MLS-C01 Exploratory Data Analysis Practice Test 13 — 15 Questions

Question 1

During EDA, a data scientist plots a scatter matrix of several numeric features and notices that two of them have a Pearson correlation of 0.97. The team plans to train a linear regression model and wants to avoid unstable, hard-to-interpret coefficients. Which action is MOST appropriate based on this finding?

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

Accepted Answer

Remove one of the two highly correlated features or combine them into a single feature before training.. A correlation near 0.97 indicates the two predictors carry almost identical linear information, which destabilizes linear regression coefficients. Removing or merging one of the redundant features directly restores stability and interpretability. Longer training, rescaling, or adding interactions do not change the collinear structure and therefore fail to solve the problem the scatter matrix revealed.

Answer

Increase the number of training epochs so the regression converges despite the correlation.

Answer

Add polynomial interaction terms between the two features to separate their effects.

Answer

Standardize both features to unit variance and retrain the model without other changes.

Question 2

A data scientist is performing EDA on a dataset stored in Amazon S3 in Parquet format. The dataset contains 200 columns, and the scientist needs to compute pairwise correlations between all numerical features to identify multicollinearity before training a linear model. The dataset is 50 GB and cannot be loaded entirely into memory on the SageMaker notebook instance (16 GB RAM). Which approach is MOST efficient and scalable for this task?

Accepted Answer

Use Amazon SageMaker Data Wrangler to import the dataset, then use its built-in correlation matrix analysis to compute pairwise correlations.. SageMaker Data Wrangler is designed for exploratory data analysis on large datasets and includes a built-in correlation matrix that handles data larger than memory by using sampling and distributed processing. It reads Parquet directly from S3 and provides an interactive visual interface. The other approaches either exceed memory limits, require custom coding, or lack statistical functions.

Answer

Use Amazon Athena to run SQL queries that compute correlations between each pair of columns, then aggregate the results.

Answer

Use AWS Glue to crawl the S3 data, then run an AWS Glue ETL job that computes correlations using PySpark and writes the results to S3.

Answer

Load the dataset into a pandas DataFrame on the notebook instance and use the .corr() method to compute pairwise correlations.

Question 3

A data scientist is performing exploratory data analysis on a large dataset stored in Amazon S3. The dataset contains missing values in several numerical columns. The scientist wants to understand the pattern of missingness and decide on an appropriate imputation strategy. The scientist plans to use Amazon SageMaker Data Wrangler. Which two actions should the scientist take to analyze missing values effectively? (Choose two.)

Accepted Answer

Use the Data Wrangler 'Missing Value' analysis to generate a report showing the percentage of missing values per column and their correlation with other features.. The scientist should use the Missing Value analysis and the Data Quality and Insights Report in SageMaker Data Wrangler. These tools provide detailed statistics and visualizations of missing values, including percentages and correlations, and offer recommendations for handling them. Other options do not directly address missing value analysis or are intended for different purposes.

Answer

Use the Data Wrangler 'Quick Model' to train a model that predicts missing values, then use the model's feature importance to infer missingness patterns.

Answer

Use the Data Wrangler 'Correlation Matrix' to compute pairwise correlations between all features, which will directly reveal the pattern of missing values.

Answer

Use the Data Wrangler 'Bias Report' to detect if missing values are disproportionately distributed across sensitive groups, which indicates the need for group-specific imputation.

Question 4

A data scientist is performing EDA on a dataset with 1,000 features and 10,000 observations. The target variable is continuous. They want to reduce dimensionality before training a model but need to preserve the interpretability of the original features as much as possible. Which technique is MOST suitable for this scenario?

Accepted Answer

Lasso regression (L1 regularization). Lasso regression (L1 regularization) is a feature selection method that shrinks less important feature coefficients to zero, effectively reducing the feature set while retaining the original features that remain. This preserves interpretability, unlike PCA or autoencoders which create new uninterpretable features. t-SNE is for visualization, not feature reduction for modeling.

Answer

t-Distributed Stochastic Neighbor Embedding (t-SNE)

Answer

Principal Component Analysis (PCA)

Answer

Autoencoder neural network

Question 5

A data scientist is performing exploratory data analysis on a dataset with 200 features and 500,000 observations. The target variable is continuous. The scientist suspects multicollinearity among features and wants to identify highly correlated feature pairs to potentially remove redundant features. The dataset is stored in Amazon S3 and is too large to load into memory. Which method is most appropriate for efficiently identifying highly correlated feature pairs?

Accepted Answer

Use Amazon SageMaker Data Wrangler to compute a correlation matrix for all features, then filter pairs with absolute correlation above a threshold.. SageMaker Data Wrangler is the most appropriate because it offers a built-in correlation matrix analysis that can handle large datasets through sampling or distributed processing. It provides a visual and filterable output, making it easy to identify highly correlated feature pairs. Other methods require custom code or are not optimized for dense correlation computations.

