WorksheetsQuiz 15 Dec
Total questions: 15
Worksheet time: 8mins
Which statistical measure is NOT typically displayed by the df.describe() function in Pandas?
Mean
Standard Deviation
Non-Null Values
75th Percentile
A Box Plot is the most suitable visualization for exploring the relationship between two continuous numerical variables.
True
False
One-Hot Encoding is a preprocessing technique used to handle continuous numerical data.
True
False
If the first two Principal Components (PC1 and PC2) in your PCA output capture a Cumulative Explained Variance (CEV) of 95%, what product design insight does this provide?
You must retain all original features.
The KNN model will definitely have 95% accuracy.
You can safely reduce the dataset from N features to 2 features.
The data is highly correlated.
Feature A has a correlation of 0.85 with the target variable, and 0.95 with Feature B. If you must reduce the dataset by one feature, which one is the least detrimental to remove?
Feature A, because its correlation with the target is lower.
Feature B, because its correlation with the target is the lowest of the two.
Neither, because the high correlation with the target means both must be kept.
The Manhattan Distance is generally preferred over the Euclidean Distance when feature differences should be penalized more severely.
True
False
What is the critical operational difference between a Correlation Filter node and a PCA node?
The Correlation Filter is non-parametric, while PCA is a classification algorithm.
The Correlation Filter performs Feature Selection (removes columns); PCA performs Feature Extraction (creates new columns).
None of the above
You find a high correlation between 'Feature X Usage' and 'Customer Retention'. What is the most immediate actionable product design insight?
Remove Feature X since it is redundant.
Focus marketing efforts on acquiring new customers.
Promote use of Feature X.
The data is heavily skewed.
Using KNN to predict whether a new customer belongs to the pre-defined 'High-Churn-Risk' group is an example of applying a machine learning model for (a) and targeted product intervention.
Adding too many engineered features may cause:
Better Accuracy
Overfitting
Faster execution
Better PCA
You have a feature, 'Session Duration', which is heavily right-skewed (many short sessions, few very long). Which scaling method is mathematically least likely to distort the data relationships before applying KNN?
Z score
Min max
Log Transformation followed by Standardization
None of the above
What is the primary conceptual effect of the 'Curse of Dimensionality' on the K-Nearest Neighbors (KNN) algorithm in very high-dimensional feature spaces?
It forces the use of the Manhattan Distance metric instead of Euclidean Distance.
It causes the model to suffer from high bias and low variance.
The distance between the nearest neighbor and the farthest neighbor becomes nearly the same.
It only impacts classification tasks, not regression tasks.
You need to select the best 5 non-redundant features for a simple linear model. Which combination of criteria provides the highest quality feature set?
Low variance (near-constant values) AND high correlation with the target.
Low correlation with the target variable AND low correlation with each other.
High correlation with the target variable AND high correlation with each other.
High correlation with the target variable AND low correlation with each other.
Binning a continuous target variable (Revenue) into a categorical variable ('Low', 'Medium', 'High') provides the greatest benefit in terms of:
Reducing the total number of features used by the model.
Preventing the Curse of Dimensionality in the feature space.
Increased model performance by eliminating the need for scaling.
Enhanced interpretability and actionability for product and marketing teams.
A product feature dataset involves counts of discreet events (e.g., Clicks, Scrolls). The data naturally follows a grid-like structure. Which distance metric is theoretically superior for KNN in this context?
Minkowski Distance (p=1.5)
Chebyshev Distance
Euclidean Distance
Manhattan Distance
