DataCaster NetCaster - by John Sweeney
DataCaster and NetCaster help you build multilayer feedforward neural
networks. DataCaster's statistical abilities should be of particular interest to
T he demand for neural network products shows no signs of abating. For those interested in applying neural
network technology to their own trading, some basic questions about any neural net software must be answered:
How difficult is it to get the data into the product? How much preprocessing must be done? What, if any, technical
indicators, time series transformations, statistical functions and graphics capabilities are available to support analysis?
And that's just to get ready for preliminary networking!
Enter DataCaster and NetCaster from Maui Analysis and Synthesis Technologies (MAST). These are companion
programs that allow you to design training files for neural networks, preprocess data, train the neural network and
use the neural net to make forecasts. Preprocessing data means analyzing and mathematically transforming your data
to make variables useful and relevant. Using a trial and error principle, the user may build several neural networks,
varying the inputs, hidden node and activation function to obtain the best predictor. DataCaster, the preprocessor to
NetCaster, can be run from NetCaster or on its own, while NetCaster builds and trains a multilayer feedforwardÝ