Answer

Use Amazon SageMaker Processing with a custom Apache Spark job to compute pairwise correlations using the MLlib library, then write the results to S3.

Answer

Use Amazon SageMaker Clarify to generate a correlation matrix as part of its bias detection report, then extract the feature correlations from the report.

Answer

Use Amazon Athena to run a SQL query that calculates the Pearson correlation coefficient for every pair of features, then export the results to Amazon QuickSight for visualization.

Question 6

A data scientist is analyzing a dataset with a categorical feature that has 50 unique levels. During EDA, they notice that some levels appear very infrequently, with counts below 10. They plan to use this feature in a tree-based model. Which encoding strategy is MOST appropriate to handle the high cardinality and rare levels?

Accepted Answer

Target encoding (mean encoding). Target encoding is most appropriate because it converts the high-cardinality categorical feature into a single numerical feature based on the target mean, capturing the relationship with the target. It handles rare levels via smoothing, reducing overfitting. One-hot and label encoding do not address rarity effectively, and frequency encoding ignores the target relationship.

Answer

One-hot encoding

Answer

Frequency encoding

Answer

Label encoding

Question 7

A team is building a regression model to predict house prices. The dataset includes a column 'zip_code' with 100 unique values. The data scientist one-hot encodes this column, resulting in 100 new binary columns. The model shows poor performance on a validation set. What is the most likely cause?

Accepted Answer

The model suffers from the curse of dimensionality due to the large number of features.. One-hot encoding 'zip_code' with 100 unique values creates 100 binary features. When combined with other features, the total number of features can be large relative to the number of training samples, leading to the curse of dimensionality. This causes the model to overfit the training data and generalize poorly to the validation set. While multicollinearity among one-hot encoded columns is often low due to their binary nature, the primary issue here is the high dimensionality relative to sample size. Option D correctly identifies this as the most likely cause.

Answer

One-hot encoding introduced multicollinearity among the binary columns.

Answer

One-hot encoding reduced the number of features, causing underfitting.

Answer

The one-hot encoding introduced high variance, but the validation set has low variance.

Question 8

During exploratory data analysis, a data scientist plots the distribution of a numerical feature and observes a heavy right skew. The feature has many outliers at the high end. Which transformation is most appropriate to reduce skewness?

Accepted Answer

Apply a log transformation to the feature.. A log transformation compresses the range of the data, reducing the impact of extreme values and pulling in the long tail of a right-skewed distribution. This makes the feature more normally distributed, which is often required for linear models and many statistical tests. It is the standard technique for handling positive-valued features with heavy right skew.

Answer

Apply z-score normalization.

Answer

Apply one-hot encoding.

Answer

Apply min-max scaling.

Question 9

A data scientist is analyzing a dataset with missing values in 30% of the rows for the 'age' column. The data scientist decides to impute the missing values with the median of the observed 'age' values. What is a potential drawback of this approach?

Accepted Answer

The imputed values may reduce the variance of the 'age' distribution.. Imputing missing values with the median of the observed data artificially concentrates imputed values around the center of the distribution. This reduces the overall variance of the 'age' column because the imputed values do not reflect the natural spread of the data, potentially distorting downstream analyses like regression or clustering that rely on variance structure.

Answer

The imputation will introduce bias if the missing values are not random.

Answer

Imputation using median is computationally expensive for large datasets.

Answer

The imputed values will increase the variance of the feature, leading to overfitting.

Question 10

A data scientist is exploring a dataset with 500 features and 10,000 samples. The data scientist computes the pairwise correlation matrix and finds that many features have correlations above 0.9. The data scientist wants to reduce the dataset to 50 features while preserving as much variance as possible. Which technique should be used?

Accepted Answer

Apply Principal Component Analysis (PCA) and keep the top 50 principal components.. Principal Component Analysis (PCA) is the correct technique because it performs an orthogonal linear transformation that projects the original 500 features into a new coordinate system where the axes (principal components) are ordered by the variance they capture. By keeping the top 50 principal components, the data scientist retains the maximum possible variance in the reduced 50-dimensional space, directly addressing the goal of preserving variance while handling high multicollinearity.

Answer

Remove all but one feature from each group of highly correlated features.

Answer

Use Linear Discriminant Analysis (LDA) to project to 50 dimensions.

Answer

Use t-Distributed Stochastic Neighbor Embedding (t-SNE) to reduce to 50 dimensions.

Question 11

During EDA, a data scientist notices that a feature has a high proportion of missing values (e.g., 70%). The feature is continuous and expected to be important based on domain knowledge. What is the best approach to handle this?

Accepted Answer

Create a binary indicator for missingness and impute the continuous values with the median.. It preserves the predictive signal from the feature while accounting for the pattern of missingness. Creating a binary indicator allows the model to learn whether missingness itself is informative, and median imputation is robust to outliers for a continuous feature. This approach avoids the bias of dropping the feature entirely and is more principled than arbitrary out-of-range imputation.

