One of those rare products that changes the way you think

"DataModeler is one of those rare products that changes the way you think.  It ends any excuse for extending an assumption of linearity in modeling beyond the domains in which it is truly appropriate.  Using brilliant and flexible design that is itself the result of years of evolution, DataModeler harnesses natural selection to hunt down from the infinite computational universe and with frightening speed and efficiency ensembles of parsimonious non-linear models that predict well. 

I have used it with success to explore optimal insurance contracts, issues in legal education and a host of other difficult problems.  Often producing results that are superior to those from neural networks and that end up more amenable to traditional statistical analysis, DataModeler never fails to astound me with its robustness, its capabilities, and its ever-growing network of documentation."

Prof. Seth J. Chandler, Foundation Professor of Law, Director of the Program on Law and Computation, University of Houston Law Center, U.S.A.

Much higher fidelity in matching predictions to data

"I have used DataModeler from Evolved Analytics in my work as a chemical engineer.  I have found that compared to other genetic programming packages, DataModeler can give me much higher fidelity in matching predictions to data.  DataModeler’s use of ensembles of models rather than a single-best-guess model offers a significant improvement in reducing residuals and improving robustness.  

On one problem, in a four input-variable reactor model, two other packages gave average residual error of about 30% while DataModeler was able to reduce this to about 4%.  In addition, DataModeler can be used to produce trust metrics; functionality unique to DataModeler compared to the other programs I have used."

Peter Kip Mercure, PhD Chemical Engineering, The Dow Chemical Company, U.S.A.

Using DataModeler with Excellent Results

"We are using Data Modeler with excellent results in both our student projects and industrial applications for several years. It has access to the powerful symbolic and numerical calculation tools and the nearly endless visualization opportunities of Mathematica. Data Modeler substantially benefits from being developed in an industrial setting to address real-world modeling problems. It is a robust tool for large and noisy data sets.

Our main projects are related to real-world industrial optimization problems with the focus on data exploration, time-series analysis, inferential sensors development, and financial data analysis. Data Modeler was used in real-world case studies such as “Collection, Preparation, and Analysis of Data from Power Suppliers” to develop a surrogate model for emission prediction in steam generators."

Prof. Dr. Thomas Bartz-Beielstein, Head of CIOP Research Center, Cologne University of Applied Sciences, Cologne, Germany

Incredible out-of-the-box performance

"With DataModeler we were able to model a data set with 32 attributes and over 10,000 rows in less than an hour. The ensemble it produced was far more accurate than anything else we've seen. This is incredible out-of-the-box performance."

Dr. Conor Ryan, Director at the Biocomputing and Developmental Systems Group in the Computer Science and Information Systems Department at the University of Limerick, Ireland.

DataModeler enabled the data to talk to us quickly

"We put DataModeler in the loop. It enabled the data to talk to us quickly. Without delay, we could translate the insights to our client's perspective and thoughtfully consider how to revise, refine and immediately iterate.

We wanted to check some implicit assumptions about our models that we thought might be wrong. With DataModeler we knew the structure of the models would be driven by the data, not by what we wanted to find. Identifying the 'real truth', not what you hope to find, is so important"

Una-May O'Reilly, PhD, Leader of the Evolutionary Design and Optimization Group at Computer Science and Artificial Intelligence Lab, MIT, U.S.A