Neural, Novel & Hybrid Algorithms for Time Series Prediction

An authoritative guide to predicting the future using neural, novel, and hybrid algorithms Expert Timothy Masters provides you with carefully paced, step-by-step advice and guidance plus the proven tools and techniques you need to develop successful applications for business forecasting, stock market prediction, engineering process control, economic cycle tracking, marketing analysis, and more. Neural, Novel & Hybrid Algorithms for Time Series Prediction provides information on: * Robust confidence intervals for predictions made with neural, ARIMA, and other models * Wavelets for detecting features that presage important events * Multivariate ARMA models for simultaneous prediction of multiple series based on multiple inputs and shocks * Hybrid ARMA/neural models to improve the accuracy of predictions * Data reduction and orthogonalization using principal components and related operations * Digital filters for preprocessing to enhance useful information and suppress noise * Diagnostic tools such as the maximum entropy spectrum and Savitzky-Golay filters for suggesting and validating prediction models * Effective preprocessing techniques for prediction with neural networks CD-ROM INCLUDES: * PREDICT-both DOS and Windows NT versions-a powerful time series program that can be easily customized to make accurate predictions in any application area * Much useful source code, including the complex-general multivariate fast Fourier transform in both C++ and Pentium-optimized assembler