A-P.N. Refenes, A.N. Burgess, and Y. Bentz. Neural Networks in Financial Engineering: A Study in Methodology. IEEE Transactions on Neural Networks, 8(6):1222--1267, 1997.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....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 knowledge of the form that this ....

....final set of archived Pareto optimal members, F T, should provide an estimate of the trade off of the risk return defined by the generating process and trading strategy. Financial forecasting (modelling the generating process of a financial time series, or process) is a popular application of NNs [1, 3, 18, 20, 23, 24, 28, 34, 39, 40, 43, 44, 47, 54]. However, in a number of studies misleading claims are made (or inferred) with regards to the actually efficiency of the models presented. Typically the accuracy of a model is described for some data set (usually in terms of Euclidean error) and an estimate of the profit generated 13 by using ....

A-P.N. Refenes, A.N. Burgess, and Y. Bentz. Neural Networks in Financial Engineering: A Study in Method- ology. IEEE Transactions on Neural Networks, 8(6):1222-1267, 1997.

....time t is defined by: t = # # S S (3) Stock prices would follow the following diffusion process : ## # # I q# q # q q#T qT 1 0 and where = S is the stock price, is the drift rate by unit of time and is the instantaneous volatility. 603 According to Refenes et al. [12], we will use technical indicators directly resulting from the outputs of the residues: t 20, t 40 : returns ; t 20 : differences of returns ; K(20) K(40) oscillators ; MM(10) MM(50) moving averages ; MME(10) MME(50) exponential moving averages. ....

Refenes A. N., Burgess A.N. and Bentz Y.: Neural Networks in Financial Engineering: A Study in Methodology. IEEE Transactions on Neural Networks 8(6) (1997) 1222-1267

....years sixties and seventies, despite the heavy use of charts and technical indicators by the professional community. On the basis of all empirical evidences, we will consider that there is some interest in trying to predict the evolution of financial asset prices, as do Refenes, Burgess and Bentz [4] in their introduction to the methods used in financial engineering. When time series prediction is viewed as a regression problem, the inputs being past values of the series and exogenous variables, one may expect useful information (for the prediction of the series) to be contained in these ....

Refenes A.P., Burgess A.N. and Bentz Y., "Neural Networks in Financial Engineering: A Study in Methodology", IEEE Transactions on Neural Networks, vol. 8, n6, pp. 1222-1267, November 1997.

.... we selected international indices of security prices (SBF 250, S P500, Topix, FTSE100, etc) exchange rates (Dollar Mark, Dollar Yen, etc) and interest rates (T Bills 3 months, US Treasury Constant Maturity 10 years, etc) We used 42 technical indicators in total [17] chosen according to [18] and [19] based on these exogenous variables and of course also on the past values of the series. We used 2600 daily data of the BEL20 index over 10 years to have a significant data set. The problem considered here is to forecast the sign of the variation of the BEL20 index at time t 5, from ....

A.N. Refenes, A.N. Burgess, Y. Bentz, Neural networks in financial engineering: a study in methodology, IEEE Transactions on Neural Networks, 8-6 (1997) 1222-1267.

....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 this ....

A-P.N. Refenes, A.N. Burgess, and Y. Bentz. Neural Networks in Financial Engineering: A Study in Methodology. IEEE Transactions on Neural Networks, 8(6):1222--1267, 1997.

....niveau du SBF demain est proche de celui d aujourd hui est trivial. Par contre, dterminer si la bourse va monter ou descendre est beaucoup plus complexe. Le rendement journalier r t du rsidu R t l instant t est dfini par : r t = 1 1 t t t R R R (3) Suivant en cela Refenes et al. [12], nous allons utiliser des indicateurs techniques directement issus des rendements des rsidus : r t , r t 10 , r t 20, r t 40 : rendements ; r t r t 5 , r t 5 r t 10 , r t 10 r t 15 , r t 15 r t 20 : diffrences de rendements ; K(20) K(40) oscillateurs ; MM(10) MM(50) moyennes ....

Refenes A. N., Burgess A.N. and Bentz Y., "Neural Networks in Financial Engineering: A Study in Methodology", IEEE Transactions on Neural Networks, vol. 8, November 1997.

....also discussed. Index Terms Neural Networks, Evolutionary Strategies, Multiple Objectives, Time Series Forecasting. I. INTRODUCTION The use of NNs in the time series forecasting domain is now well established, there are already review papers on this matter [2] as well as methodology studies [8, 10]. The main attribute which separates NN time series modelling from traditional Econometric methods, and the reason practitioners most often site for their use, is their ability to generate non linear relationships between time series input variables with little or no a priori knowledge of the form ....

