CCNA Exploratory Data Analysis Questions

Option A is correct because PCA is a linear technique that preserves global variance. Option B is wrong because t-SNE focuses on local structure. Option C is wrong because LDA requires labels.

Option D is wrong because Autoencoders are more complex and not primarily for visualization.

Option B is correct because VIF quantifies multicollinearity by measuring how much the variance of a coefficient is inflated due to correlation with other features. Option A is wrong because PCA creates new features, not identification of redundant ones. Option C is wrong because correlation matrix shows pairwise correlations but VIF is more comprehensive.

Option D is wrong because Chi-square is for categorical features.

Option B is correct because Athena writes query results to an S3 bucket, which requires s3:PutObject permission on the output bucket. Option A is wrong because the policy includes necessary Athena permissions. Option C is wrong because glue:GetTable is already included.

Option D is wrong because s3:ListBucket is included.

The chi-square test of independence is appropriate for testing association between categorical features and a binary target. ANOVA is for continuous target. Mutual information measures dependency but is not a hypothesis test.

Point-biserial correlation is for continuous and binary. Cramér's V is a measure of association after chi-square.

Option D is correct because understanding the distribution of features across churn and non-churn classes helps identify which features drive churn. Option A is wrong because PCA reduces dimensionality but does not address imbalance. Option B is wrong because cross-validation is a modeling step, not EDA.

Option C is wrong because removing outliers may worsen imbalance.

A skewness > 1 indicates a highly right-skewed (positive skew) distribution, meaning the tail extends to the right. Option A is wrong because left skew is negative. Option B is wrong because symmetric distributions have skewness near 0.

Option D is wrong because skewness describes shape, not the presence of outliers specifically.

Option B is correct because using SageMaker Data Wrangler's sampling capabilities allows faster exploration while preserving statistical properties for feature selection. Option A is wrong because reducing to a single partition loses time series context. Option C is wrong because moving to a smaller instance may cause memory issues.

Option D is wrong because random sample of rows may break time series ordering.

Option B is correct because checking the schema and data type conversion can reveal issues like unquoted commas or wrong format. Option A is wrong because the issue is likely with data types, not encryption. Option C is wrong because adding a partition won't fix data type issues.

Option D is wrong because crawling does not change data types if schema is inferred incorrectly.

Option C is correct because SageMaker Data Wrangler provides a visual interface to apply transformations like log scaling without writing code. Option A is wrong because SageMaker Ground Truth is for labeling, not transformation. Option B is wrong because SageMaker JumpStart is for pre-built models.

Option D is wrong because SageMaker Clarify is for bias detection and explainability.

AWS Glue DataBrew provides a visual interface to flatten nested JSON and compute derived metrics like session duration without writing code. Option B (Athena with JSON SerDe) can query but requires SQL that handles arrays. Option C (EMR with Spark) is more complex.

Option D (QuickSight) is visualization only.

Option D is correct because Amazon S3 Select can query a subset of rows from S3 without loading the entire dataset, enabling quick profiling. Option A is wrong because loading full data is expensive and slow. Option B is wrong because Glue ETL processes full dataset.

Option C is wrong because converting to Parquet adds overhead for initial profiling.

Using Spark on SageMaker allows distributed processing of large data. Option A is wrong because pandas may run out of memory. Option B is wrong because Athena requires SQL queries and external setup.

Option D is wrong because DataBrew is for profiling but not custom duplicate analysis.

Amazon QuickSight is a serverless, machine learning-powered business intelligence service that allows users to create interactive dashboards and visualizations without writing code. Option A is wrong because SageMaker Studio requires coding for custom visualizations. Option B is wrong because AWS Glue is used for ETL, not visualization.

Option D is wrong because Amazon Athena is a query service. Option E is wrong because Amazon EMR is a big data platform not primarily for visualization.

Option B is correct because AWS Glue DataBrew provides a visual interface to profile data and compute summary statistics without writing code. Option A is wrong because Amazon Athena requires SQL queries and more manual effort. Option C is wrong because Amazon EMR requires cluster setup and management.

Option D is wrong because Amazon SageMaker Data Wrangler is a good tool but requires more configuration than DataBrew for simple summary statistics.

Option A is correct because the log transformation is commonly used to stabilize variance when variance increases with the mean. Box-Cox (B) is more general but requires positive data. Differencing (C) is for trend/seasonality, not variance.

Min-max scaling (D) does not stabilize variance.

Visualizing class imbalance and identifying missing values are fundamental EDA steps. Option B (PCA) is for dimensionality reduction, not initial EDA. Option D (one-hot encoding) is for categorical variables, but not an EDA action.

