Neural Networks in Finance book cover

Neural Networks in Finance

Gaining Predictive Edge in the Market

This book explores the intuitive appeal of neural networks and the genetic algorithm in finance. It demonstrates how neural networks used in combination with evolutionary computation outperform classical econometric methods for accuracy in forecasting, classification and dimensionality reduction. McNelis utilizes a variety of examples, from forecasting automobile production and corporate bond spread, to inflation and deflation processes in Hong Kong and Japan, to credit card default in Germany to bank failures in Texas, to cap-floor volatilities in New York and Hong Kong.

Upper division undergraduates and MBA students, as well as the rapidly growing number of financial engineering programs, whose curricula emphasize quantitative applications in financial economics and markets

Hardbound, 256 Pages

Published: December 2004

Imprint: Academic Press

ISBN: 978-0-12-485967-8


  • "This book clarifies many of the mysteries of Neural Networks and related optimization techniques for researchers in both economics and finance. It contains many practical examples backed up with computer programs for readers to explore. I recommend it to anyone who wants to understand methods used in nonlinear forecasting." -- Blake LeBaron, Professor of Finance, Brandeis University "An important addition to the select collection of books on financial econometrics, Paul Mcnelis' volume, Neural Networks in Finance, serves as an important reference on neural network models of nonlinear dynamics as a practical econometric tool for better decision-making in financial markets." -- Roberto S. Mariano, Dean of School of Economics and Social Sciences & Vice-Provost for Research, Singapore Management University; Professor Emeritus of Economics, University of Pennsylvania "This book represents an impressive step forward in the exposition and application of evolutionary computational tools. The author illustrates the potency of evolutionary computational tools through multiple examples, which contrast the predictive outcomes from the evolutionary approach with others of a linear and general non-linear variety. The book will be of utmost appeal to both academics throughout the social sciences as well as practitioners, especially in the area of finance." -- Carlos Asilis, Portfolio Manager, VegaPlus Capital Partners; formerly Chief Investment Strategist, JPMorgan Chase " excellent, easy-to read introduction to the math behind neural networks." - Financial Engineering News


