Georgios Sermpinis

Forecasting foreign exchange rates with adaptive neural networks using radial-basis functions and Particle Swarm Optimization

The motivation for this paper is to introduce a hybrid neural network architecture of Particle Swarm Optimization and Adaptive Radial Basis Function (ARBF–PSO), a time varying leverage trading strategy based on Glosten, Jagannathan and Runkle (GJR) volatility forecasts and a neural network fitness function for financial forecasting purposes. This is done by benchmarking the ARBF–PSO results with those of three different neural networks architectures, a Nearest Neighbors algorithm (k-NN), an autoregressive moving average model (ARMA), a moving average convergence/divergence model (MACD) plus a nai¨ve strategy. More specifically, the trading and statistical performance of all models is investigated in a forecast simulation of the EUR/USD, EUR/GBP and EUR/JPY ECB exchange rate fixing time series over the period January 1999–March 2011 using the last 2years for out-of-sample testing.

Higher order and recurrent neural architectures for trading the EUR/USD exchange rate

The motivation for this paper is to investigate the use of alternative novel neural network architectures when applied to the task of forecasting and trading the Euro/Dollar (EUR/USD) exchange rate. This is done by benchmarking three different neural network designs representing a Higher Order Neural Network (HONN), a Psi Sigma Network and a Recurrent Network (RNN) with three successful architectures, the traditional Multilayer Perceptron (MLP), the Softmax and the Gaussian Mixture (GM) models. More specifically, the trading performance of the six models is investigated in a forecast and trading simulation competition on the EUR/USD time series over a period of 8 years. These results are also benchmarked with more traditional models such as a moving average convergence divergence technical model (MACD), an autoregressive moving average model (ARMA) and a logistic regression model (LOGIT). As it turns out, the MLP, the HONN, the Psi Sigma and the RNN models all do well and outperform the more traditional models in a simple trading simulation exercise. However, when more sophisticated trading strategies using confirmation filters and leverage are applied, the GM network produces remarkable results and outperforms all the other network architectures.

Forecasting and trading the EUR/USD exchange rate with Gene Expression and Psi Sigma Neural Networks

Highlights? We investigate the use of Psi Sigma Neural Network and the Gene Expression. ? We benchmark their results with five different linear and non-linear models. ? We introduce a time-varying leverage strategy. The motivation for this paper is to investigate the use of two promising classes of artificial intelligence models, the Psi Sigma Neural Network (PSI) and the Gene Expression algorithm (GEP), when applied to the task of forecasting and trading the EUR/USD exchange rate. This is done by benchmarking their results with a Multi-Layer Perceptron (MLP), a Recurrent Neural Network (RNN), a genetic programming algorithm (GP), an autoregressive moving average model (ARMA) plus a naive strategy. We also examine if the introduction of a time-varying leverage strategy can improve the trading performance of our models.

Modeling, forecasting and trading the EUR exchange rates with hybrid rolling genetic algorithms—Support vector regression forecast combinations

The motivation of this paper is to introduce a hybrid Rolling Genetic Algorithm-Support Vector Regression (RG-SVR) model for optimal parameter selection and feature subset combination. The algorithm is applied to the task of forecasting and trading the EUR/USD, EUR/GBP and EUR/JPY exchange rates. The proposed methodology genetically searches over a feature space (pool of individual forecasts) and then combines the optimal feature subsets (SVR forecast combinations) for each exchange rate. This is achieved by applying a fitness function specialized for financial purposes and adopting a sliding window approach. The individual forecasts are derived from several linear and non-linear models. RG-SVR is benchmarked against genetically and non-genetically optimized SVRs and SVMs models that are dominating the relevant literature, along with the robust ARBF-PSO neural network. The statistical and trading performance of all models is investigated during the period of 1999–2012. As it turns out, RG-SVR presents the best performance in terms of statistical accuracy and trading efficiency for all the exchange rates under study. This superiority confirms the success of the implemented fitness function and training procedure, while it validates the benefits of the proposed algorithm.

Modelling and trading the EUR/USD exchange rate at the ECB fixing

The motivation for this paper is to investigate the use of alternative novel neural network (NN) architectures when applied to the task of forecasting and trading the euro/dollar (EUR/USD) exchange rate, using the European Central Bank (ECB) fixing series with only auto-regressive terms as inputs. This is done by benchmarking four different NN designs representing a higher-order neural network (HONN), a Psi Sigma Network and a recurrent neural network with the classic multilayer perception (MLP) and some traditional techniques, either statistical such as an auto-regressive moving average model, or technical such as a moving average convergence/divergence model, plus a naïve strategy. More specifically, the trading performance of all models is investigated in a forecast and trading simulation on the EUR/USD ECB fixing time series over the period 1999–2007 using the last one and half years for out-of-sample testing, an original feature of this paper. We use the EUR/USD daily fixing by the ECB as many financial institutions are ready to trade at this level and it is therefore possible to leave orders with a bank for business to be transacted on that basis. As it turns out, the MLP does remarkably well and outperforms all other models in a simple trading simulation exercise. However, when more sophisticated trading strategies using confirmation filters and leverage are applied, the HONN network produces better results and outperforms all other NN and traditional statistical models in terms of annualized return.

