Today I'm going to introduce two new algorithms of IBM SPSS Modeler 17.1: GLE and Linear-AS
Generalized Linear Engine (GLE)
GLE provides a variety of statistical models such as linear regression for normally distributed responses, logistic models for binary data, log linear models for count data any many more through its very general model formulation.
GLE models are good when you need to fit a well-defined parameterized model to your data and need to find those parameters on a large data set without over-fitting
GLE supports:
- Binary, Categorical and Numeric targets
- Automated model selection based on your target field’s distribution, finds distribution & link function
- Automated field selection and overfit prevention (L1 (LASSO), L2 (Ridge Regression) & L1 + L2 (Elastic Net)
- PMML generation
- and is scoreable via the database Scoring Adapters
Linear-AS
The Linear-AS Node can be used to build Linear Regression models – ie when you need to explore the relationship between one continuous target variable and a list of continuous (covariate) and categorical (factor) predictors.
Note: Linear can also be used in conjunction with Analytic Server but only through PSM (pass, stream, merge) to create multi-threaded split or averaged ensemble models. Linear-AS truly parallelises the building of a single model
Linear-AS supports:
- Automatic detection of 2-way interaction terms
- Numeric Target only
- PMML generation
- and is scoreable via the database Scoring Adapters
#Algorithms#SPSSModeler#WatsonStudio