Clicky

Rulex Algorithms

Rulex provides a rich selection of proprietary and enhanced standard algorithms for both rules-based and math-based predictive analytics. Rulex supports these powerful algorithms with intuitive, graphical tools and highly automated modeling functions to enable business/process experts and citizen data scientists to easily make accurate, fast predictions for augmented, automated, and autonomous decisions.

  • Logic Learning Machine
    • Flagship proprietary training and modeling algorithm for producing conditional logic-base predictive models.
  • Decision Tree
    • Proprietary construction algorithm that improves the C4.5 approach to achieve better rule models.
  • Uniclassifier®
    • Revolutionary proprietary one-class training algorithm for unsupervised anomaly detection and rules discovery.
  • K-Nearest Neighbor
    • Proprietary enhancements enable treatment of both ordered and categorical variables through a suitable balance between distances.
  • Neural Networks
    • Proprietary optimizations enable training that achieves better models in less time.
  • Logic Learning Machine
    • Flagship proprietary training and modeling algorithm for producing conditional logic-based predictive models.
  • Piecewise Linear
    • Proprietary training algorithm that reconstructs a piecewise linear model from data, where different behaviors in the domain space are triggered by the value of categorical variables.
  • Auto Regressive
    • Proprietary preprocessing and postprocessing to identify linear models, especially for period identification. Considered Croston models are not standard and allow a better forecast of spiking time series, such as demand series.
  • K-Nearest Neighbor
    • Proprietary enhancements enable treatment of both ordered and categorical variables through a suitable balance between distances.
  • Neural Networks
    • Proprietary optimizations enable training that achieves better models in less time.
  • Label Clustering
    • Proprietary algorithm that adopts the modified K-means technique of Standard Clustering for grouping patterns labeled by categorical variables. It can be used for predicting a vector of values instead of single ones, as in Regression models.
  • Standard Clustering
    • Proprietary enhancements enable treatment of both ordered and categorical variables through a suitable balance between distances.
  • Projection Clustering
    • Proprietary algorithm that adopts a modified K-means technique for grouping patterns to allow projection on selected variables and produce a clustering in each of them.
  • Rule Optimization
    • Proprietary algorithm that optimizes rule-based models built with Logic Learning Machine or Decision Tree algorithms, while respecting constraints chosen by the user.
  • Discretize
    • Proprietary ADID (Attribute Driven Incremental Discretization) algorithm for a near-optimal discretization of ordered variables in a classification problem.
  • Join
    • Proprietary implementation of non-standard join functions for processing key relationships. In addition to less than, greater than, and different from operators, it handles relationships between strings, phonetic and soft operators for faster text processing.
  • Moving Window
    • Proprietary algorithm enables static models like Classification and Regression to analyze and forecast time series.
  • Reshape To Long
    • Very useful non-standard task for unrolling data according to designated variables.
  • Reshape To Wide
    • Very useful non-standard task for performing the opposite operation of Reshape To Long.