For some predictive modeling functions, it is very important to normalize the numeric input variables; that is, to rescale them so they have a similar mean and standard deviation. If you do not normalize input variables, the effect of variables with a large magnitude or a large standard deviation may dominate the model and reduce the accuracy of its predictions.
Normalize input variables before calling the following functions:
- Canopy
- KMeans Functions
- KMeans
- KMeansPredict
- KNN
- Generalized Linear Model (GLM) Functions
- GLM
- GLMPredict_MLE
- GLML1L2
- GLML1L2Predict
- Least Angle Regression (LAR) Functions
- LAR
- LARPredict
- Linear Regression Functions
- Linear Regression
- LinRegPredict
- Principal Component Analysis (PCA) Functions
- PCA
- PCAScore
- Support Vector Machine (SVM) Functions
- SVMSparse
- SVMSparsePredict_MLE
- SVMSparseSummary
- SVMDense
- SVMDensePredict
- SVMDenseSummary
The MLE Scale functions are designed to make normalization easy. For an example of using Scale functions to normalize input variables, see PCA Example.