The module provides two additional workflows: the Create Predictor workflow guides the user through the creation and validation of linear and non-linear prediction models, the Apply Predictor workflow through the definition and application of scoring models to new data and through the evaluation of scoring results in comparison to control groups.
Main functions and features
- Linear multivariate regression models with detailed model statistics
- Local regressions over self-organizing map (SOM) “receptive fields” to resolve non-linearities
- Visualization of local model statistics and non-linearity diagnostics
- Administration of data partitioning into training and test sets
- Graphical validation and comparison of created models
- Interactive definition of score groups and objective functions
- Application of scoring models to new data and export of scoring results
- Visualization and evaluation of application results using score charts and gains charts
The patented Local SOM Prediction technology combines the non-linear data representation of the SOM with linear regression “white-box” models, designed for explaining residual variance in the global linear model. Read more about this technology here.