CCNA Exploratory Data Analysis Questions

Option B is correct because the rank transformation maps values to their ranks, eliminating the impact of outliers while preserving order. Option A is wrong because Box-Cox requires positive values and may not reduce outlier influence. Option C is wrong because log transformation can reduce skew but still allows outliers to remain influential.

Option D is wrong because min-max scaling does not reduce outlier influence; it compresses the range.

Differencing is a common technique to remove trends and seasonality to make a time series stationary. Option B is wrong because logarithmic transformation stabilizes variance but does not remove trends. Option C is wrong because min-max scaling does not address trends.

Option D is wrong because one-hot encoding is for categorical variables. Option E is wrong because PCA is for dimensionality reduction.

Using Amazon Athena with SQL queries allows serverless querying of data in S3, and is cost-effective (pay per query). It can compute summary statistics directly on the data without moving it. Option A is wrong because SageMaker Data Wrangler requires importing data into SageMaker, which may incur transfer costs and time.

Option B is wrong because Glue DataBrew also processes data but may be more expensive for large datasets. Option D is wrong because S3 Select works on single objects and is limited. Option E is wrong because launching an EMR cluster adds overhead and cost for a simple task.

Option B is correct because MICE (Multiple Imputation by Chained Equations) models each variable with missing values as a function of other variables, preserving relationships. Option A is wrong because listwise deletion can introduce bias. Option C is wrong because mean/median imputation reduces variance.

Option D is wrong because forward-fill is for time series.

Option A is correct because the correlation matrix shows pairwise Pearson correlations, which can indicate high collinearity. Option B is wrong because VIF requires more variables than observations. Option C is wrong because chi-square is for categorical.

Option D is wrong because covariance alone is scale-dependent.

Option A is correct because median imputation is robust to outliers and preserves the dataset size. Option B is wrong because dropping rows with missing values would lose 14% of data. Option C is wrong because mean imputation can be affected by outliers.

Option D is wrong because KNN imputation may introduce bias and is computationally expensive.

Options B and C are correct. High correlation can lead to multicollinearity, so removing one feature (B) or combining them (C) are valid approaches. Option A is wrong because increasing regularization is a remedy for multicollinearity but does not directly address the correlation.

Option D is wrong because scaling does not affect correlation. Option E is wrong because adding interaction terms can increase multicollinearity.

Option B is correct because AWS Glue provides built-in transforms in its ETL jobs to parse timestamps and extract date/time components. Option A is wrong because Athena is a query engine and can extract date parts using SQL, but that requires writing SQL queries, not a no-code solution. Option C is wrong because SageMaker Data Wrangler is a visual tool that can create features, but it requires a SageMaker Studio environment.

Option D is wrong because QuickSight is a visualization tool, not for feature engineering.

Option A is correct because Amazon SageMaker Canvas supports time-series forecasting with built-in decomposition. Option B is wrong because AWS Glue does not have time-series decomposition. Option C is wrong because Amazon QuickSight does not decompose time series.

Option D is wrong because Amazon Athena does not have decomposition functions.

Option C is correct because removing stop words (common words like 'the', 'and') ensures that the word cloud highlights meaningful words. Stemming (A) may not be necessary for a word cloud. Tokenization (B) is fundamental but not the most critical for meaningfulness.

POS tagging (D) is overkill.

Options A, C, and E are correct. A: Correlation matrix quantifies linear relationships. C: Pair plots allow visual inspection of multiple relationships.

E: Box plots by target show distribution differences. B is wrong because PCA is for dimensionality reduction, not EDA of relationships. D is wrong because clustering is unsupervised and not directly for feature-target relationships.

Option C is correct because a box plot explicitly shows median, quartiles, and outliers. Option A is wrong because scatter plots show relationships between two variables, not distribution. Option B is wrong because bar charts are for categorical data.

Option D is wrong because heatmaps show correlations, not distribution.

