Data Quality and Its Effect on the ML Model CT-AI Practice Questions (Page 1)

Questions 1–5

1. 1foundation · easy

   Which of the following is a technique for handling outliers in a dataset?

   • ACapping or winsorizing extreme values
   • BIncreasing the learning rate
   • CUsing a different loss function
   • DAdding more features
   Reveal Solution Discussion Select an answer first
2. 2application · medium

   A bank is training a credit-risk model on historical loan data. The dataset contains 10,000 records, but 40% of the 'annual_income' field is missing. The data engineer decides to impute the missing values with the mean of the available incomes. After training, the model shows unexpectedly poor performance on high-income applicants. What is the MOST likely cause of this degradation?

   • AThe dataset became too small after removing rows with missing values, reducing statistical power.
   • BMean imputation reduced the variance of the income feature, making the model less able to distinguish high-income applicants.
   • CThe timeliness of the income data was compromised because the imputed values came from a different time period.
   • DThe model overfitted to the missing-value indicator that was created during imputation.
   Reveal Solution Discussion Select an answer first
4. 3application · medium

   A telecom company is building a model to predict customer churn. The dataset has a 'contract_type' field with values 'Month-to-month', 'One year', 'Two year', and 'Monthly'. The team notices that 'Month-to-month' and 'Monthly' refer to the same thing. What data quality issue is present, and what is the BEST action?

   • AAccuracy issue — the data is factually wrong and should be corrected from an external source.
   • BCompleteness issue — impute the missing 'Monthly' values with 'Month-to-month'.
   • CConsistency issue — standardize the values to a single label before training.
   • DValidity issue — reject all 'Monthly' records because they are invalid.
   Reveal Solution Discussion Select an answer first
5. 4application · medium

   A manufacturing company is building a predictive-maintenance model. The sensor data contains occasional spikes (e.g., temperature jumps from 80°C to 200°C and back in one reading) that are known to be sensor glitches, not real events. The team wants to clean the data. Which technique is MOST appropriate?

   • ACap all temperatures above 100°C at 100°C to reduce the spike impact.
   • BRemove the spike values and interpolate between the neighboring valid readings.
   • CReplace all spike values with the global mean temperature.
   • DKeep the spikes because they are real outliers that the model should learn to handle.
   Reveal Solution Discussion Select an answer first
6. 5application · medium

   A healthcare startup is building a model to predict patient readmission. The dataset contains a 'heart_rate' column where a small number of values are recorded as 0, which is physiologically impossible. The team wants to handle these values before training. Which approach BEST preserves the integrity of the data while minimizing information loss?

   • ARemove the rows containing 0 values from the dataset entirely.
   • BTreat the 0 values as missing and apply multiple imputation using other patient features.
   • CFlag the 0 values as outliers and cap them at the 1st percentile of the observed distribution.
   • DReplace the 0 values with the median heart rate of the entire dataset.
   Reveal Solution Discussion Select an answer first