Refenes, A-P.N., Burgess, A.N. and Bentz, Y. "Neural Networks in Financial Engineering: A Study in Methodology", IEEE Transactions on Neural Networks, Vol. 8, No. 6, pp1222-1267,1997.

....Simple theoretical trading strategies are also mentioned, highlighting real applications of the system. 1. INTRODUCTION Time series forecasting is an important research area in several domains. Recently, neural networks and other advanced methods on prediction have been used in financial domains [1 3]. Peters [4] notes that most financial markets are not Gaussian in nature and tend to have sharper peaks and fat tails, a phenomenon well known in practice. In the face of such evidence, a number of traditional methods based on Gaussian normality assumption have limitations making accurate ....

A. N. Refenes, N. Burgess and Y. Bentz, "Neural networks in financial engineering: A study in methodology," IEEE Transactions on Neural Networks, vol. 8, no. 6, pp. 1222-1267, 1997.

....PMRS approach of pattern matching of up to 5 historic differences ( values) which therefore uses the information contained in the most recent 6 lags. In other circumstances the 6 lags chosen would be those with the highest partial autocorrelation function value when correlated with the actual [5]) In our study, neural networks have two hidden layers with 5 sigmoidal nodes in each and the networks are fully connected. The learning rate was set at 0.05 with a momentum of 0.5. The Neural Networks training was stopped when the combined RMSE on the test and training set had fallen by less ....

Refenes A.N., Burgess N. and Bentz Y. Neural networks in financial engineering: A study in methodology. IEEE Transactions on Neural Networks 1997; 8:6:1222-1267.

....The results are discussed on three benchmark series and the real US S P financial index. 1. Introduction Forecasting is important in several domains and a large number of studies have used classical statistical methods for predicting series behaviour. Advanced methods such as neural networks [1,7,12], genetic algorithms [4] Markov models [9] and fuzzy methods have also been frequently used [4] Farmer and Sidorowich [6] have found that chaotic time series prediction is several order of magnitude better using local approximation techniques than universal approximators. Local approximation ....

....in correctly predicting the output is measured using t statistic. The five most important variables are selected for further analysis. Here we make an assumption that linear input selection is appropriate for a further non linear analysis; this is in line with the work done by Refenes et al. [12]. Hence, for both networks there are five inputs. There is one output which in the first network is a 0 or 1 to predict whether the series goes up or down, and a real number for the second network to predict the actual changes. The number of hidden nodes in Figure 7 are selected using the ....

Refenes, A. N., Burgess, A. N. and Bentz, Y. "Neural networks in financial engineering: A study in methodology," IEEE Transactions on Neural Networks, vol. 8, no. 6, 1997.

....Behaviour Profiling Nearest Neighbours Time Series Prediction and Estimation III I. INTRODUCTION Forecasting is important in several domains and a large number of studies have used classical statistical methods for predicting series behaviour [1] Advanced methods such as neural networks [2, 3], Markov models [4] genetic algorithms and fuzzy methods have also been frequently used [5] Farmer and Sidorowich [6] have found that chaotic time series prediction is several orders of magnitude better using local approximation techniques than universal approximators. Local approximation refers ....

A. N. Refenes, A. N. Burgess and Y. Bentz, "Neural networks in financial engineering: A study in methodology," IEEE Transactions on Neural Networks, vol. 8, no. 6, pp. 1222-1267 (1997).

.... techniques such as ARIMA models have been extensively used for forecasting (Delurgio, 1988) It has been realised that statistical techniques have limited capabilities when modelling time series data, and more advanced methods including neural networks have been frequently used (Azoff, 1994; Refenes et al. 1997). Farmer and Sidorowich (1988) state that local approximation and nearest neighbour techniques can be used in forecasting to give results several orders of magnitude better than conventional statistical techniques. The main characteristic of a nearest neighbour method lies in the fact that it is a ....

Refenes, A. N., Burgess, A. N. and Bentz, Y. 1997. "Neural networks in financial engineering: A study in methodology," IEEE Transactions on Neural Networks, vol. 8, no. 6, pp. 1222-1267.

....in correctly predicting the output is measured using t statistic. The five most important variables are selected for further analysis. Here we make an assumption that linear input selection is appropriate for a further non linear analysis; this is in line with the work done by Refenes et al. [14]. Hence, for both networks there are five inputs. There is one output which in the first network is a 0 or 1 to predict whether the series goes up or down, and a real number for the second network to predict the actual changes. The number of hidden nodes are selected using the procedure discussed ....

