Model
1. Linear Regression
2. Logistic Regression
3. SVM
4. QDA (LDA ~ Linear Regression)
5. Random Forest
6. Neural Network
7. Naive Bayes
Feature Engineering
1. Include polynomial term: X^2
2. Interactive Term: X1*X2
3. Sqrt(abs(X))*sign(X)
4. Log(abs(X)+1)*sign(X)
Regularization
1. L1; L1 then OLS on chosen x's
2. L2
3. L1 + L2
4. Adaptive L1
5. Forward Stepwise / Backward Stepwise
6. Preselect with OLS then L1
Model Selection
1. CV,
2. AIC, AICc
3. BIC,
4. Cp
Performance Measure
1. AUC
2. Accuracy
3. F1 Score, Mean Average Precision, Cohen's Kappa
4. LogLoss / Deviance
5. Mutual Information, KL convergence
6. Cluster Correlation as defined in the paper
7. Own method of double centering the joint count table (for binary vs. binary)
Algorithm
1. Newton
1.1. Quasi-Newton
1.2. BFGS
1.3. LBFGS
2. Gradient Descent
2.1. Stochastic Gradient Descent
2.2. Coordinate Descent
2.3. Conjugate Descent
3. Least Angle Regression
No comments:
Post a Comment