Using Artificial Intelligence to Improve Real-World Performance

Book Review
By: Lou Mendelsohn

Neural networks are one of the most innovative analytical tools to surface in the financial arena. Editors Robert Trippi and Efraim Turban have put together a collection of recent articles from industry and academic experts dealing with the application of artificial intelligence to real-world finance and investing. This work, which begins with an introductory section on artificial neural networks, examines the application to areas including analysis of financial conditions, business failure prediction, debt-risk assessment, security market analysis and financial forecasting. Neural Networks is an excellent survey for those new to neural network applications in the finance industry. It also provides those readers familiar with neural networks or finance with a springboard into a variety of topics. Traders will be particularly interested in Neural Networks' Part 5 and Part 6, titled "Security market applications" and "Neural network approaches to financial forecasting," respectively. Although many chapters are not specifically concerned with the stock and futures markets, much of the material is relevant in a more general context. Throughout the book, the authors discuss various issues that are common across neural network applications including the selection of inputs, methods of preprocessing, choice of architecture and result interpretation.

NEURAL NET OVERVIEW AND ANALYSIS
Part 1, titled "Neural network overview," contains three chapters that cover the fundamentals of artificial neural networks. The first chapter examines the basic concept behind neural networks and contains a solid overview of the structure, inner workings and application of neural networks. The second chapter reviews neural networks, then briefly discusses applications of this technology for corporate finance, financial institutions and professional investors. The third chapter concerns itself with the application of neural networks to the task of credit approval. Part 2, "Analysis of financial conditions," has four chapters involving the application of neural networks to a variety of financial situations. In the latter two, neural networks are used exclusively in two financial analysis tasks. R. Berry and Duarte Trigueiros describe a system that uses neural networks to extract knowledge from accounting reports, while George Klemic examines the use of neural networks to help the IRS identify those debtors most likely to become delinquent. This chapter describes the methodology used by the IRS to create neural computing systems and provides a cost-benefit analysis of the particular system-under consideration.

Part 3, titled "Business failure prediction," includes four chapters that involve bankruptcy prediction, while another two involve prediction of bank and thrift failure. Part 4, "Debt risk assessment," contains four chapters that explore the use of neural networks to assess risk. The first two chapters concern the task of bond rating, while the last two chapters concentrate on risk assessment in mortgage underwriting, one of which examines the development and use of a neural network system that utilizes a hierarchical organization of multiple neural nets, where the nets at each level act as a panel of judges to generate risk classifications.

FOR TRADERS
Part 5, titled "Security market applications," is the longest section and will also be of most interest to stock and futures traders. Of the eight chapters here, four deal with stock market prediction, three with prediction in the futures markets and one with testing arbitrage pricing theory. The first chapter depicts a study that uses neural networks to predict daily stock returns for IBM, using a simple model to bring out relevant issues without undo complication. The next chapter compares a neural network approach with the multiple discriminant analysis (MDA) model to predict stock price performance. Here, the authors show that the neural network approach yielded better performance than MDA and also offer some insight into the effects of various network architectures on performance. Another chapter examines the use of modular neural networks in a stock market trading system. The authors discuss work to develop a TOPIX buying and selling prediction system that utilizes a number of neural network modules to generate signals. The chapter discusses many aspects of training and testing of the networks, including the types of prediction simulations used as well as a look at how to extract information from the trained networks. The next chapter focuses on stock price pattern recognition. The researchers use a less-familiar type of neural network especially designed for problems that are dependent on a temporal context. These networks are successfully used to recognize triangle patterns on candlestick charts.

The next three chapters in Part 5 deal with the futures markets. In one, W.E. Bosarge, Jr., explores some theoretical issues behind market prediction, explaining that adaptive processes such as neural networks may be used to exploit nonlinear structures, like those found in the financial markets. After a discussion of chaos and the efficient market model, Bosarge describes the use of neural networks to take advantage of these nonlinear structures. The following chapter discusses a commodity trading model based on a hybrid neural network/expert system. Here, the authors discuss a labor-intensive method of selecting quality training data for their neural networks coupled with the use of an expert system to handle risk management. They then show their system's theoretical potential in trading the Standard & Poor's 500. The next chapter examines continued work on training a neural network to recognize a buy and sell pattern for the live cattle market and includes training results and actual trading performance of the network for six months in 1991. The final chapter in Part 5 examines the use of neural networks in testing arbitrage pricing theory.

Part 6, titled "Neural network approaches to financial forecasting," contains three chapters, the first of which introduces neural networks as an alternative to regression analysis, one of the most popular quantitative methods used in finance. The next chapter presents an empirical study that compares neural networks to the traditional Box-Jenkins forecasting model for predicting time series; a testing method used to determine which parameters and architecture to use for neural networks is also explained. The remaining chapter examines constructive learning and an application of that type of learning to currency exchange rate forecasting. Neural networks typically have a static or fixed architecture, but because different architectures can have very different results, finding a good architecture can be something of an art. Constructive learning is an attempt to overcome some of these disadvantages by allowing the network to "grow" as it learns. This type of learning is then compared with fixed architecture networks and statistical forecasting.

Though not an exhaustive treatment of the subject, Neural Networks in Finance and Investing provides the reader with an informative cross-section of real-world neural network applications in the financial industry. It contains articles that help put this new technology in perspective and offers suggestions on, and insight into, how neural networks can be applied in the trading arena.

Louis Mendelsohn is president of Market Technologies, Wesley Chapel, FL, a research, development and consulting firm involved in the application of artificial intelligence to financial market analysis.

Reprinted from Technical Analysis of
Stocks & Commodities magazine. (C) 1993 Technical Analysis, Inc.,
4757 California Avenue S.W., Seattle, WA 98116-4499, (800) 832-4642.

Want to see how you can use VantagePoint
to give you an edge in your trading? Click Here >>>