Option E (gradient boosting) is modeling, not EDA.

Correlation analysis with the target variable is a quick way to rank features. Option B is wrong because PCA is unsupervised and does not rank features by importance to target. Option C is wrong because t-SNE is for visualization.

Option D is wrong because k-means is clustering.

Log transformation compresses the range of values and reduces the impact of outliers. Standardization (z-score) does not reduce outlier impact. Min-max scaling is sensitive to outliers.

Square root transformation is less effective than log for large outliers. Binning loses information.

Option B is correct because Amazon Athena allows running SQL queries directly on data in S3, including counting duplicates. Option A is wrong because AWS Glue is an ETL service, not a query engine. Option C is wrong because Amazon Redshift Spectrum can query data in S3 but requires a Redshift cluster.

Option D is wrong because Amazon SageMaker Studio is an IDE, not a query service.

Options B and C are correct because a high correlation indicates a strong linear relationship and suggests multicollinearity. Option A is wrong because correlation does not imply causation. Option D is wrong because a high positive correlation means they move together, not opposite.

Option E is wrong because correlation measures linear relationship, not independence.

Option B is correct because median imputation is robust to outliers and simple for numerical features. Option A is wrong because dropping rows with any missing values would remove 25% of data. Option C is wrong because mode is for categorical.

Option D is wrong because model-based imputation is complex for initial EDA.

Option D is correct because the number of rows can vary between files, leading to different file sizes. Option A is wrong because different column types would cause inconsistent schemas, but the engineer says same number of columns. Option B is wrong because compression would be applied uniformly.

Option C is wrong because S3 does not add headers multiple times.

Option A is correct because log transformation is commonly used to reduce right skewness. Option B is wrong because min-max scaling does not change distribution shape. Option C is wrong because one-hot encoding is for categorical variables.

Option D is wrong because standardization does not reduce skewness.

Option B is correct because Amazon SageMaker provides built-in algorithms like K-Means for clustering, and the scientist can use custom code with libraries like Geopy to compute distances. Option A is wrong because Amazon Athena with PostGIS is for querying geospatial data, not for clustering. Option C is wrong because Amazon Location Service is for maps and location tracking, not for analytical clustering.

Option D is wrong because AWS Glue is for ETL, not for analysis and clustering.

A box plot with log scale is effective for skewed data as it shows outliers and distribution shape after transformation. Histogram with log scale also works. KDE is similar to histogram.

Q-Q plot checks normality. Scatter plot is for two variables.

Partitioning the table on a frequently filtered column limits the data scanned to relevant partitions. Option A is wrong because compressing reduces storage but not scan size unless combined with columnar format. Option C is wrong because converting to JSON does not reduce scan.

Option D is wrong because Parquet is columnar and can reduce scan, but without partitioning, Athena still scans entire columns.

Option D is correct because Amazon SageMaker Processing jobs can run custom scripts that can handle high cardinality categorical features using libraries like pandas. Option A is wrong because SageMaker Data Wrangler is a visual tool, but it may not be as flexible for custom grouping logic. Option B is wrong because SageMaker Autopilot is for automated ML, not for custom data processing.

Option C is wrong because AWS Glue Studio is a visual ETL tool, but it may not be as tightly integrated with SageMaker.

Option B is correct because Amazon Comprehend can perform sentiment analysis and extract key phrases. Option A is wrong because Amazon Rekognition is for image/video analysis. Option C is wrong because Amazon Polly is a text-to-speech service.

Option D is wrong because Amazon Transcribe is for speech-to-text.

Option D is correct because the AUC ROC curve is independent of class distribution and provides a robust measure of separability between classes. Option A is wrong because accuracy is misleading for imbalanced data. Option B is wrong because RMSE is for regression.

Option C is wrong because F1 score depends on a threshold and can be affected by sampling.

Option B is correct because the file size is approximately 1 GB (1073741824 bytes = 1 GB), and pandas typically requires 3-5x the file size in memory for CSV parsing, which would exceed the 8 GB RAM. Option A is wrong because the file is accessible (HTTP 200). Option C is wrong because the content type is text/csv, which is correct.

Option D is wrong because the metadata indicates 10 million rows, but the file size is the main issue.

Option B is correct because near-zero variance features provide little information and can be removed. Option A is wrong because high variance is often useful. Option C is wrong because high correlation between two features might warrant removal of one, but variance is not the direct indicator.

Option D is wrong because negative covariance is still informative.

