Help
Options
Didn't know it?
click below
click below
Knew it?
click below
click below
Don't Know
Remaining cards (0)
Know
retry
shuffle
restart
0:00
Embed Code - If you would like this activity on your web page, copy the script below and paste it into your web page.
Normal Size Small Size show me how
Normal Size Small Size show me how
MIS 665 Final2SS
Terms from Summary Sheets
| Term | Definition |
|---|---|
| _More_ or _Less_ features with similar R^2 "wins?" | Less |
| What do we fit data on? | Train data (NEVER test) |
| What do we check before modeling to catch multicollinearity | Correlation Heatmap |
| Is a lower or higher MSE better? | Lower |
| Higher R^2 means | more variance explained |
| StnadardScaler changes ___, NOT ___ | coef scale, model accuracy |
| Parsimony | fewer features with similar R^2 is preferred |
| Lasso (alpha=1) | automatic feature selection via L1 penalty |
| DATA LEAKAGE | never fit_transform on full X --> Split first! |
| KNN requires ___ | StandardScaler (uses distance) |
| Logistic Regression gives ____, not just class labels | probabilities |
| SelectKBest: fit on training data only | leakage rule |
| 0.5 is __ guess, 1.0 is ___ guess | random, perfect |
| Use _____ for reliable accuracy | cross_val_score(cv=10) |
| what avoids dummy trap in OneHotEncoder | drop='first' |
| ALWAYS do what before clustering | standardize |
| ______ features dominate distance | high-variance |
| Clustering is supervised/unsupervised | unsupervised (NO Y) |
| Silhouette score of ___+ is strong, ____ is reasonable | 0.71+, 0.51-0.70 |
| K-Means++ init reduces | sensitivity to random start |
| Profile clusters on _________ data for meaning | original (unscaled) |
| PCA is ____-based, MUST scale first | variance |
| Fit PCA on ___ data only, transform both separately | training |
| n_components=0/90 | auto-selects fewest PCs for 90% variance |
| PCA replaces feature names | use loadings to interpret |
| Embeddings understand ___ | synonyms |
| Pipeline: | represent text > reduce dims > build classifier |
| A silhouette score of 0.50 generally indicates | moderately well-separated clusters |
| In healthcare, which metric is most important? | Recall (measure of how many actual positives were caught) |
Created by:
lexi.welte