Font size
WorksheetsData Science Fundamentals Worksheet
Total questions: 91
Worksheet time: 46mins
What is the primary goal of data science?
To collect as much data as possible
To use data to gain insights and make decisions
To replace human judgment entirely
To create complex algorithms without application
Which of the following is an example of unstructured data?
A CSV file of customer purchases
A SQL database table
A tweet containing text and emojis
An Excel spreadsheet with sales figures
What does the "Venn Diagram of Data Science" include?
Hacking skills, math/statistics, and domain expertise
Coding, visualization, and cloud computing
Data mining, AI, and robotics
SQL, Python, and R
Which type of data is "zip code" classified as?
Quantitative continuous
Quantitative discrete
Qualitative nominal
Qualitative ordinal
What is the key assumption of the Central Limit Theorem?
Data must be normally distributed
Sample means will approximate normality as sample size increases
Population variance must be known
Samples must be biased
In hypothesis testing, what does a p-value < 0.05 typically indicate?
Accept the null hypothesis
Reject the null hypothesis
Which metric measures "how often a classifier is correct when it predicts positive"?
Recall
F1-score
Precision
Accuracy
What does Occam’s Razor suggest in modeling?
Use the most complex model available
The simplest explanation is usually best
Always prioritize high accuracy over interpretability
Add more features to improve performance
Which distance metric is calculated as √(Σ(x_i - y_i)²)?
Manhattan distance
Euclidean distance
Cosine similarity
Hamming distance
What is the purpose of z-score normalization?
To convert categorical data to numerical
To scale data to have mean=0 and standard deviation=1
To remove outliers permanently
To reduce dimensionality
Which of the following is NOT a requirement for a distance metric?
Positivity (d(x,y) ≥ 0)
Symmetry (d(x,y) = d(y,x))
Triangle inequality
Linear correlation
What does a correlation coefficient of -0.9 indicate?
Strong positive relationship
Weak negative relationship
Strong negative relationship
No relationship
Which plot is best for visualizing relationships between two numerical variables?
Histogram
Box plot
Scatter plot
Pie chart
What is the main weakness of pie charts?
They are hard to read and compare
They only work for small datasets
They require 3D rendering
They cannot show proportions
Which rule states that ~68% of data falls within ±1 standard deviation of the mean?
Bayes’ Theorem
Law of Large Numbers
Empirical Rule
Central Limit Theorem
In a confusion matrix, what does "False Negative" represent?
Correctly predicted negative cases
Incorrectly predicted positive cases
Correctly predicted positive cases
Incorrectly predicted negative cases
The harmonic mean of precision and recall is called:
F1 score
Accuracy
Specificity
ROC curve
Which sampling method ensures every member has an equal chance of selection?
Stratified sampling
Random sampling
Convenience sampling
Snowball sampling
What is the purpose of the "bias-variance tradeoff"?
To maximize model complexity
To balance underfitting and overfitting
To eliminate all errors
To ignore outliers
Which of the following is a categorical variable?
Temperature in Celsius
Height in cm
Blood type (A/B/AB/O)
Income in dollars
What does a high coefficient of variation (CV) indicate?
Low variability relative to the mean
High variability relative to the mean
Perfect correlation
Normal distribution
Which test checks if two categorical variables are independent?
t-test
ANOVA
Chi-square test
Kolmogorov-Smirnov test
What is the primary use of a box plot?
To show trends over time
To visualize the distribution and outliers
To compare proportions
To display correlations
Which of the following is a supervised learning algorithm?
K-means clustering
Logistic regression
Principal Component Analysis (PCA)
Apriori algorithm
What does "imputation" refer to in data cleaning?
Removing missing values
Estimating missing values
Standardizing scales
Encoding categories
What is the primary goal of logistic regression?
To predict continuous outcomes
To classify data into discrete categories
To reduce dimensionality
To cluster unlabeled data
Which activation function is commonly used in logistic regression?
ReLU
Sigmoid
Tanh
Linear
Which of the following is the main purpose of L2 regularization?
Increase model complexity
Reduce overfitting by penalizing large coefficients
Ignore feature correlations
Maximize training accuracy
In K-means clustering, what does "K" represent?
Number of features
Number of clusters
Learning rate
Iteration count
Which evaluation metric is most appropriate for imbalanced datasets?
Accuracy
F1-score
Mean squared error (MSE)
R-squared
What is the key assumption of Naive Bayes classifiers?