Option B is correct. The command lists objects in the bucket under prefix 'data/' that are greater than 1,000,000 bytes (Size > '1000000') and returns their keys. Option A is wrong because size is in bytes, not 1 MB (1 MB = 1,048,576 bytes).

Option C is wrong because it returns keys, not counts. Option D is wrong because it filters by size, not date.

Option B is correct because a histogram with a log scale can handle skewed data. Option A is wrong because a box plot shows quartiles but not the full distribution shape. Option C is wrong because a scatter plot is for two variables.

Option D is wrong because a line chart is for time series.

Chi-squared test is for categorical features and mutual information can handle both. Option C is wrong because PCA is unsupervised. Option D is wrong because k-means is clustering.

Option E is wrong because LSTM is a model, not for EDA.

Option D is correct because the negative values are errors and likely distort the distribution; removing them is straightforward and valid. Option A is wrong because taking absolute values would incorrectly treat errors as legitimate high values. Option B is wrong because negative values are not missing, so imputation is not appropriate.

Option C is wrong because capping may still retain erroneous values.

Option B is correct because the policy requires that all PutObject requests include server-side encryption with AES256. Option A is wrong because the policy allows GetObject and PutObject, but with a condition. Option C is wrong because the condition is on the encryption, not on the bucket.

Option D is wrong because the error is AccessDenied, not NoSuchBucket.

Option C is correct because PCA is linear, fast, and preserves global variance structure. Option A is wrong because t-SNE is non-linear, slower, and focuses on local structure. Option B is wrong because UMAP can be slow and is also non-linear.

Option D is wrong because LDA is supervised and requires labels.

Option C is correct because missingness related to the value itself means that the missing data are systematically different; creating a 'missing' category allows the model to learn the pattern. Option A is wrong because mean imputation reduces variance and ignores the systematic difference. Option B is wrong because median imputation has similar issues.

Option D is wrong because random imputation introduces noise without capturing the missingness pattern.

Log transformation is commonly applied to right-skewed data to make it more normally distributed, which can improve model performance. Option A (StandardScaler) is for scaling, not skewness. Option B (MinMaxScaler) also doesn't address skewness.

Option D (One-hot encoding) is for categorical variables.

Options A, C, and E are correct. Log transformation, Box-Cox, and Yeo-Johnson are common transformations for right-skewed data. Option B is wrong because StandardScaler only standardizes, does not reduce skewness.

Option D is wrong because MinMax scaling does not affect skewness.

Using Amazon Athena with columnar formats like Parquet after converting from CSV reduces query costs and improves performance. Option A (SageMaker Data Wrangler) may struggle with 2000 columns. Option B (AWS Glue ETL) is more expensive and slower for simple statistics.

Option D (Redshift Spectrum) requires setting up a Redshift cluster, which is overkill.

Option D is correct because the large discrepancy between training and test accuracy indicates overfitting, and pruning the decision tree (e.g., limiting max_depth) reduces overfitting. Option A is wrong because more features may worsen overfitting. Option B is wrong because more data may help but is not the first step; also data is limited.

Option C is wrong because cross-validation is a technique to evaluate model performance but does not directly fix overfitting; pruning does.

Checking for duplicate rows and analyzing missing values are fundamental steps in EDA. Option B is wrong because dropping user_id before analysis may lose information. Option C is wrong because training a model is not part of EDA.

Option D is wrong because PCA is not for data quality.

Option A is correct because Amazon QuickSight integrates with S3 and can generate histograms and summary statistics without needing to move data. Option B is wrong because SageMaker is for model building, not quick ad-hoc analysis. Option C is wrong because Athena is for querying, not visualization.

Option D is wrong because Glue is for ETL, not analysis.

Options A, C, and E are correct. A: Checking data types is fundamental. C: Descriptive statistics summarize distributions.

E: Visualizing with histograms reveals patterns. B: Splitting into train/test is modeling, not EDA. D: Building a linear model is modeling, not EDA.

Correct options: A and D. VIF (A) quantifies multicollinearity; PCA (D) creates orthogonal components. Option B is wrong because correlation matrix only shows pairwise correlations.

Option C is wrong because Lasso does feature selection but does not identify multicollinearity. Option E is wrong because RFE is for feature selection, not multicollinearity detection.

Option A is correct because scatter plots of feature pairs can reveal linear relationships. Option C is correct because correlation matrix with heatmap quantifies pairwise correlations. Option B is wrong because histogram shows distribution, not relationship.

Option D is wrong because VIF is a statistical test, but the question asks for EDA techniques (visual). Option E is wrong because box plot shows univariate distribution.

