Automated input and model selection since 1995
Your data. The right inputs. The right model.
The NeuroGenetic Optimizer automatically searches candidate inputs, model types, and internal structures to build selected predictive model committees. No programming or neural-network design expertise required.
The model-design problem
A model is easy. The right model is hard.
Training one neural network does not establish that the useful inputs were selected, that the right model type was chosen, or that better internal structures were considered.
The NGO was created to automate those decisions for people who understand their data and their problem, not necessarily neural-network design.
Point-and-click workflow
From historical data to new predictions.
The NGO gives domain experts a complete predictive-modeling process without requiring programming or manual neural-network construction.
Load your data
Use delimited text files or Excel worksheets with labeled columns.
Prepare and inspect
Chart, transform, align, filter, and prepare variables with included tools.
Define the question
Select the target and mark candidate, required, or ignored inputs.
Let NGO search
Build and evaluate alternative input combinations, model types, and structures.
Predict and understand
Review selected models, sensitivity, validation, response surfaces, and new predictions.
What the NGO delivers
More than one fitted model.
The NeuroGenetic Optimizer provides a complete environment for preparing data, constructing alternatives, selecting strong models, understanding input sensitivity, validating performance, and producing predictions.
Input selection
Optional inputs can be included or excluded during the search so weak variables do not have to remain in every model.
Model selection
Searches model types and internal structures rather than requiring the user to specify one architecture in advance.
Model committees
Retains selected physical models and combines their outputs into a system model.
Multiple validation stages
Uses separate modeling, optimization, and selection data, with an additional holdback option for external validation.
Input sensitivity
Shows which inputs have the greatest effect on the model output and where further investigation may be valuable.
Time-series modeling
Includes lead-lag analysis, temporal model options, preprocessing functions, and tools for sequential data.
Visualization and reporting
Provides predicted-versus-actual charts, error analysis, response surfaces, performance measures, and Word reports.
New predictions
Apply the saved modeling system to new data and export predictions from the system or selected physical models.
Broad predictive applications
One modeling system. Many kinds of questions.
The NeuroGenetic Optimizer has been sold and applied across industries because the underlying job is universal: predict an outcome from candidate data when the useful inputs and model design are not known in advance.
Insurance
Model claims, risk, response, or other outcomes influenced by many interacting variables.
Marketing
Predict customer behavior and investigate which characteristics influence the modeled response.
Finance
Model nonlinear relationships and time-dependent outcomes using historical data.
Entertainment and hospitality
Predict attendance, demand, revenue, or other operational and commercial outcomes.
Manufacturing
Predict quality and product properties or identify variables associated with undesirable performance.
Your application
Choose a measurable target and provide candidate inputs. NeuroGenetic handles the model search.
“As an evaluator of neural networks in R, I still find NGO far superior in all its capabilities. I absolutely love it.”
Brian K. Lee, MBA, PRM, CFA
Build your model
You know the question and the data. Let NeuroGenetic search for the model.
Request a trial for a serious business, government, or academic evaluation, or purchase NGO 3.0 directly.
Trial requests are reviewed to protect the software and support serious evaluations.