MLS-C01 Exploratory Data Analysis Practice Test 8 — 15 Questions

Question 1

A data scientist is investigating an application that logs errors to Amazon CloudWatch Logs. The data scientist runs the CloudWatch Logs Insights query shown in the exhibit. The query returns no results, even though the data scientist knows errors have occurred. What is the most likely cause?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

The filter pattern is case-sensitive and the log messages use a different case for 'error'.. CloudWatch Logs Insights filter commands are case-sensitive by default. If the application logs messages like 'Error' or 'ERROR' but the query filters for lowercase 'error', the filter will not match any events, producing zero results. The fix is to use a case-insensitive pattern such as '?ERROR ?error' or normalize the log output.

Answer

The stats count() function is misspelled.

Answer

The query sorts by timestamp descending, which hides results.

Answer

The bin(5m) function is not supported in CloudWatch Logs Insights.

Question 2

A data scientist is analyzing a dataset with a target variable that is highly imbalanced (only 1% positive class). The goal is to build a binary classifier. During exploratory data analysis, which metric is MOST appropriate to evaluate the performance of different sampling strategies before model training?

Accepted Answer

Area Under the Receiver Operating Characteristic Curve (AUC ROC). AUC ROC is the most appropriate metric for evaluating sampling strategies on a highly imbalanced dataset because it measures the model's ability to discriminate between classes across all classification thresholds, independent of class distribution. It is robust to imbalance and provides a threshold-independent assessment, making it ideal for comparing how different sampling techniques (e.g., SMOTE, undersampling) affect separability before committing to a model.

Answer

Root Mean Squared Error (RMSE)

Answer

F1 score

Answer

Accuracy

Question 3

A data scientist is exploring a dataset with a large number of features. The scientist suspects that some features are redundant because they are highly correlated with each other. Which technique should the scientist use during EDA to identify and remove such redundant features?

Accepted Answer

Variance Inflation Factor (VIF). Variance Inflation Factor (VIF) quantifies multicollinearity by measuring how much the variance of a coefficient is inflated due to correlation with other features. Features with high VIF (typically >5 or >10) are considered highly correlated and can be removed. Option A is incorrect because chi-square test is used for testing independence between categorical variables, not for identifying redundant features. Option B is incorrect because PCA reduces dimensionality by creating new uncorrelated features, but it does not directly identify which original features are redundant. Option C is incorrect because while a correlation matrix heatmap can show pairwise correlations, it does not account for multicollinearity among multiple features; VIF is more comprehensive.

Answer

Chi-square test

Answer

Principal Component Analysis (PCA)

Answer

Correlation matrix heatmap

Question 4

A data scientist is performing EDA on a dataset with 500,000 rows and 20 columns. The dataset contains missing values in some columns. Which TWO approaches are appropriate for handling missing data during EDA? (Choose 2)

Accepted Answer

Create an indicator column to flag whether the value was missing, then impute with a placeholder. Option C is correct because creating a missing-indicator (dummy) column preserves the information that a value was missing—missingness itself can be informative—while imputing with a placeholder keeps all 500,000 rows available for EDA; this is a standard, low-risk technique during exploratory analysis. Option E is correct because median imputation is robust to outliers and skew for numerical columns, and mode imputation is the appropriate central-tendency choice for categorical columns, making it a sound approach for mixed-type data with missing values. Option A is not appropriate here because forward fill assumes an ordered time series and would propagate stale values across unrelated rows, which is invalid for general EDA. Option B is not appropriate because listwise deletion on 500,000 rows with missing values across 20 columns can discard large amounts of data and bias results. Option D is not appropriate because mean imputation is sensitive to outliers and skew and cannot be applied to categorical columns.

Answer

Use forward fill to propagate the last observed value

Answer

Remove all rows with any missing value (listwise deletion)

Answer

Impute missing values with the mean of each column

Question 5

A data scientist is evaluating feature engineering options for a dataset containing a categorical variable 'education_level' with values: High School, Bachelor, Master, PhD. The target variable is continuous. Which THREE encoding methods are appropriate for this ordinal categorical variable? (Choose 3)

Accepted Answer

One-hot encoding. Label encoding (D) is appropriate because 'education_level' is ordinal, so mapping High School=0, Bachelor=1, Master=2, PhD=3 preserves the natural ordering and lets models exploit the monotonic relationship with the continuous target. One-hot encoding (A) is also valid: it creates a separate binary column per level, avoids imposing a false numeric magnitude, and works well with linear models and regularized regressors for a low-cardinality variable like this. Target encoding (B) is appropriate as well because it replaces each category with the mean of the continuous target per category, capturing the category-target relationship directly and reducing dimensionality while handling the ordinal levels effectively. Hash encoding (C) is not suitable here since it is designed for high-cardinality features and introduces collisions without preserving the ordinal structure, and binary encoding (E) is unnecessary for only four categories and, like hashing, does not respect the ordinal nature of the variable.

