A Dea-Cascor Model for High-Frequency Stock Trading:Computational Experiments in the U.S.Stock Market

The paper presents results of computer-assisted portfolio management simulation based on using a DEACascormathematical model. The model uses the Data Envelopment Analysis (DEA) ratio as a neuronwith memory and combines it with Cascade Correlation Neural Network (Cascor) to forecast stockprices. The model is designed for using in high-frequency stock trading. It utilizes ability of DEA toconcentrate multi-faceted information in one indicator scaled to the interval [0,1], a DEA efficiencyindex, and is aimed to compress market information. Cascor combines data of several consecutiveperiods using its flexible structure and generates a buy - sell strategy. The paper presents results of thesimulation of a 50-stock portfolio during a period of 60 consecutive trade days chosen during one of themost problematic period of the U.S. stock market operation. Obtained results allow for optimismregarding its practical use for high-frequency stock trading provided availability of a convenientcomputer – based support.