Answer

Remove the feature entirely to avoid bias.

Answer

Impute missing values with -1 since it is out of range.

Answer

Drop all rows with missing values in that feature.

Question 12

A data scientist is performing EDA on a dataset with 1,000 features and 10,000 rows. The target variable is binary. After checking for multicollinearity, the scientist finds many pairs of features with correlation > 0.95. Which action should be taken to prepare the data for modeling?

Accepted Answer

For each highly correlated pair, remove one feature based on domain knowledge or higher correlation with target.. When features are highly correlated (e.g., > 0.95), they introduce multicollinearity, which can destabilize coefficient estimates in linear models and reduce interpretability. Removing one feature from each correlated pair based on domain knowledge or its correlation with the target variable preserves predictive power while reducing redundancy. This approach is more targeted than PCA, which transforms features into uncorrelated components but sacrifices interpretability and may not align with the binary target.

Answer

Apply PCA to all features to decorrelate them.

Answer

Standardize all features using StandardScaler.

Answer

Randomly drop half of the correlated features.

Question 13

Which TWO of the following are appropriate techniques for detecting outliers in a univariate continuous feature?

Accepted Answer

Use Z-score and flag values with absolute Z-score > 3.. Option B is correct because the Z-score measures how many standard deviations a value lies from the mean, and in a univariate continuous feature a common rule flags observations with |Z| > 3 as outliers (roughly 0.3% of a normal distribution). Option E is correct because the IQR method is a standard univariate technique: it computes Q1 and Q3, defines the fences Q1 - 1.5*IQR and Q3 + 1.5*IQR, and flags values outside them as outliers. Option A is not appropriate because Random Forest is a supervised classifier requiring labeled outliers, which is not a univariate detection technique. Option C is wrong because one standard deviation is far too narrow (it would flag about 32% of normally distributed data), not a valid outlier threshold. Option D is not appropriate because DBSCAN is a multivariate density-based clustering algorithm, and its default parameters are not designed for univariate outlier detection.

Answer

Apply a Random Forest classifier to predict outliers.

Answer

Remove any value that is more than one standard deviation from the mean.

Answer

Use DBSCAN clustering with default parameters.

Question 14

A machine learning engineer is working on a customer churn prediction project. The dataset contains 100,000 records with 15 features, including customer demographics, account information, and usage patterns. The target variable 'churned' is binary with 15% positive examples. During EDA, the engineer notices that the feature 'tenure' (number of months the customer has been with the company) has a multimodal distribution with peaks at 1, 12, 24, and 36 months. Also, the feature 'monthly_charges' has a strong positive correlation with 'total_charges' (correlation coefficient = 0.95). The engineer wants to build a logistic regression model. Which preprocessing steps should the engineer take to address these issues? (Select TWO.)

Accepted Answer

Bin the 'tenure' feature into categorical groups (e.g., 0-6, 7-12, 13-24, 25-36, 36+) to capture the non-linear relationship.. Binning the 'tenure' feature into categorical groups (e.g., 0-6, 7-12, 13-24, 25-36, 36+) captures the multimodal distribution and non-linear relationship with churn, which logistic regression (a linear model) cannot model directly. This transforms the feature into a format that allows the model to learn different churn probabilities for each tenure segment without imposing a linear assumption. Option B is correct because 'monthly_charges' and 'total_charges' have a very high positive correlation (0.95), leading to multicollinearity. Removing one of these features (e.g., 'total_charges') reduces multicollinearity, which improves the stability and interpretability of logistic regression coefficients.

Answer

Apply log transformation to the 'tenure' feature to make it unimodal.

Answer

Create polynomial features up to degree 3 for 'tenure' to capture non-linearity.

Answer

Standardize all numerical features to have mean 0 and variance 1.

Question 15

A machine learning engineer is working on a regression problem to predict house prices. The dataset contains 500,000 rows and 20 features, including 'sqft_living', 'bedrooms', 'bathrooms', 'floors', 'waterfront', 'view', 'condition', 'grade', 'yr_built', 'zipcode', and 'lat'. After performing exploratory data analysis, the engineer notices that the 'sqft_living' feature has a right-skewed distribution with a long tail. The 'zipcode' feature is categorical with 70 unique values. The 'lat' feature is continuous. The engineer wants to prepare the data for a linear regression model. Which action should the engineer take to improve model performance?

Accepted Answer

Apply a log transformation to the 'sqft_living' feature.. Linear regression assumes that features are approximately normally distributed, and a right-skewed distribution like 'sqft_living' can violate this assumption, leading to poor model performance. Applying a log transformation compresses the long tail, making the distribution more symmetric and helping the model learn a linear relationship between the feature and the target. This is a standard preprocessing step for skewed features in regression tasks.

Answer

Remove the 'sqft_living' feature because it violates the normality assumption.

Answer

One-hot encode the 'zipcode' feature to capture location effects.

Answer

Apply standard scaling (z-score) to the 'sqft_living' feature.