Ward Systems Group, Inc.

Ward Systems Group, Inc. was established in 1988 to develop and sell analytical software that utilizes artificial intelligence (AI) for science, business, medicine, education, and financial trading. Their advanced neural network and genetic algorithm software has given thousands of customers a competitive edge in solving real problems using prediction, classification, modeling, estimation and optimization techniques. Consulting assistance is given to every customer in both using the programs and understanding the strategies that ensure effective use of the software.

Ward Systems Group develops and sells a unique program called ChaosHunter that is designed to produce readable formulas to model numeric data using either Evolution Strategy or Swarm optimization techniques. ChaosHunter is capable of producing buy/sell signals for trading, predicting a future value of a time series, classifying data into different categories, or discovering a formula based upon a set of inputs.

Ward Systems Group, Inc. markets a technical analysis program, NeuroShell Trader, which has continuously won first place for the past 11 years in the Technical Analysis of Stock’s and Commodities competition for artificial intelligence software. The proprietary neural network in NeuroShell Trader trains itself on the data, creates its own rules, and generates trading signals. The genetic algorithm optimizers select the best trading rules, neural network predictions, and indicator parameter values based on the individual instruments being modeled. NeuroShell Trader connects to a variety of datafeed and trading partners that includes coverage of a broad range of international markets, as well as major US stock indexes, futures, and FOREX.

The original NeuroShell software aimed to make backpropagation neural network technology available to problem solvers in a wide variety of fields even if the user did not have a mathematics background. NeuroShell 2 software evolved into a program with 12 different backpropagation networks, as well as PNN, GRNN, GMBH and Kohonen networks for solving different types of problems, all of which required experience based tweaking to get the model right. Working with our customers we learned that they had to run many different models in order to find a solution for their problem, so our research efforts continued to find faster algorithms that required less parameter settings. The result was a neural network architecture called TurboProp 2™, a proprietary algorithm that is very fast, very accurate, and has absolutely no learning parameters to worry about setting correctly. TurboProp 2 generalizes very well on new data. It is ready to help you solve your problems in minutes. The second type of neural network used in the AI Trilogy family of products, which we call the “genetic method,” uses neural networks developed by Dr. Don Specht. In the NeuroShell Predictor, we use Specht’s General Regression Neural Nets (GRNN). In the NeuroShell Classifier we use Specht’s Probabilistic Neural Nets (PNN). We pioneered the use of genetic algorithms to train GRNN and PNN nets so that they are excellent at finding the contribution of each variable in the training data, and eliminating worthless variables.