Option C is correct because autocorrelation function (ACF) can show peaks at lag 7 indicating weekly seasonality. Option A is wrong because line plot can show patterns but is subjective. Option B is wrong because moving average smooths data and may hide seasonality.

Option D is wrong because spectral analysis is for frequency, but ACF is simpler.

Options A and C are correct: Z-score identifies outliers based on standard deviations from the mean; IQR uses the interquartile range. DBSCAN (B) is for multivariate clustering. Mahalanobis distance (D) is multivariate.

Cook's distance (E) is for regression influence.

Using Data Wrangler's built-in data quality and insights report is the most efficient way to get statistics and detect issues without custom code. Option B (Athena) requires writing SQL queries. Option C (QuickSight) needs exporting.

Option D (EMR) is overkill.

Option C is correct because understanding the distribution of the target variable is essential for imbalanced datasets to choose appropriate sampling techniques or evaluation metrics. Option A is wrong because correlation analysis is less critical than target distribution. Option B is wrong because PCA is a dimensionality reduction technique not primarily for EDA.

Option D is wrong because missing value imputation is important but not the most critical for imbalance.

Option D is correct because a log transformation is commonly used for right-skewed positive data to reduce skewness. Option A is wrong because StandardScaler does not change skewness. Option B is wrong because Min-Max scaling does not change shape.

Option C is wrong because a square transformation would increase skewness.

Option A is correct because Amazon Comprehend is a natural language processing (NLP) service that can perform sentiment analysis directly. Option B is wrong because Amazon Lex is for conversational interfaces, not text analysis. Option C is wrong because Amazon Rekognition is for image and video analysis.

Option D is wrong because Amazon Textract is for extracting text from documents, not sentiment.

Option B is correct because principal component analysis (PCA) is a dimensionality reduction technique that identifies the principal components capturing the most variance. Option A is wrong because correlation matrix only shows pairwise linear relationships, not importance. Option C is wrong because regularization can shrink coefficients but is not a dedicated dimensionality reduction technique.

Option D is wrong because random forests can provide feature importance but are not a dimensionality reduction technique per se.

Option C is correct because Amazon Athena allows querying Redshift data directly via federated queries, returning only aggregated results, avoiding the need to move large datasets. Option A is wrong because reading all data to a local DataFrame would exceed memory. Option B is wrong because writing to S3 and then reading with pandas still requires loading all data into memory.

Option D is wrong because SageMaker notebook's local memory is still limited.

Option D is correct because converting the string to a date type in the query allows Athena to use partition pruning if the table is partitioned by date, reducing scanned data. Option A is wrong because CAST in SELECT still scans all data. Option B is wrong because creating a view does not reduce data scanned.

Option C is wrong because converting to Parquet is beneficial but not the most direct for the given task.

Options A and C are correct. Visualizing missing data patterns (A) helps understand the missing mechanism. Using imputation methods like mean/median (C) is common during EDA.

Option B is wrong because deleting all rows with missing values may discard too much data. Option D is wrong because ignoring missing values can lead to errors. Option E is wrong because replacing with -1 can distort data.

Option B is correct because log transformation is commonly used to reduce right skewness by compressing the long tail. Option A is wrong because the square root transformation is less effective for severe skewness. Option C is wrong because Box-Cox requires all values positive and is a family that includes log, but the log is a specific case.

Option D is wrong because standardization does not change the shape of the distribution.

Options A and D are correct. Imputing with mean/median is a common technique, and using a model to predict missing values is also valid. Option B is wrong because dropping all rows with missing values can discard too much data.

Option C is wrong because filling with zeros may not be appropriate for all features. Option E is wrong because removing the feature entirely may lose important information.

Option B is correct because the file is large but the query returned only one row, suggesting the table's partition mapping is wrong; the WHERE clause on date may not match the actual partition. Option A is wrong because if the table were external, it would still read all data. Option C is wrong because 1 MB file likely has many rows.

Option D is wrong because Athena supports CSV.

