• Predictive data-mining & goal optimization for Business, Life sciences and Genomics.
  • Reduces high dimensional problems to low dimensional solutions.
  • Simplifying complexity.
  • Say goodbye to regression, you are unlikely to need it again.

How TheGmax works

TheGmax achieves its objectives by imitating the biological operations of genetic inheritance that drive the adaptation and evolution of living things.

The application of such operations to 'computer code sequences' instead of 'genetic code sequences' enables software algorithms (models) to spontaneously self-modify and adapt.

When applied continuously with fixity of purpose to a large population of algorithms, they are encouraged to become supremely adapted to their intended purpose.

In other words, TheGmax enables computers to evolve their own software.

The naturalist, Charles Darwin, named this process "evolution by natural selection." We call it Genetic Modeling.

This approach to the analysis of data exhibits some extraordinary auto-emergent properties that are extremely beneficial and most unlikely to be found in any other modeling paradigm. We list them in the adjacent panel.


The benefits that make TheGmax unique

1. TheGmax can deliberately target non-parametric and rank-based objectives without needing to construct intermediate models of least squares or product moment correlation.

Therefore, the evolved models are uniquely adapted to optimize the real purpose for which they are actually intended.

2. The model structure itself is an 'unknown' and TheGmax is free to explore a truly infinite variety of alternatives.

Consequently, the models tend to need fewer variables because unsuspected relationships and hidden structures are allowed to express themselves.

3. The identification of key variables and the manner of their involvement is automatic, fast and wholly independent of the number of candidate variables.

Thus the age-old "Curse of Dimensionality" is eliminated. It doesn't matter whether there are 10 or 10,000 variables; the time required to locate the key drivers is unaffected.

4. No prior assumptions are required concerning either the variables or the errors-of-prediction. The discovered models are entirely distribution free.

So non-technical people can achieve excellent results with little training.

5. TheGmax is immune to missing data, tolerant of noise, resistant to outliers and impervious to the type and format of the source data.

Hence, there is no requirement to pre-process or clean the source data.

6. The models are rendered in simple arithmetic.

Which means there are no black boxes and the models are understandable.

7. The models are robust and generalise well to unseen data.

Predictions are reliable.