M. Adya and F. Collopy. How Effective are Neural Networks at Forecasting and Prediction? A Review and Evalution. International Journal of Forecasting, 17:481--495, 1998.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....adaptive topologies, multiple objectives, time series forecast ing. Corresponding author. I INTRODUCTION The use of neural networks (NNs) in the time series forecasting domain is now well established. There are a number of review papers in this area (for example, Adya and Collopy [1]) as well as methodology studies [39, 43] The main attribute which distinguishes NN time series modelling from traditional econometric methods is their ability to generate non linear relationships between a vector of time series input variables and a dependent series, with little or no a priori ....

....problem are now discussed. 1.2 Problems with the linear combination of errors approach Figure i illustrates the current approach to multi objective training in NN regression and classification. Consider the situation where a number of errors measures (objectives) are used that lie in the range [0,1]. Given that the practitioner wishes to minimise these errors, the typical approach in linear sum back propagation is to minimise the composite error ec. In the D error case (where there are D errors to be minimised) this is: D e C Ole 1 1 O2e 2 1 . ODe D , 0 i I , ViO ai 1 ....

[Article contains additional citation context not shown here]

M. Adya and F. Collopy. How Effective are Neural Networks at Forecasting and Prediction? A Review and Evalution. International Journal of Forecasting, 17:481-495, 1998.

....Network regression models to describe a market generation process in relation to the forecasting of its risk and return. I. INTRODUCTION The use of Neural Networks (NNs) in the time series forecasting domain is now well established, with a number of recent review and methodology studies (e.g. [1], 2] 3] The main attribute which differentiates NN time series modelling from traditional econometric methods is their ability to generate non linear relationships between a vector of time series input variables and a dependent series, with little or no a priori knowledge of the form that ....

....the problems associated with the current approach to multi objectivity in NN regression is provided in Figure 1(a) and 1(b) a) b) Fig. 1. a) Two dimensional error surface 1. b) Two dimensional error surface 2. Consider the situation where two error measures are used that lie in the range [0,1]. Given that the practitioner wishes to minimise errors, the typical approach in linear sum backpropagation is to minimise the composite error , in the D error measure case (where the errors are to be minimised) this is calculated as follows: 8026 6 157 158 159 7 ....

[Article contains additional citation context not shown here]

M. Adya and F. Collopy. How Effective are Neural Networks at Forecasting and Prediction? A Review and Evalution. International Journal of Forecasting, 17:481--495, 1998.

....approaches are fundamentally based on the concept that the most accurate forecast, if not the actual realised (target) value, is that with the smallest Euclidean distance from the actual. When measuring predictor performance however, practitioners use a whole range of different error 2 measures [1]. These error measures on the whole tend to reflect the preferences of potential end users of the forecast model, as opposed to just the Euclidean measure. Recent approaches to time series forecasting using NNs have introduced limited augmentations to the traditional gradient descent algorithm in ....

....This is because end users of time series forecasts commonly have other aims than just Euclidean distance minimised forecasts. Examples of some of these different aims will be discussed in this section. A popular error measure for time series forecasting, due to its unit free nature, is MAPE [1]. If a practitioner has a preference for this measure of accuracy it may also make sense that their model is trained with it in mind. Due to their differing calculation methods, minimising the Euclidean error of a model does not mean that the MAPE error of a model is minimised) The calculation ....

[Article contains additional citation context not shown here]

Adya, M. and Collopy, F., "How Effective are Neural Networks at Forecasting and Prediction? A Review and Evalution", Journal of Forecasting, Vol. 17, pp481-495, 1998.

....technology and they offer a new avenue to explore the dynamics of a variety of financial applications. The backpropagation algorithm [17] has emerged as one of the most widely used learning procedures for multi layer networks. They have been shown to have great potential for financial forecasting [1]. Neural networks can make contributions to the maximization of returns, while reducing costs, and limiting risks. They can simulate fundamental and technical analysis methods using fundamental and technical indicators as inputs. Consumer price index, foreign reserve, GDP, export and import volume ....

M. Adya, F. Collopy, "How Effective are Neural Networks at Forecasting and Prediction? A Review and Evaluation", Journal of Forecasting, Vol. 17, 1998, pp481-495.

....considered mature. In particular, Refenes97] identifies the lack of established procedures for performing tests of statistical significance on the estimated network parameters as the main research issue that must be addressed if neural networks are to become commonplace in financial econometrics. [Adya98] also notes several other disadvantages including the fact that there is no single configuration that is adequate for all domains forcing topology determination to proceed by trial and error in a fairly ad hoc manner. Other issues where care must also be taken include their susceptibility to local ....

M Adya, F Collopy, "How Effective are Neural Networks at Forecasting and Prediction? A Review and Evaluation", Journal of Forecasting, Vol.17 # 5/6. Sep-Nov. 1998, pp.481-495

No context found.

M. Adya and F. Collopy. How Effective are Neural Networks at Forecasting and Prediction? A Review and Evalution. International Journal of Forecasting, 17:481--495, 1998.

No context found.

M. Adya and F. Collopy, "How Effective are Neural Networks at Forecasting and Prediction? A Review and Evalution," International Journal of Forecasting, vol. 17, pp. 481--495, 1998.

No context found.

Adya, M. and F. Collopy 1998. "How effective are neural networks at forecasting and prediction? A review and evaluation." Journal of Forecasting 17(5-6): 481-495.

No context found.

M. Adya and F. Collopy. How Effective are Neural Networks at Forecasting and Prediction? A Review and Evalution. International Journal of Forecasting, 17:481--495, 1998.

No context found.

M. Adya and F. Collopy. How Effective are Neural Networks at Forecasting and Prediction? A Review and Evalution. International Journal of Forecasting, 17:481--495, 1998.

No context found.

M Adya and F Collopy. "How Effective are Neural Networks at Forecasting and Prediction? A Review and Evaluation", Journal of Forecasting, 17(5/6):481--495, 1998.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC