Use powerful modeling techniques and algorithms, including automated variable selection, virtual variable transformation and rigorous built-in model validation. Both software and workflow were designed from the ground up to provide everything you need to build, access, report and tune predictive models.

Benefit to the Business

Business analysts and statisticians can quickly build, assess, deploy and manage multiple predictive models or scores. These models can easily incorporate mixed data sources for better-informed business decisions.

Key Features

• Model development
  
  
  • Flexible workflow enables rapid definition and building of models
  • Ability to copy and edit existing models or create new ones in minutes
  • Model using mixed data
  • Model using time series data
  • Model using index-of-max data transformations
  • Model using existing clusters or strategies
• Variables
  
  
  • Create, modify and transform variables using built-in expression language
  • Extended editing of grouping variables
  • Automated data mining and variable selection
  • Random, query-based, and set-theory-based operations methods of sampling datasets
• Algorithms:
  
  
  • Non-monotone (can automatically transform variables) and non-linear (can identify complex patterns). These characteristics enable the models to detect patterns that traditional algorithms would miss
  • Designed to get the most out of customer behavior data, including text, standard structured variables (attributes) and patterns or sequences of behavior
  • Includes classic algorithms, such as linear regression and binomial regression
• Reason code generation using score card modeling techniques



• Automatic updating of model performance results and model validation reports every time a model is rebuilt



• Lift charts, deciles and odds charts, delta lift curves, variance scatter plots, variable box plots, partial dependence plots and relative importance measures for key predictors, and more
• Model compliance. View the list of key predictors that were selected by a model, and remove them from the model if appropriate. This includes models using text features
• Random, query-based, and set-theory-based operations methods of sampling datasets
• Export of decision design into an application instruction package. Put models, including any data transformations, into production without coding.