Preprocessing Data For Neural Networks by Lou Mendelsohn
Today's global markets demand new analytical tools for survival and profit as prevailing methods of analysis lose their luster. Here, Stocks & Commodities contributor Lou Mendelsohn explains how an emerging method of synergistic market analysis can be applied to neural networks for financial forecasting and discusses how to select and combine various types of market information and transform the data into a format appropriate for neural network training.
With the rise of artificial intelligence technology and the growing interrelated markets of the 1990s
offering unprecedented trading opportunities, technical analysis simply based on single market historical
testing is no longer enough. To meet the trading challenge in today's global markets, technical analysis
must be redefined. I propose a multidimensional method of analysis known as synergistic market
analysis, which utilizes artificial intelligence technologies, including neural networks, to synthesize
technical, fundamental and intermarket data. Synergistic analysis can quantify and discern underlying
relationships and patterns between related markets, capturing information that reflects global market
dynamics, which can markedly improve trading performance.
Previously, I discussed the selection of neural network paradigms and architectures for synergistic
trading. This time, I will explore several issues related to input data selection and preprocessing. Because
available technical, intermarket and fundamental market data is extensive and useful preprocessing
methods quite extensive, here are some ways to handle input data effectively and efficiently in
developing neural networks.