Mutual information can capture any kind of dependency (including nonlinear) between features and target. Option A (Pearson correlation) only linear. Option B (Chi-squared test) is for categorical features.

Option D (ANOVA) is for comparing means across groups.

The table is not partitioned (empty PartitionKeys), and the SerDe expects comma-delimited files. If some files use a different delimiter (e.g., tab), those rows may not be parsed correctly, leading to fewer rows counted. Option A is wrong because the table schema matches the data.

Option B is wrong because Athena can handle many small files, but count should still include all rows. Option D is wrong because Athena does not skip files based on size unless explicitly filtered.

Option A is correct because side-by-side boxplots show distribution and outliers across categories. Option C is correct because overlapping histograms with transparency show distribution shapes. Option B is wrong because scatter plots require two numerical variables.

Option D is wrong because heatmaps are for correlation or contingency tables. Option E is wrong because bar charts are for categorical data, not numerical distributions.

Option B is correct because analyzing the missing pattern (e.g., MCAR, MAR, MNAR) guides imputation strategy. Option A is wrong because mode imputation is a method, not a diagnostic step. Option C is wrong because label encoding is for modeling, not imputation decision.

Option D is wrong because dropping rows may lose data; need to understand missingness first.

Options B and D are correct. Option A is wrong because scatter plot is for two variables. Option C is wrong because heatmap shows correlation.

Option E is wrong because bar chart is for categorical data.

Option C is correct because mutual information can capture any dependency. Option A is wrong because Pearson correlation only captures linear. Option B is wrong because Spearman captures monotonic but not all non-linear.

Option D is wrong because ANOVA is for categorical vs continuous.

Option C is correct because log transformation is commonly used to reduce right skewness. Option A is wrong because standardization does not change distribution shape. Option B is wrong because min-max scaling does not change shape.

Option D is wrong because one-hot encoding is for categorical variables, not continuous.

Option C is correct because Isolation Forest is an ensemble method that isolates anomalies effectively in high-dimensional spaces without assuming distribution. Option A is wrong because Z-score assumes normal distribution. Option B is wrong because IQR is univariate.

Option D is wrong because DBSCAN is for clustering, not specifically for outlier detection.

Options B and E are correct. B: Visualizing the distribution helps identify shape and outliers. E: Reporting missing value counts is a standard EDA step.

A is wrong because imputation should be done after analysis. C is wrong because removing outliers without analysis is premature. D is wrong because normalization is not a first step.

Option B is correct because the ETag can be used to verify file integrity. Option A is wrong because ContentLength shows size in bytes, ContentLength 1048576 is 1 MB. Option C is wrong because ETag is not a version ID; S3 versioning uses VersionId.

Option D is wrong because preprocessed metadata indicates the file has been processed.

Option C is correct because Data Wrangler allows specifying custom missing value symbols during import. Option A is wrong because dropping rows is premature. Option B is wrong because replacing after import is less efficient.

Option D is wrong because imputation is needed after treating missing values.

Option C is correct because the policy allows s3:GetObject only on the 'my-training-data' bucket, but Athena writes results to a different S3 location (specified in the query configuration). Without permission to read from that output bucket, the notebook cannot load the results. Option A is wrong because Athena actions are allowed.

Option B is wrong because listing the training data bucket is allowed. Option D is wrong because reading individual objects from the training data bucket is allowed.

Option D is correct because SMOTE generates synthetic samples for the minority class. Option A is wrong because it discards majority class data. Option B is wrong because scaling does not balance classes.

Option C is wrong because encoding is for categorical features.

Option A is correct because using SPICE in-memory engine speeds up queries by caching data. Option B is wrong because sampling loses accuracy. Option C is wrong because reducing bin count is a trade-off.

Option D is wrong because Athena is query engine, but QuickSight already uses it; SPICE is better.

Option C is correct because object dtype typically indicates string or mixed types; converting to numeric allows mathematical operations. Option A is wrong because object dtype does not necessarily indicate missing values. Option B is wrong because object dtype is not automatically numeric.

