answer first