Answer

Hash encoding (using feature hashing)

Answer

Binary encoding (convert to binary representation)

Question 6

Refer to the exhibit. A data scientist plans to read this CSV file into memory for exploratory data analysis using pandas. The instance has 8 GB of RAM. What is the MOST likely issue the scientist will encounter?

Accepted Answer

The file is too large to load into memory on this instance. The most likely issue is that the CSV file is too large to fit into the 8 GB of RAM available on the instance. Pandas loads entire DataFrames into memory, so a file approaching or exceeding available RAM will trigger a MemoryError or cause severe swapping. This is a classic pandas limitation for exploratory analysis on constrained instances.

Answer

The file contains too many rows for pandas to handle

Answer

The file is not in CSV format despite the ContentType

Answer

The file is not accessible because of insufficient permissions

Question 7

Refer to the exhibit. A data scientist examines a sample of data and notices that all columns are numeric. The scientist wants to check for multicollinearity. Which statistic should be computed from this sample?

Accepted Answer

Correlation matrix (Pearson). The correlation matrix shows pairwise Pearson correlations, which can indicate high collinearity. Option B is wrong because chi-square is for categorical variables, not numeric. Option C is wrong because VIF requires more variables than observations (or typically used after regression). Option D is wrong because covariance alone is scale-dependent.

Answer

Chi-square test of independence

Answer

Variance Inflation Factor (VIF)

Answer

Covariance matrix

Question 8

A team is using Amazon SageMaker Data Wrangler to perform exploratory data analysis on a large dataset stored in S3. The dataset contains missing values, outliers, and categorical variables with high cardinality. The team wants to understand data distributions and relationships before modeling. Which combination of Data Wrangler features should they use?

Accepted Answer

Generate a data quality report, view histograms, and create scatter plots for selected features.. SageMaker Data Wrangler provides built-in features for exploratory data analysis, including data quality reports (with summary statistics and missing value analysis), histograms for distribution visualization, and scatter plots to explore relationships between features. These are ideal for understanding distributions and correlations early in the pipeline. Option B is incorrect because dropping rows is a data cleaning transformation, not an EDA step, and box plots alone are insufficient for understanding relationships. Option C is incorrect because imputation and one-hot encoding are data preparation transformations applied after EDA. Option D is incorrect because while Data Wrangler generates a data quality report, it does not directly include correlation heatmaps; scatter plots (as in A) are a more direct way to assess relationships.

Answer

Drop rows with missing values and visualize box plots for numerical features.

Answer

Use imputation to handle missing values and one-hot encoding for categorical features.

Answer

Generate a data quality report and a correlation heatmap.

Question 9

A data scientist uses Amazon QuickSight to visualize a dataset and observes that a numerical feature has a skewness of 2.5 and a kurtosis of 8. Which transformation should they apply to make the distribution more normal?

Accepted Answer

Apply a log transformation.. A skewness of 2.5 indicates strong right-skew, and a kurtosis of 8 indicates heavy tails; a log transformation compresses the right tail and is the standard remedy for right-skewed positive-valued features. Log transforms are simple, interpretable, and effective when the data is strictly positive, which is typical for the kind of feature exhibiting this skew.

Answer

Standardize the feature using Z-score normalization.

Answer

Apply a Box-Cox transformation with lambda=0.5.

Answer

Apply Min-Max scaling to the range [0,1].

Question 10

A data scientist wants to identify outliers in a dataset with 1,000 samples and 5 numerical features. Which technique is most appropriate for univariate outlier detection?

Accepted Answer

Interquartile range (IQR) method. The IQR method, where outliers are defined as points below Q1 - 1.5*IQR or above Q3 + 1.5*IQR, is appropriate for univariate outlier detection as it does not assume a specific distribution and is robust to extreme values. PCA (A) is a dimensionality reduction technique, not for outlier detection. Mahalanobis distance (C) is for multivariate outliers. Z-score with threshold 3 (D) assumes normality and is sensitive to extreme outliers.

Answer

Principal component analysis (PCA)

Answer

Mahalanobis distance

Answer

Z-score with a threshold of 3

Question 11

A data scientist is working on a binary classification problem with a highly imbalanced dataset (1% positive class). They have applied oversampling using SMOTE and trained a logistic regression model. The model achieves 99% accuracy on the test set, but the recall for the positive class is only 5%. What is the most likely cause?

