Advantages and Disadvantages of Classifiers Flashcards
What are 3 advantages of K-nn classifier
- Simple and flexible classifiers
- Good classification performance
- Simple to adapt classifier complexity
4 disadvantages of K-nn classifier
- Need large training sets
- Complete training set has to be stored
- Distances to all training objects need to be calculated
- Features have to be scaled
- Value for k has to be optimized
3 advantages Naive Bayes
- Handles high dimensional feature spaces
- Fast training time
- Can handle continuous and discrete data
1 disadvantage Naive Bayes
- Can not handle correlated features
Sensitivity
TP/(TP+FN), is equal to true positive rate
Specificity
TN/(TN+FP), is equal to true negative rate
True positive rate
TP/(TP+FN), is equal to sensitivity
True negative rate
TN/(TN+FP), is equal to specificity
Precision
TP/(TP+FP)
4 Advantages of Decision Trees
- Interpretable
- Automatic feature selection
- Easy to work with discrete and missing values
- Faste
3 Disadvantages of Decision Trees
- Unstable
- Can not model linear relationsships efficiently
- Greedy
4 Advantages of Random Forests
- Low bias
- Works for different types of data
- Scale invariant
- Few hyperparameters
2 Disadvantages of Random Forests
- Computationally expensive
- Harder to interpret
3 Advantages of MLP
- Flexible model class
- Good emperical performance
- Easily adaptable to different learning settings
4 Disadvantages of MLP
- Computationally expensive
- Lots of hyperparameters
- No convergence to unique optimum
- Hard to interpret
2 Advantages of K-means clustering
- Fast
- Very simple
3 Disadvantages of K-means clustering
- Finds only convex shapes
- Sensitive to initialization
- Can get stuck in local minima
3 Advantages of Hierarchical clustering
- Dendrogram gives overview of all possible clusterings
- Linkage type allows to find clusters of varying shapes
- Different dissimilarity measures can be used
2 Disadvantages of Hierarchical clustering
- Computationally intensive
- Clustering limited to “hierarchical nestings”
2 Advantages of PCA
- Reflects our intuitions about the data
- Dramatic reduction in size of data
2 Disadvantages of PCA
- Too expensive for many applications
- Hard to understand assumptions behind the methods