Data Science Practice Questions & Answers

Sample Data Science Questions

Free sample questions with the correct answer highlighted.

1. Q1. You load a 5 GB CSV file on a machine with 8 GB RAM using Pandas. Memory usage spikes and the kernel crashes. What is the most memory-efficient way to process this file in Pandas?

   • A. Convert the CSV file to an array using np.loadtxt() first
   • B. Load data with pd.read_excel() instead of pd.read_csv()
   • C. Use df.dropna() before loading the file into memory
   • D. Use pd.read_csv() with chunksize parameter to iterate over data blocks ✓

2. Q2. In a dataset containing customer records, the salary column has missing values primarily for entry-level roles where experience_years is 0. What mechanism of missingness is present, and how should you impute it?

   • A. Structural missingness; replace all missing values with global mean
   • B. Missing Not at Random (MNAR); fill all missing values with 0 using .fillna(0)
   • C. Missing at Random (MAR); group by experience_years and impute using median ✓
   • D. Missing Completely at Random (MCAR); drop all missing rows using .dropna()

3. Q3. You have a high-cardinality string column 'user_status' containing values 'Active', 'Pending', and 'Inactive' across 10 million rows in Pandas. How can you significantly cut RAM usage?

   • A. Convert column to string type via df['user_status'].astype('string')
   • B. Convert column to integer using pd.to_numeric()
   • C. Apply string strip via df['user_status'].apply(lambda x: x.strip())
   • D. Convert column to categorical type via df['user_status'].astype('category') ✓

4. Q4. You need to perform a join between two DataFrames: sales and returns. You want to keep every sale record regardless of whether it was returned, but duplicate sale rows if multiple returns exist. Which code achieves this?

   • A. pd.concat([sales, returns], axis=1, join='inner')
   • B. sales.join(returns, lsuffix='_sales', how='right')
   • C. pd.merge(sales, returns, on='transaction_id', how='left') ✓
   • D. pd.merge(sales, returns, on='transaction_id', how='outer')

5. Q5. A customer transaction column 'transaction_date' is stored as object strings '2023-05-12 14:30:00'. What is the correct way to convert it and extract the day of the week name?

   • A. df['transaction_date'].astype('datetime').weekday_name()
   • B. pd.DatetimeIndex(df['transaction_date']).get_day()
   • C. df['transaction_date'].apply(lambda x: x.day_name())
   • D. pd.to_datetime(df['transaction_date']).dt.day_name() ✓

6. Q6. You execute df.groupby('department').agg({'salary': ['mean', 'std']}). What index structure does the resulting DataFrame column index have?

   • A. Flat Index
   • B. DatetimeIndex
   • C. RangeIndex
   • D. MultiIndex (Hierarchical Column Index) ✓

7. Q7. While cleaning string data in Pandas, df.drop_duplicates(subset=['email']) fails to remove duplicates that look identical. What is the most common hidden cause?

   • A. The inplace=True parameter was omitted during function call
   • B. drop_duplicates only works on numerical data types
   • C. Inconsistent leading or trailing whitespace in text entries ✓
   • D. Pandas cannot drop duplicate string values across multiple CPU cores

8. Q8. You wish to transform a wide dataset containing columns ['user_id', 'Q1_sales', 'Q2_sales', 'Q3_sales'] into a long dataset with columns ['user_id', 'Quarter', 'Sales']. Which Pandas method is designed for this?

   • A. pd.pivot()
   • B. pd.concat()
   • C. pd.melt() ✓
   • D. pd.crosstab()