Table 3

RESULTS OF GENETIC ADAPTIVE NEURAL NETWORK (GANN) APPROXIMATION

OF EURODOLLAR CALL FUTURES OPTION TRADED ON LIFFE

Panel A: Training Data Set

   

CS OPM Approximation

GANN Approximation

Sub-sample

Description

Number of

Observations

Mean Squared

Error (MSE)

Mean Absolute

Error (MAE)

Mean Squared

Error (MSE)

Mean Absolute

Error (MAE)

Complete

Sample

2000

0.02012

0.07743

0.00047

0.01589

M* > .01

790

0.02939

0.10903

0.00066

0.02012

.01 < M < 1

517

0.03381

0.11526

0.00072

0.02083

M > 1

271

0.02116

0.09785

0.00055

0.01871

M < -.01

1202

0.01348

0.05568

0.00033

0.01306

-1 < M < -.01

577

0.02641

0.09886

0.00048

0.01597

M < -1

619

0.00155

0.01575

0.00019

0.01028

-.01 < M < .01

8

0.10399

0.22549

0.00140

0.02345

Panel B: Holdout Data Set, HOLDOUT1

Complete

Sample

6887

0.01759

0.07162

0.00051

0.01677

M* > .01

2698

0.02822

0.10593

0.00071

0.02098

.01 < M < 1

1782

0.03290

0.11229

0.00076

0.02159

M > 1

910

0.01912

0.09361

0.00061

0.01978

M < -.01

4159

0.01051

0.04892

0.00038

0.01399

-1 < M < -.01

1840

0.02174

0.09017

0.00059

0.01722

M < -1

2312

0.00159

0.01608

0.00022

0.01143

-.01 < M < .01

30

0.04373

0.13311

0.00117

0.02360

* M is a measure of the degree of moneyness. For example, if the strike rate (100 - K) is greater than the futures rate (F(t)), a call option on a 3-month Eurodollar futures contract will be in-the-money. Therefore, M is defined as (100 - K) - F(t), and when M > .01 the call option is in-the-money.

[Back to Document]