Features are normally distributed
Features are conditionally independent
Data must be linearly separable
Labels are ordinal
Which algorithm seeks the "maximum margin" hyperplane in classification?
Decision Trees
Support Vector Machines (SVM)
K-nearest neighbors
Random Forest
What does "bagging" in ensemble methods refer to?
Combining models with weighted votes
Training models on random subsets of data with replacement
Iteratively reweighting misclassified samples
Reducing features via PCA
Which of the following is a disadvantage of decision trees?
High interpretability
Prone to overfitting
Handles only numerical data
Requires feature scaling
What is the purpose of the ROC curve?
To visualize precision-recall tradeoffs
To compare classifiers at different thresholds
To measure clustering compactness
To reduce feature dimensions
Which technique helps address class imbalance?
Removing minority class samples
Oversampling the minority class
Ignoring class weights
Using Euclidean distance exclusively
What does "gradient descent" optimize in linear regression?
Sum of absolute errors
Sum of squared errors
Number of features
Correlation coefficient
Which statement about the "bias-variance tradeoff" is true?
High bias implies overfitting
High variance implies underfitting
Increasing model complexity reduces bias but may increase variance
Simple models always have low variance
The main advantage of non-parametric models is:
They do not assume a specific form for the underlying data distribution.
They require fewer data points than parametric models.
They are always faster to train than parametric models.
They guarantee higher accuracy than parametric models.
Which visualization is best for showing parts of a whole?
Scatter plot
Stacked bar chart
Histogram
Box plot
What does a high "silhouette score" indicate in clustering?
Poor separation between clusters
Well-defined clusters
Overfitting
Random clustering
Which method is used for dimensionality reduction?
LASSO regression
Principal Component Analysis (PCA)
K-means clustering
Logistic regression
What is the key difference between supervised and unsupervised learning?
Supervised learning uses labeled data
Unsupervised learning always has higher accuracy
Supervised learning ignores features
Unsupervised learning requires a test set
Which of the following is a hyperparameter?
Coefficient in linear regression
Learning rate in gradient descent
Mean of the dataset
Output prediction
What problem does "multicollinearity" cause in regression?
Reduces model interpretability
Makes coefficient estimates unstable
Increases training speed
Eliminates outliers
Which statement about "deep learning" is true?
Requires minimal data
Automatically performs feature engineering
Only works with tabular data
Avoids non-linear transformations
What is the purpose of a "validation set"?
To test the final model
To tune hyperparameters
To replace the training set
To remove outliers
Which metric is used to evaluate regression models?
Precision
Recall
Mean Absolute Error (MAE)
F1-score
What does "one-hot encoding" do?
Converts categorical variables to binary vectors
Normalizes numerical data
Reduces dimensionality
Clusters similar features
50. The algorithm that is most sensitive to feature scaling is:
K-Nearest Neighbors
Decision Tree
Naive Bayes
Random Forest
Which of the following is an unsupervised learning algorithm?
Decision Trees
Random Forest
K-means clustering
Naive Bayes
In Bayesian statistics, what does the "prior probability" represent?
The probability after observing data
The initial belief before observing data
The likelihood of the null hypothesis
The error rate of the model
Which distribution is commonly used to model binary outcomes?
Normal distribution
Poisson distribution
Bernoulli distribution
Uniform distribution
What is the key idea behind the "kernel trick" in SVM?
It reduces computational complexity
It transforms data into higher dimensions without explicit calculation
It eliminates the need for feature scaling
It works only with linear decision boundaries
Which of the following is NOT a loss function for regression?
Mean Squared Error (MSE)
Cross-entropy
Mean Absolute Error (MAE)
Huber loss
What does "entropy" measure in decision trees?
The purity of a node
The depth of the tree
Which statement about "AdaBoost" is true?
It gives equal weight to all classifiers
It focuses on misclassified samples in subsequent iterations
It requires features to be uncorrelated
It only works with deep neural networks
What is the primary use of the Kolmogorov-Smirnov test?
To compare two distributions
To evaluate classifier accuracy
To reduce dimensionality
To impute missing values
Which probability rule states P(A|B) = P(B|A)P(A)/P(B)?
Law of Total Probability
Bayes’ Theorem
Central Limit Theorem
Empirical Rule
What is the main challenge of "curse of dimensionality"?
Increased computational cost and sparsity of data
Decreased model interpretability
Overfitting due to small sample sizes
All of the above
Which technique is used to handle overfitting in decision trees?