A. N. Refenes, N. Burgess and Y. Bentz, Neural networks in financial engineering: A study in methodology, IEEE Transactions on Neural Networks, 8(6), pp. 12221267, (1997).

.... past, conventional statistical techniques such as ARIMA models have been extensively used for forecasting [1] It has been realised that statistical techniques have limited capabilities when modelling time series data, and more advanced methods including neural networks have been frequently used [2,3]. Farmer and Sidorowich [4] state that local approximation and nearest neighbour techniques can be used in forecasting to give results several orders of magnitude better than conventional statistical techniques. The main characteristic of a nearest neighbour method lies in the fact that it is a ....

Refenes, A. N., Burgess, A. N. and Bentz, Y., "Neural networks in financial engineering: A study in methodology," IEEE Transactions on Neural Networks, vol. 8, no. 6, pp. 1222-1267, 1997.

....3 1. MOTIVATION Time series forecasting is an important research area in several domains. Traditionally, forecasting research and practice has been dominated by statistical methods. More recently, neural networks and other advanced methods on prediction have been used in financial domains [1 3]. As we get to know more about the dynamic nature of the financial markets, the weaknesses of traditional methods become apparent. In the last few years, research has focussed on understanding the nature of financial markets before applying methods of forecasting in domains including stock ....

....this section we first discuss the various issues involved in developing an optimal multi layer perceptron architecture for forecasting. This forecasting model will be compared with PMRS. Neural Networks In the recent past, neural networks have been extensively used for forecasting stock markets [1 3, 11]. In practice, financial markets depend heavily on the standard neural network architectures and training algorithms; it will be realistic to assume that a multi layer perceptron with backpropagation is a market standard. In this paper we use the standard MLP architecture for comparison on ....

[Article contains additional citation context not shown here]

A. N. Refenes, N. Burgess and Y. Bentz, "Neural networks in financial engineering: A study in methodology," IEEE Transactions on Neural Networks, vol. 8, no. 6, pp. 1222-1267, 1997.

No context found.

A-P.N. Refenes, A.N. Burgess, and Y. Bentz. Neural Networks in Financial Engineering: A Study in Methodology. IEEE Transactions on Neural Networks, 8(6):1222--1267, 1997.

No context found.

A.-P. Refenes, A. Burgess, and Y. Bentz, "Neural Networks in Financial Engineering: A Study in Methodology," IEEE Transactions on Neural Networks, vol. 8, no. 6, pp. 1222--1267, 1997.

No context found.

A. N. Refenes, N. Burgess and Y. Bentz, Neural networks in financial engineering: A study in methodology, IEEE Transactions on Neural Networks, 8(6), pp. 12221267, (1997).

No context found.

A. N. Refenes, N. Burgess and Y. Bentz, "Neural networks in financial engineering: A study in methodology," IEEE Transactions on Neural Networks, vol. 8, no. 6, pp. 1222-1267, 1997.

No context found.

A-P.N. Refenes, A.N. Burgess, and Y. Bentz. Neural Networks in Financial Engineering: A Study in Methodology. IEEE Transactions on Neural Networks, 8(6):1222--1267, 1997.

No context found.

A-P.N. Refenes, A.N. Burgess, and Y. Bentz. Neural Networks in Financial Engineering: A Study in Methodology. IEEE Transactions on Neural Networks, 8(6):1222--1267, 1997.

No context found.

A.P. Refenes, A.N. Burgess, Y. Bentz, "Neural Networks in Financial Engineering: A Study in Methodology",IEEE transactions on Neural networks, Vol. 8, n. 6, pp.1222-1267, November 1997.

No context found.

A.N. Refenes, A.N. Burgess and Y. Bentz, \Neural Networks in Financial Engineering: A Study in Methodology," IEEE Trans. on Neural Networks, vol. 8, pp. 1222-1267, 1997.

No context found.

A.P. Refenes, A.N.Burgess and Y. Bentz (1997) Neural Networks in Financial Engineering: A Study in Methodology, IEEE Transactions on Neural Networks, vol. 8, n6, pp. 1222-1267.

No context found.

A.P.N. Refenes, A.N. Burgess and Y. Bentz, (1997), Neural Networks in Financial Engineering: A Study in Methodology, IEEE Transactions on Neural Networks 8(6).

First 50 documents

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