IBM A1000-120 - Assessment: Data Science Foundations Practice Exam 2025: Latest Questions

A retail team wants to describe what happened to weekly sales over the last year and summarize performance with simple metrics. Which type of analytics are they performing?

A data scientist is evaluating the spread of customer wait times. The distribution is skewed with occasional extreme delays. Which measure is the most robust summary of central tendency?

You need to combine a customer table with an orders table so that all customers appear even if they have placed no orders. Which join should you use (customers is the left table)?

A binary classifier outputs probabilities for the positive class. The business wants to flag the top 5% highest-risk cases for manual review. What is the most direct way to meet this requirement?

A dataset contains both numeric features (income, age) and categorical features (city, plan type). You want to build a linear regression model. What is the recommended approach for the categorical features?

A team computes a 95% confidence interval for the mean delivery time as (28.2, 31.6) minutes. Which interpretation is correct?

A scatter plot of two variables shows a strong curved (nonlinear) relationship. You compute Pearson correlation and get a value near 0. What is the best conclusion?

You build a model to predict churn and achieve 98% accuracy. Later you discover only 2% of customers churned in the dataset. What is the most important next step?

A company trains a model to predict loan default. During training, the data scientist normalizes numeric features using the mean and standard deviation computed on the entire dataset before splitting into train/test. The test results look unusually strong. What is the most likely problem?

You are analyzing customer spending with a histogram and notice two distinct peaks. A colleague suggests using the overall mean to represent 'typical' spending. What is the best response?

A retail team wants a quick way to describe what data science is to non-technical stakeholders. Which description best matches data science at a foundational level?

A dataset contains a column "country" with values like "US", "United States", "U.S.", and "USA". Before analysis, what is the most appropriate first step?

A team trains a classification model to predict rare fraud events. They report 98% accuracy, but fraud is only 1% of transactions. What is the biggest issue with using accuracy alone here?

You are comparing customer spend between two groups: customers who received a coupon and those who did not. The spend distribution is heavily right-skewed with many outliers. Which measure is most robust for comparing central tendency across groups?

A data scientist builds a regression model to predict house prices. During evaluation, they notice the model performs much better on the training set than on a held-out test set. What is the most likely explanation?

A dashboard shows monthly revenue trends over three years. The product owner asks for a visualization that best highlights seasonality patterns (repeating monthly fluctuations). Which visualization is most appropriate?

A dataset has missing values in a numeric feature "income". The feature is important, and missingness is about 5% of rows. Which approach is generally a reasonable baseline imputation strategy before modeling?

A marketing analyst calculates a correlation coefficient of -0.85 between ad spend and conversions and concludes that increasing spend reduces conversions. What is the best critique of this conclusion?

A data scientist builds a model to predict customer churn. They perform feature scaling (standardization) on the entire dataset before splitting into train and test sets. Test performance looks unusually high. What is the most likely problem and fix?

A team wants to estimate the effect of a new recommendation algorithm on average order value (AOV). They roll it out to all users at once and compare AOV before vs. after. The business also launched a holiday promotion during the same period. What is the best next step to isolate the algorithm’s impact?

A data science team is given a dataset with 5,000 rows and a target label. They discover that 4,900 records are labeled "No" and only 100 are labeled "Yes". Which metric is the best initial choice to evaluate a classification model in this situation?

A dataset contains a "customer_id" field and a "purchase_amount" field. During preprocessing, a model performs extremely well in training but poorly on new data. A review shows that "customer_id" was included as a numeric feature. What is the most likely issue?

A new analyst reports: "Our model is 95% accurate." However, the business cares most about catching rare fraudulent transactions and can tolerate some false positives. Which follow-up question best tests whether the model is meeting the business goal?

A data scientist wants to compare the distribution of incomes between two groups (Group A and Group B). The income variable is strongly right-skewed with several extreme outliers. Which visualization is most appropriate to compare distributions while being robust to outliers?

You are preparing a linear regression model. Two predictors are highly correlated (multicollinearity), and the model’s coefficients vary dramatically with small changes in training data. Which approach is a common way to mitigate this issue while keeping both predictors?