Pruning
Increasing tree depth
Removing features
Using larger datasets only
In hypothesis testing, what is a Type I error?
Failing to reject a false null hypothesis
Rejecting a true null hypothesis
Accepting the alternative hypothesis
Incorrectly calculating the p-value
Which of the following is a non-parametric test?
t-test
ANOVA
Wilcoxon rank-sum test
Chi-square test
What does "AUC-ROC" measure?
The area under the precision-recall curve
The model’s ability to distinguish between classes
The variance of the residuals
The clustering compactness
Which statement about "feature engineering" is true?
It is unnecessary for deep learning models
It involves creating new features from raw data
It only applies to numerical data
It reduces model accuracy
What is the purpose of "stratified sampling"?
To ensure each subgroup is proportionally represented
To randomly shuffle data
To remove outliers
To reduce feature dimensions
Which of the following is a disadvantage of KNN?
High computational cost for large datasets
Assumes linear separability
What does "multiclass classification" involve?
Predicting more than two classes
Using multiple binary classifiers
Both A and B
None of the above
Which distribution describes the number of events in fixed intervals?
Normal distribution
Binomial distribution
Poisson distribution
Exponential distribution
What is the key advantage of "random forests" over single decision trees?
Reduced overfitting through ensemble averaging
Faster training time
Lower memory usage
No hyperparameters to tune
Which metric is most sensitive to outliers in regression?
Mean Absolute Error (MAE)
Mean Squared Error (MSE)
R-squared
F1-score
What is the primary goal of "dimensionality reduction"?
To remove all categorical features
To retain important information while reducing features
To increase model complexity
To eliminate the need for training data
Which statement about "cross-validation" is true?
It reduces the need for a test set
It provides a more robust estimate of model performance
It is only used for clustering
It requires labeled data in unsupervised learning
What does "TF-IDF" stand for in text processing?
Term Frequency-Inverse Document Frequency
Text Feature-Indexed Data Format
Time-Frequency Inverse Distribution
Tokenized Frequency-Integrated Data Frame
Which algorithm is used for anomaly detection?
Linear Regression
K-means
Isolation Forest
Logistic Regression
What is the main purpose of "data augmentation"?
To delete redundant data
To artificially increase training data diversity
To reduce dataset size
To replace missing values
What does a z-score of 2.0 indicate?
The value is 2 standard deviations below the mean
The value is 2 standard deviations above the mean
The value is in the bottom 5% of the distribution
The value is exactly at the mean
Which R function correctly calculates the mean of a vector x?
median(x)
mean(x)
The Naive Bayes classifier assumes:
Features are normally distributed
Features are conditionally independent given the class
All features are equally important
The data has no missing values
In hypothesis testing, what does the p-value represent?
The probability that the null hypothesis is true
The probability of observing the data if the null hypothesis is true
The effect size of the alternative hypothesis
The confidence level
Which R code would you use to remove rows with missing values from a dataframe df?
df[complete.cases(df), ]
df %>% drop_na()
na.omit(df)
All of the above
The Central Limit Theorem states that:
Sample means are normally distributed if n ≥ 30
Population distributions must be normal
Variance decreases with larger samples
Confidence intervals always contain the true mean
Which metric is most appropriate for evaluating an imbalanced binary classifier?
Accuracy
F1-score
R-squared
Euclidean distance
What does the empirical rule state about normally distributed data?
68% falls within ±1 SD, 95% within ±2 SD
The mean equals the median
Skewness must be zero
It only applies to large samples
In decision trees, how is the optimal split chosen?
By maximizing information gain
By minimizing Gini impurity
Both A and B
By random selection
Which R package is used for creating heatmaps?
ggplot2
dplyr
heatmap() (base R)
Both A and C
The Kolmogorov-Smirnov test compares:
Two distributions
Means of paired samples
Variance across groups
Correlation coefficients
What does mutate() do in R (dplyr)?
Adds new variables to a dataframe
Filters rows
Sorts columns
Merges tables
Which of the following formulas is correct?
TP / (TP + FP)
TP / (TP + FN)
(TP + TN) / total
FP / (FP + TN)
Which statement about SVM is true?
It only works with linear decision boundaries
It maximizes the margin between classes
It ignores outliers
It requires features to be uncorrelated
What does this R code do?
Filters high salaries
Calculates average salary by department
Creates a scatter plot
Removes missing values
