Genetic datamining - model dicovery and goal optimsation

TheGmax is the first complete evolutionary computing tool that empowers computers to model data without having to tell them how to do it.

Able to reduce many thousands of variables to a handful of key drivers in minutes, TheGmax discovers formulae that optimise user goals in a single step.

Say goodbye to regression and neural nets; TheGmax represents a quantum shift in the ability to rapidly mine and model data sets of any size.

Model Discovery

TheGmax uses evolutionary programming to discover models (formulae) that are uniquely adapted to best fit data. Thus unlike conventional methods, which force data to fit predetermined structures, TheGmax is free to explore a truly infinite variety of structural alternatives.

Models so evolved are non-parametric, assumption free, immune to missing data, resistant to noise and have unparalleled predictive power.

Variable Selection

The most powerful auto-emergent property exhibited by TheGmax is its ability to automatically identify the key components and patterns in huge data sets.

Irrespective of whether there are 10 or 10,000 variables, the time required to identify the key drivers is unaffected. The process is very fast, thorough and efficient.

Some people use TheGmax for this feature alone.

Goal Optimisation

The freedoms gained by letting the data dictate its own model structure permit the deliberate targeting of non-parametric and rank-based objectives with the same ease that conventional methods minimise squared error.

Thus for example TheGmax can target directly:

- Areas under ROC curves (binary dependent)

- Cumulative lifts (binary and continuous dependants).

- Multiple class discrimination (categorical dependent).

- Rank based correlation measures.

- Order based measures of entropy.

File Scoring

This is where TheGmax earns its keep by putting the discovered models to work.

The value of any model lies in ability to make predictions and the built-in companion utility, Gscore, is the productivity tool that automatically makes the predictions for you.

Gscore seamlessly applies discovered models to external files and databases using simple point-and-click controls.

It scores the records with their predicted percentiles and their predicted values and offers flexible filtering for both the input and the output.

GOAL OPTIMISATION

MODEL DISCOVERY

VARIABLE SELECTION

DATABASE & FILE SCORING