Predicting stock returns by classifier ensembles
Applied Soft Computing
| Hi-index | 0.00 |
Combining unbiased forecasts of continuous variables necessarily reduces the forecast error variance below that of a typical individual forecast. However, this does not necessarily hold for forecasts of discrete variables, or where the costs of errors are not directly related to the error variance. This paper investigates the benefits of combining forecasts of outperforming shares, based on one linear and four non-linear statistical classification techniques, including neural network and recursive partitioning methods. All produce excess returns. Combining by simple ‘majority voting’ improves accuracy and profitability. Much greater gains come from applying the ‘unanimity principle’, whereby a share is not held in the high-performing portfolio unless all classifiers agree. Copyright © 2007 John Wiley & Sons, Ltd.