Option D is wrong because normalization is done after conversion.

Option A is correct because ANOVA F-statistic tests if means across groups are significantly different. Option B is wrong because chi-square is for categorical features. Option C is wrong because mutual information is for feature selection but doesn't directly test separability.

Option D is wrong because variance ratio is not standard.

Option C is correct because dropping one of the highly correlated features reduces redundancy and mitigates multicollinearity. Option A is wrong because scaling does not address collinearity. Option B is wrong because PCA creates orthogonal components but reduces interpretability; dropping a feature is more straightforward.

Option D is wrong because adding polynomial features increases correlation.

Options B and D are correct. B: Visualizing missing patterns (e.g., with a missingno matrix) is a good EDA practice. D: Understanding the mechanism (MCAR, MAR, MNAR) is critical for choosing imputation method.

A is wrong because listwise deletion can introduce bias if data not MCAR. C is wrong because mean imputation reduces variance. E is wrong because missing values should be handled before model training, not after.

Option C is correct because a histogram or density plot reveals skewness visually. Option A is wrong because count plot is for categorical data. Option B is wrong because scatter plot shows relationship between two variables.

Option D is wrong because heatmap shows correlation, not skewness.

Option C is correct because the IQR method identifies outliers as points outside 1.5*IQR from the quartiles, which is robust to non-normal distributions. Option A is incorrect because Z-score assumes normality and is sensitive to extreme outliers. Option B is incorrect because Mahalanobis distance is for multivariate outliers.

Option D is incorrect because PCA is for dimensionality reduction, not outlier detection.

Option B is correct because checking the partition metadata using SHOW PARTITIONS or querying the information_schema helps verify that the partition structure is correct and that queries are using partition pruning. Option A is incorrect because compressing data reduces storage but does not directly affect partition pruning. Option C is incorrect because converting to Parquet improves columnar scanning but does not address partition misuse.

Option D is incorrect because increasing workers does not fix incorrect partition usage.

Option A is correct because manually checking files confirms data presence. Option B is correct because MSCK REPAIR TABLE adds partitions not yet registered. Option C is correct because incorrect data types can cause filters to exclude rows.

Option D is wrong because the query does not automatically update partitions. Option E is wrong because converting to Parquet is not a troubleshooting step.

Option A is correct because a histogram with hue (color) for each group can show the separate distributions. Option B is wrong because a box plot shows summary statistics but not the shape. Option C is wrong because a scatter plot is for two continuous variables.

Option D is wrong because a bar chart is for categorical data.

SMOTE (Synthetic Minority Over-sampling Technique) is a popular method for handling imbalanced datasets by generating synthetic samples for the minority class. Option A is wrong because removing outliers does not address class imbalance. Option B is wrong because feature scaling does not affect imbalance.

Option D is wrong because PCA is for dimensionality reduction, not imbalance. Option E is wrong because cross-validation is a model evaluation technique, not an imbalance solution.

Option D is correct because stratified random sampling ensures the proportion of classes is preserved in both training and test sets. Option A is wrong because simple random sampling may result in uneven distribution. Option B is wrong because oversampling should be done after splitting to avoid data leakage.

Option C is wrong because k-fold cross-validation is not a split method.

Options C and D are correct because imputation with median/mode is robust, and flagging missingness with an indicator variable preserves information. Option A is wrong because listwise deletion can introduce bias and reduce sample size. Option B is wrong because mean imputation is sensitive to outliers.

Option E is wrong because forward fill is for time series, not general EDA.

Option B is correct because converting to datetime allows extracting day-of-week. Option A is wrong because binning into hours loses day information. Option C is wrong because converting to string does not facilitate analysis.

Option D is wrong because scaling does not help.

Option B is correct because Amazon QuickSight is a BI service that can connect to S3 data and create histograms without coding. Option A is wrong because Amazon SageMaker Studio requires code or notebook. Option C is wrong because Amazon Athena outputs query results, not visualizations directly.

Option D is wrong because AWS Glue is for ETL, not visualization.

Option B is correct because binning the transaction amount and computing fraud rates per bin can reveal non-linear relationships that correlation might miss. Option A is wrong because log transformation does not reveal relationship with target. Option C is wrong because correlation is already computed and can mask non-linearity.

Option D is wrong because feature importance from a tree model is more appropriate after feature engineering, not during EDA.

Option C is correct because using SageMaker Processing with a custom script can distribute the computation across multiple instances, making it efficient for large datasets. Option A is wrong because loading all images into memory on a single notebook instance is not feasible. Option B is wrong because Athena is designed for structured data, not images.

Option D is wrong because AWS Glue is for ETL on tabular data, not image processing.

Option D is correct because mutual information captures any kind of dependency, including non-linear. Option A is wrong because Pearson correlation only measures linear relationships. Option B is wrong because PCA is for dimensionality reduction, not feature-target relationship.

Option C is wrong because correlation matrix is pairwise among features, not with target.

Option D is correct because a box plot shows distribution of continuous feature across binary classes. Option A is wrong because scatter plot is for two continuous variables. Option B is wrong because histogram shows distribution of one variable.

Option C is wrong because bar chart is for categorical features.

Checking for class imbalance is critical because it can cause a model to predict the majority class and still achieve high accuracy, but fail on the minority class in unseen data. Option A is wrong because log transformation is for skewness, not class imbalance. Option C is wrong because a correlation matrix helps with multicollinearity.

Option D is wrong because missing value heatmaps show missing data patterns.

Option C is correct because Amazon Athena allows running SQL queries directly on data in S3, and the COUNT and CASE statements can compute missing value percentages efficiently without moving data. Option A is wrong because AWS Glue Crawler only catalogs metadata, not performing data analysis. Option B is wrong because SageMaker Notebook requires manual coding and is less efficient for quick checks.

Option D is wrong because QuickSight is a visualization tool, not for direct SQL-based analysis.

Option C is correct because high correlation indicates multicollinearity, which can be addressed by removing one feature. Option A is wrong because scaling does not help. Option B is wrong because interaction terms increase multicollinearity.

Option D is wrong because regularization helps but is not the first step; removal is simpler.

Option B is correct because a histogram shows the frequency distribution of a continuous variable. Option A is wrong because a scatter plot shows relationship between two variables. Option C is wrong because a box plot shows summary statistics but not full distribution.

Option D is wrong because a bar chart is for categorical data.

Mean imputation is suitable for normally distributed data as it preserves the mean. Median is robust to outliers, not normality. Mode is for categorical data.

Standard deviation is not an imputation method. KNN imputation is non-parametric.

One-hot encoding 'zip_code' with 100 unique values creates 100 binary features. With only 100 features, the dataset is not high-dimensional enough to cause the curse of dimensionality, which typically requires thousands of features. The poor performance is more likely due to other issues like overfitting or data leakage, not the curse of dimensionality.

Option D is incorrect because the curse of dimensionality is not the most likely cause in this scenario.

Option D is correct because domain knowledge and documentation are the most reliable ways to understand the meaning of zeros. Option A is wrong because statistical methods cannot distinguish missing vs actual zero without context. Option B is wrong because metadata may not have this detail.

Option C is wrong because comparing to other columns might be misleading.

Options A and B are correct. A: Correlation matrix shows pairwise correlations; high values indicate collinearity. B: Variance Inflation Factor (VIF) quantifies how much a feature is explained by others.

C: PCA reduces dimensionality but does not detect collinearity directly; D: t-tests compare means; E: Chi-square tests independence for categorical variables.

Options A and C are correct. Scaling before splitting is a classic source of data leakage. Using target information to filter features is also leakage.

Option B is wrong because proper cross-validation prevents leakage if done correctly. Option D is wrong because time-based split is a valid way to prevent leakage in time series. Option E is wrong because SMOTE should be applied after splitting to avoid leakage.