Accepted Answer

SMOTE was applied before splitting the data into training and test sets. Applying SMOTE before the train/test split causes synthetic minority samples generated from the full dataset to leak into the test set, and more importantly, it means the test set is no longer representative of the true 1% class distribution — the model can achieve high accuracy by predicting the majority class while the test set's inflated minority representation masks the poor recall. The 5% recall on the positive class is the classic symptom of data leakage from pre-split resampling combined with the model defaulting to the majority class.

Answer

The model is overfitting due to lack of regularization

Answer

Accuracy is not a suitable metric for imbalanced data

Answer

Logistic regression is inappropriate for imbalanced datasets

Question 12

A company has a large dataset of customer transactions stored in Amazon Redshift. A data scientist wants to perform EDA using Python libraries like pandas and matplotlib. The dataset is too large to fit into memory on a single EC2 instance. What is the most efficient approach?

Accepted Answer

Use Amazon Athena Federated Query to run SQL queries against Redshift and retrieve aggregated results. Amazon Athena Federated Query allows querying Amazon Redshift directly using SQL without moving or loading the entire dataset, and the data scientist can retrieve only aggregated or filtered results into pandas for visualization. This avoids the memory constraint because the heavy computation (aggregation, filtering) happens on Redshift's compute engine, and only the reduced result set is transferred to the notebook. This is the most efficient approach for EDA on large datasets that cannot fit in memory.

Answer

Launch an Amazon SageMaker notebook instance with an attached EBS volume large enough to hold the data

Answer

Use a SQLAlchemy connection to read the entire table into a pandas DataFrame and sample it

Answer

Export the Redshift table to Amazon S3 in Parquet format, then use pandas to read the Parquet files

Question 13

A data scientist is building a regression model to predict house prices. The dataset includes a feature 'zip_code' with 1,000 unique values. What is the best way to handle this categorical feature in the exploratory data analysis phase?

Accepted Answer

Apply target encoding using the mean house price per zip code. Target encoding (mean house price per zip code) is the best approach for a high-cardinality categorical feature like zip_code with 1,000 unique values because it captures the predictive relationship between the category and the target while producing a single numeric column. This avoids the dimensionality explosion of one-hot encoding and the arbitrary ordinality of label encoding. It is especially effective in the EDA phase to quickly assess the feature's signal.

Answer

One-hot encode the zip_code feature

Answer

Replace zip_code with the frequency of each zip code in the dataset

Answer

Use label encoding: assign each zip code a unique integer

Question 14

A data scientist is analyzing a dataset for a binary classification problem. The dataset has 10,000 samples and 200 features. After splitting into training (80%) and test (20%), the data scientist trains a decision tree classifier and achieves 100% accuracy on the training set but only 55% on the test set. Which step should the data scientist take first to address this issue?

Accepted Answer

Prune the decision tree to reduce complexity. The model is overfitting: it perfectly memorizes the training data (100% accuracy) but generalizes poorly to unseen data (55% test accuracy). A decision tree with unlimited depth can create leaves for every training sample, capturing noise. The first step to address overfitting is to reduce model complexity, typically by pruning the tree (e.g., cost-complexity pruning with ccp_alpha in scikit-learn) or setting a max_depth. Pruning directly targets the root cause—excessive complexity—without requiring new data or features.

Answer

Use cross-validation to evaluate model performance

Answer

Collect more training data

Answer

Add more features to the model

Question 15

Which THREE statements about data leakage in machine learning are correct? (Select THREE.)

Accepted Answer

Using the target variable to filter features before splitting leads to data leakage. Option A is correct because using the target variable to filter or select features before the train-test split leaks information about the labels into the feature selection process, so the model indirectly sees test-set target information and performance estimates become optimistically biased. Option B is correct because SMOTE is a resampling technique that synthesizes minority-class examples; it must be applied only to the training set after splitting, since applying it before splitting would create synthetic samples derived from data that later appears in the test set, causing leakage. Option C is correct because computing standardization statistics (mean and standard deviation) on the entire dataset before splitting allows the test set's distribution to influence the training preprocessing, which is a classic form of data leakage; the scaler should be fit only on the training data and then applied to the test data. Option D is not correct because cross-validation is an evaluation strategy, not a leakage-prevention mechanism; leakage can still occur inside each fold if preprocessing, feature selection, or resampling is done before the fold split. Option E is not correct because for time series data a random train-test split violates temporal ordering and leaks future information into the past; a chronological split (or time-series-aware cross-validation) is required to avoid leakage.

Answer

Using cross-validation eliminates all possible data leakage

Answer

For time series data, using a random train-test split is recommended to avoid data leakage