Data Quality and Its Effect on the ML Model CT-AI Practice Questions (Page 2)

Questions 6–10

1. 6application · medium

   An e-commerce company is training a model to recommend products. The training data contains duplicate user sessions: the same user's browsing history appears multiple times, but with slightly different timestamps. The model performs well on the training set but poorly on new users. What is the MOST likely cause?

   • AThe data was not timely, so the model learned outdated user preferences.
   • BThe duplicates caused the model to overfit to the repeated user patterns, reducing generalization to unseen users.
   • CThe duplicates introduced noise that the model could not filter out, reducing overall accuracy.
   • DThe timestamps were inconsistent, violating the consistency dimension and confusing the model.
   Reveal Solution Discussion Select an answer first
2. 7application · medium

   A streaming service is building a model to predict user engagement. The 'watch_time' feature has 20% missing values. The team is considering two imputation methods: (A) mean imputation, or (B) regression imputation using 'number_of_sessions' as a predictor. Which method is MORE appropriate, and why?

   • AMean imputation — it preserves the variance of the feature.
   • BRegression imputation — it is guaranteed to produce unbiased estimates.
   • CRegression imputation — it uses the relationship with other features to estimate more realistic values.
   • DMean imputation — it is simpler and faster to implement.
   Reveal Solution Discussion Select an answer first
4. 8foundation · easy

   Which metric is commonly used to evaluate the completeness of a dataset?

   • AF1 score
   • BModel accuracy
   • CPercentage of missing values
   • DTraining time
   Reveal Solution Discussion Select an answer first
5. 9application · medium

   A data science team is evaluating whether their dataset is suitable for training a fraud-detection model. They compute the following metrics: 98% of records are complete, 95% of values pass validity checks, and the target variable has a 1% fraud rate. Which additional assessment would be MOST valuable before proceeding?

   • AAssess the class imbalance and its potential impact on model training.
   • BCheck the consistency of the data across different source systems.
   • CCount the number of duplicate transaction IDs to ensure data integrity.
   • DVerify that the timestamps of transactions are within the expected time window.
   Reveal Solution Discussion Select an answer first
6. 10foundation · easy

   What is the purpose of evaluating data quality before training an ML model?

   • ATo reduce the size of the dataset
   • BTo eliminate the need for model validation
   • CTo ensure the data is suitable for the intended ML task
   • DTo increase the model's training speed
   Reveal Solution Discussion Select an answer first