Option B is correct because Amazon SageMaker Data Wrangler can connect to S3, perform correlation analysis, and generate a correlation matrix without writing code. Option A is wrong because AWS Glue ETL is for ETL pipelines, not interactive exploration. Option C is wrong because Amazon Athena is for SQL queries, not correlation analysis.

Option D is wrong because Amazon QuickSight is for visualization, not statistical correlation.

Option D is correct because log transformation is commonly used for right-skewed data to reduce skewness. Option A is wrong because min-max scaling does not change distribution shape. Option B is wrong because square transformation increases skewness.

Option C is wrong because binning loses information.

For large datasets, undersampling the majority class is feasible and can be effective. Option A is wrong because SMOTE generates synthetic samples but may be computationally expensive for 10M rows. Option B is wrong because class weights are set during training, not EDA.

Option D is wrong because collecting more data is not guaranteed to fix imbalance.

Option D is correct because sine and cosine transformations capture cyclic patterns like time of day or day of year. Option A is wrong because one-hot encoding creates many sparse features. Option B is wrong because converting to Unix timestamp loses cyclic nature.

Option C is wrong because binning loses granularity.

Options A and C are correct. VIF is a standard measure for detecting multicollinearity; removing features with high VIF reduces multicollinearity. Ridge regression (L2 regularization) can handle multicollinearity by penalizing large coefficients.

Option B is wrong because PCA reduces dimensionality but makes models less interpretable and does not directly handle multicollinearity in the original features. Option D is wrong because stepwise selection does not specifically address multicollinearity. Option E is wrong because standardization does not affect collinearity.

Option B is correct because the F1 score balances precision and recall, which is suitable for imbalanced datasets. Option A is wrong because accuracy can be misleading with imbalance. Option C is wrong because precision alone ignores recall.

Option D is wrong because AUC-ROC may still be high even with poor recall.

Options A and D are correct. Sampling reduces memory usage and speeds up interactive analysis. Option B is wrong because sampling can miss rare events.

Option C is wrong because model accuracy typically decreases with less data. Option E is wrong because sampling does not remove bias from the original data.

Option C is correct because count encoding uses frequency counts, which can capture information for rare categories without creating high dimensionality. One-hot encoding (A) would create 1,000 columns. Target encoding (B) can cause overfitting.

Label encoding (D) implies ordinality.

Option B is correct because Amazon Kinesis Data Analytics enables running SQL queries on streaming data in real-time. Option A is wrong because Athena is for batch queries on S3. Option C is wrong because Glue is for ETL, not real-time SQL.

Option D is wrong because SageMaker is for ML model training, not streaming SQL.

Option B is correct because Multiple Imputation by Chained Equations (MICE) accounts for relationships between features and preserves variability. Option A is wrong because mean imputation can bias estimates when data is not missing completely at random. Option C is wrong because dropping rows reduces sample size and may introduce bias.

Option D is wrong because last observation carried forward is for time series.

Option B is correct because a correlation of 0.85 indicates a strong positive linear relationship. Option A is wrong because 0.85 is far from 1. Option C is wrong because correlation does not imply causation.

Option D is wrong because 0.85 is not weak.

Option A is correct because high correlation indicates multicollinearity; removing one feature reduces redundancy. Option C is correct because PCA can create uncorrelated components. Option B is wrong because adding interaction terms would increase multicollinearity.

Option D is wrong because correlation does not imply missing values. Option E is wrong because log transformation does not address correlation between features.

Option D is correct because Isolation Forest is effective for high-dimensional data and uses tree-based isolation. Option A is wrong because Z-score assumes normality and is univariate. Option B is wrong because IQR is univariate and not suitable for high dimensions.

Option C is wrong because Mahalanobis distance assumes multivariate normality and is sensitive to dimensionality.

Option A is correct because checking linearity (e.g., scatter plots of features vs. target) is fundamental for linear model assumptions. Option B is wrong because multicollinearity affects linear regression but not tree models. Option C is wrong because class imbalance is for classification.

Option D is wrong because outlier detection is important but not the primary factor for model selection.

Correct options: A, B, D. Log transform (A), Box-Cox (B), and Yeo-Johnson (D) are common for normalizing skewed data. Option C is wrong because standardization (Z-score) does not change distribution shape.

Option E is wrong because min-max scaling does not fix skewness.

Option B is correct because box plots show outliers. Option D is correct because summary statistics (min, max, percentiles) reveal extreme values. Option A is wrong because removing outliers before understanding context is premature.

Option C is wrong because log transformation changes scale but does not help identify outliers. Option E is wrong because mean imputation would distort distribution.