Prediction of stock market indexes by analogues complexing

On the base of observations in the period of February, 22 to May 30, 1995 (66 days) the analogue complexing algorithm was used. Table shows prediction error (MAD [%]) of four variables (Dollar, Dax, F.A.Z., Dow Jones) and the mean prediction error over 7 variables. The width of the patterns varies from 6 to 15 days.

Width 6 7 8 9 10 11 12 13 14 15
Dollar 2.61 3.28 2.62 2.79 2.91 2.91 1.86 1.48 2.16 8.96
F.A.Z. 1.42 2.61 1.59 1.48 1.19 1.39 1.43 1.87 0.72 1.12
Dax 1.48 1.70 1.7 2.31 2.96 2.76 2.76 2.61 2.37 1.12
Dow J 1.36 1.71 1.62 6.98 5.39 4.65 4.97 3.85 3.36 2.79
Mean 1.17 1.57 1.58 2.36 2.46 2.08 1.94 1.79 1.63 1.88

The forecasts are obtained by means of linear combination of the continuations of 5 selected analogues patterns, where the unknown weights gj of complexing are estimated by means of parametric algorithms.