|  NOV 1989

Neural networks: A trading perspective by Carol H. Halquist & George F. Schmoll, II

Neural networks: A trading perspective by Carol H. Halquist & George F. Schmoll, III Neural networks and neural network simulators are used in signal processing, expert systems, modeling and forecasting. They may be used to solve complex, well-defined problems which may be impossible to solve algorithmically. They are primarily beneficial when the relation between input data and output data is not clearly understood, but a large number of sample sets of data are available. Neural networks are an alternative method used in signal processing, particularly in such fields as speech, vision and pattern recognition or noise filtering including stock or commodity market data. They may be used to ascertain the level of seemingly random or uncorrelated noise in data and, thus, produce a filtering mechanism to help clarify underlying relationships contained in the data. Neural networks have been used as self-optimizing expert system generators for models that are difficult or supposedly impossible to define with standard rule-based expert systems. Rule-based expert systems require a knowledgeable expert to define the game rules. It may be possible to poll or measure a number of expert traders to generate an expert system to pick and trade the markets in unison with the best of the experts. Neural networks have been used to model and forecast financial and economic time series data such as interest rates, stocks, options, mutual funds and commodity futures prices as well as cash markets (Figure 1). These forecasts have proven comparable with other methods currently in use to predict likely prices and position stop-loss buy/sell prices.

by Carol H. Halquist & George F. Schmoll, II