Your task is to detect fraudulent activities in financial transactions. What would be the considerations in choosing between AI, Machine Learning, or Deep Learning for this task?
- AI, for its expert systems
- Deep Learning, for its complex pattern recognition
- Machine Learning, for its ability to learn from historical data
- nan
Machine Learning can be trained on historical data to detect patterns indicative of fraudulent activities, making it a suitable choice for this task.
Your Logistic Regression model is suffering from separation, causing some estimated Odds Ratios to be extremely large. How could you handle this issue?
- By adding more variables
- By applying regularization techniques
- By increasing the size of the dataset
- By removing all predictors
Separation in Logistic Regression can lead to overly large coefficient estimates. Applying regularization techniques, such as Ridge or Lasso, can help in constraining the coefficient estimates and mitigate this issue.
In which type of Machine Learning does the model learn from labeled data?
- Reinforcement Learning
- Semi-Supervised Learning
- Supervised Learning
- Unsupervised Learning
Supervised Learning involves training a model on a labeled dataset, meaning the desired output for each input is known. The model learns to predict the output from the input data.
In what situations would ElasticNet be preferred over Ridge or Lasso?
- When all features are equally important
- When features are uncorrelated
- When model complexity is not a concern
- When multicollinearity is high
ElasticNet is preferred when there's multicollinearity and you want to balance between Ridge and Lasso, as it combines the properties of both.
The ________ component in PCA explains the highest amount of variance within the data.
- first
- last
- median
- random
The "first" principal component in PCA explains the highest amount of variance within the data. It is aligned with the direction of the maximum spread of the data and forms the most substantial part of the dataset's structure.
What is classification and how does it differ from regression?
- Predicting a category, differs by number of variables
- Predicting a category, differs by output type
- Predicting a number, differs by algorithm
- Predicting a number, differs by input type
Classification aims to predict a categorical outcome, such as 'yes' or 'no', whereas regression predicts a continuous numerical value, such as a price or weight. While both are predictive modeling techniques, the key difference is in the type of output they produce. This makes classification suitable for discrete decisions, while regression is used for forecasting continuous quantities.
What is the main principle behind the K-Nearest Neighbors algorithm?
- Calculating correlations
- Finding nearest points
- Grouping similar objects
- Minimizing error
The main principle of KNN is to classify a new object by assigning it to the most common class among its K nearest neighbors.
In a situation where you have a large dataset with only a small portion of labeled data, which learning paradigm would be most appropriate and why?
- Reinforcement Learning
- Semi-Supervised Learning
- Supervised Learning
- Unsupervised Learning
Semi-Supervised Learning combines both labeled and unlabeled data, making it appropriate for scenarios with limited labeled data.
Bagging stands for Bootstrap __________, which involves creating subsets of the original dataset and training individual models on them.
- Adjustment
- Aggregation
- Algorithm
- Alignment
Bagging, or Bootstrap Aggregation, involves creating subsets of the original dataset through bootstrapping and training individual models on these subsets, which are then combined to make the final prediction.
How can you test for multicollinearity in Simple Linear Regression, and why is it important?
- By Checking Accuracy, Improves Prediction
- By Checking Residuals, Reduces Overfitting
- By Checking Variance Inflation Factor (VIF), Prevents Unstable Estimates
- By Examining Correlations between Variables, Prevents Confounding Effects
Multicollinearity can be detected by checking the Variance Inflation Factor (VIF). It is important as multicollinearity can lead to unstable estimates and make it difficult to interpret individual coefficients.