25303

9. = Model is too simple and fails to capture underlying patterns
11. = Techniques to make model decisions understandable to humans
13. = Ensemble of decision trees that reduces overfitting
17. = Technique that transforms features into orthogonal components
19. = Harmonic mean of precision and recall
20. = Proportion of true positives correctly identified from actual positives
21. = Sequential ensemble method that builds models to correct previous errors
24. = Penalizing model complexity to prevent overfitting
25. = Model learns noise and performs poorly on new data
26. = Method to estimate model performance by splitting data into folds
28. = Visual that shows cumulative gain from model-based targeting
29. = Loss function that penalizes false classifications with probability estimates

1. = Choosing the most relevant variables for model building
2. = Classifier that finds the best boundary between classes
3. = Tracking model performance over time after deployment
4. = Analyzing relationships and networks between entities
5. = Regularization that shrinks coefficients but doesn’t zero them out
6. = Table showing true vs predicted classification outcomes
7. = Practices for deploying, monitoring and maintaining machine learning in production
8. = Measure showing improvement of model versus random selection
10. = Combining multiple models to improve prediction accuracy
11. = Initial investigations to discover patterns, spot anomalies and test hypotheses
12. = Local approximation method for explaining black-box models
14. = Reducing the number of input variables while retaining information
15. = Regularization technique that can set some coefficients to zero
16. = Dividing data into training and testing subsets
18. = Process of releasing a model for real-world use
22. = Game-theoretic method to explain individual predictions
23. = Proportion of true positives among predicted positives
27. = Metric measuring classifier’s ability across all thresholds