  • Preface1 Introduction1.1 Forecasting, Classification and Dimensionality Reduction1.2 Synergies1.3 The Interface Problems1.4 Plan of the BookEconometric Foundations2 What Are Neural Networks2.1 Linear Regression Model2.2 GARCH Nonlinear Models2.2.1 Polynomial Approximation2.2.2 Orthogonal Polynomials2.3 Model Typology2.4 What Is A Neural Network2.4.1 Feedforward Networks2.4.2 Squasher Functions2.4.3 Radial Basis Functions2.4.4 Ridgelet Networks2.4.5 Jump Connections2.4.6 Multilayered Feedforward Networks2.4.7 Recurrent Networks2.4.8 Networks with Multiple Outputs2.5 Neural Network Smooth-Transition Regime-Switching Models2.5.1 Smooth Transition Regime Switching Models2.5.2 Neural Network Extensions2.6 Nonlinear Principal Components: \ Intrinsic Dimensionality2.6.1 Linear Principal Components2.6.2 Nonlinear Principal Components2.6.3 Application to Asset Pricing2.7 Neural Networks and Discrete Choice2.7.1 Discriminant Analysis2.7.2 Logit Regression2.7.3 Probit Regression2.7.4 Weibull Regression2.7.5 Neural Network Models for Discrete Choice2.7.6 Models with Multinomial Ordered ChoiceCriticism and Data Mining2.9 Conclusion2.9.1 Matlab Program Notes2.9.2 Suggested Exercises3 Estimation of a Network with Evolutionary Computation3.1 Data Preprocessing3.1.1 Stationarity: Dickey-Fuller Test3.1.2 Seasonal Adjustment: Correction for Calendar Effects3.1.3 Scaling of Data3.2 The Nonlinear Estimation Problem3.2.1 Local Gradient-Based Search: \ The Quasi- Backpropagation 46 Simulated Annealing 48 3.2.3 Evolutionary Stochastic Search: The Genetic AlgorithmPopulation creationSelectionCrossoverMutationElection tournamentElitismConvergence3.2.4 Evolutionary Genetic Algorithms3.2.5 Hybridization: Coupling Gradient- and Genetic Search Methods3.3 Repeated Estimation and Thick Models3.4 Matlab Examples: Numerical Performance 53 3.4.1 Numerical Optimization3.4.2 Approximation with Networks 54 3.5 Conclusion3.5.1 Matlab Program Notes3.5.2 Suggested Exercises4 Evaluation of Network Estimation4.1 In-Sample Criteria4.1.1 Goodness of Fit Measure4.1.2 Hannan-Quinn Information Criterion4.1.3 Serial Independence and Homoskedasticity: and McLeod-Li Tests4.1.4 SymmetryNormality4.1.6 Neural Network Test for Neglected Nonlinearity: Lee-White-Granger Test4.1.7 Brock-Deckert-Scheinkman Test for Nonlinear Patterns4.1.8 Summary of in-sample criteria4.1.9 Matlab Example4.2 Out-of-Sample Criteria4.2.1 Recursive Methodology4.2.2 Root Mean Squared Error Statistic4.2.3 Diebold-Mariano Test for Out of Sample Errors4.2.4 Harvey, Leybourne, and Newbold "Size Correction" of Diebold-Mariano Test4.2.5 Out-of-Sample Comparison with Nested Models4.2.6 Success Ratio for Sign Predictions: Directional Accuracy4.2.7 Predictive Stochastic\ Complexitysubsection umberline 4.2.8 Cross-Validation and the Method 69 How Large for Predictive Accuracy4.3 Interpretive Criteria and Significance of Results4.3.1 Analytic Derivatives4.3.2 Finite Differences4.3.3 Does It Matter4.3.4 Matlab Example: Analytic and Finite Differences4.3.5 Bootstrapping for Assessing Significance4.4 Implementation Strategy4.5 Conclusion4.5.1 Matlab Program Notes4.5.2 Suggested Exercises1em Applications and Examples5 Estimation and Forecasting with Artificial Data5.1 Introduction5.2 Stochastic Chaos Model5.2.1 In-Sample Performance5.2.2 Out-of-Sample Performance5.3 Stochastic Volatility/Jump Diffusion Model5.3.1 In-Sample Performance5.3.2 Out-of-Sample Performance5.4 The Markov Regime Switching Model5.4.1 In-Sample Performance5.4.2 Out-of-Sample Performance5.5 VRS Model5.5.1 In-Sample Performance5.6 Distorted Long Memory Model5.6.1 In-Sample Performance5.6.2 Out-of-Sample Performance5.7 BSOP Model: Implied Volatility Forecasting5.7.1 In-Sample Performance5.7.2 Out-of-Sample Performance5.8 Conclusion5.8.1 Matlab Program Notes5.8.2 Suggested Exercises6 Times Series: Examples from Industry and Finance6.1 Forecasting Production in the Automotive Industry6.1.1 The Data6.1.2 Models of Quantity Adjustment6.1.3 In-Sample Performance6.1.4 Out-of-Sample Performance6.1.5 Interpretation of Results6.2 Corporate Bonds: Which Spreads? 110 6.2.1 The Data6.2.2 A Model for the Adjustment of SpreadsIn-Sample Performance6.2.4 Out-of-Sample Performance6.2.5 Interpretation of Results6.3 Conclusion6.3.1 Matlab Program Notes6.3.2 Suggested Exercises7 Inflation and Deflation: Hong Kong and Japan7.1 Hong Kong7.1.1 The Data7.1.2 Model Specification7.1.3 In-Sample Performance7.1.4 Out-of-Sample Performance7.1.5 Interpretation of Results7.2 Japan7.2.1 The Data7.2.2 Model Specification7.2.3 In-Sample Performance7.2.4 Out-of-Sample Performance7.2.5 Interpretation of Results7.3 Conclusion7.3.1 Matlab Program Notes7.3.2 Suggested Exercises8 Classification: \ Credit Card Default and Bank Failures8.1 Credit Card Risk8.1.1 The Data8.1.2 In-Sample Performance8.1.3 Out-of-Sample Performance8.1.4 Interpretation of Results8.2 Banking Intervention8.2.1 The Data8.2.2 In-Sample Performance8.2.3 Out-of-Sample Performance8.2.4 Interpretation of Results8.3 Conclusion8.3.1 Matlab Program Notes8.3.2 Suggested Exercises9 Dimensionality Reduction and Implied Volatility Forecasting9.1 Hong Kong9.1.1 The Data9.1.2 In-Sample Performance9.1.3 Out-of-Sample Performance9.2 United States9.2.1 The Data9.2.2 In-Sample Performance9.2.3 Out-of-Sample Performance9.3 Conclusion9.3.1 Matlab Program Notes9.3.2 Suggested Exercises


advert image