Operational risk: Emerging markets, sectors and measurement

The role of decision support systems in mitigating operational risks in firms is well established. However, there is a lack of investment in decision support systems in emerging markets, even though inadequate operational risk management is a key cause of discouraging external investment. This has also been exacerbated by insufficient understanding of operational risk in emerging markets, which can be attributed to past operational risk measurement techniques, limited studies on emerging markets and inadequate data.

Modelling commodity value at risk with higher order neural networks

The motivation for this article is to investigate the use of a promising class of Neural Network (NN) models, Higher Order Neural Networks (HONNs), when applied to the task of forecasting the 1-day ahead Value at Risk (VaR) of the brent oil and gold bullion series with only autoregressive terms as inputs. This is done by benchmarking their results with those of a different NN design, the Multilayer Perceptron (MLP), an Extreme Value Theory (EVT) model along with some traditional techniques, such as an Autoregressive Moving Average Model-Generalized Autoregressive Conditional Heteroscedasticity (ARMA-GARCH) (1,1) model and the RiskMetrics volatility. In addition to these, we also examine two hybrid NNs-RiskMetrics volatility models. More specifically, the forecasting performance of all models for computing the VaR of the brent oil and the gold bullion is examined over the period 2002 to 2008 using the last year for out-of-sample testing. The evaluation of our models is done by using a series of backtesting algorithms and two loss functions: a violation ratio calculating when the realized return exceeds the forecast VaR and an average squared violation magnitude function, firstly introduced in this article, computing the average magnitude of the violations. As it turns out, the hybrid HONNs-RiskMetrics model does remarkably well and outperforms all other models in forecasting the VaR of gold and oil at both the 5% and 1% confidence levels, providing an accurate number of independent violations which also have the lowest magnitude on average. The pure HONNs and MLPs along with the hybrid MLP-RiskMetrics model also give satisfactory forecasts in most cases.

Modelling and trading the realised volatility of the FTSE100 futures with higher order neural networks

The motivation for this article is the investigation of the use of a promising class of neural network (NN) models, higher order neural networks (HONNs), when applied to the task of forecasting and trading the 21-day-ahead realised volatility of the FTSE 100 futures index. This is done by benchmarking their results with those of two different NN designs, the multi-layer perceptron (MLP) and the recurrent neural network (RNN), along with a traditional technique, RiskMetrics. More specifically, the forecasting and trading performance of all models is examined over the eight FTSE 100 futures maturities of the period 2007–2008 using the realised volatility of the last 21 trading days of each maturity as the out-of-sample target. The statistical evaluation of our models is done by using a series of measures such as the mean absolute error, the mean absolute percentage error, the root-mean-squared error and the Theil U-statistic. Then we apply a simple trading strategy to exploit our forecasts based on trading at-the-money call options on FTSE 100 futures. As it turns out, HONNs demonstrate a remarkable performance and outperform all other models not only in terms of statistical accuracy but also in terms of trading efficiency. We also note that both the RNNs and MLPs provide sufficient results in the trading application in terms of cumulative profit and average profit per trade.

Stock market prediction using evolutionary support vector machines: an application to the ASE20 index

The main motivation for this paper is to introduce a novel hybrid method for the prediction of the directional movement of financial assets with an application to the ASE20 Greek stock index. Specifically, we use an alternative computational methodology named evolutionary support vector machine (ESVM) stock predictor for modeling and trading the ASE20 Greek stock index extending the universe of the examined inputs to include autoregressive inputs and moving averages of the ASE20 index and other four financial indices. The proposed hybrid method consists of a combination of genetic algorithms with support vector machines modified to uncover effective short-term trading models and overcome the limitations of existing methods. For comparison purposes, the trading performance of the ESVM stock predictor is benchmarked with four traditional strategies (a naïve strategy, a buy and hold strategy, a moving average convergence/divergence and an autoregressive moving average model), and a multilayer perceptron neural network model. As it turns out, the proposed methodology produces a higher trading performance, even during the financial crisis period, in terms of annualized return and information ratio, while providing information about the relationship between the ASE20 index and DAX30, NIKKEI225, FTSE100 and S&P500 indices.

European Exchange Trading Funds Trading with Locally Weighted Support Vector Regression

In this paper, two different Locally Weighted Support Vector Regression (wSVR) algorithms are generated and applied to the task of forecasting and trading five European Exchange Traded Funds. The trading application covers the recent European Monetary Union debt crisis. The performance of the proposed models is benchmarked against traditional Support Vector Regression (SVR) models. The Radial Basis Function, the Wavelet and the Mahalanobis kernel are explored and tested as SVR kernels. Finally, a novel statistical SVR input selection procedure is introduced based on a principal component analysis and the Hansen, Lunde, and Nason (2011) model confidence test. The results demonstrate the superiority of the wSVR models over the traditional SVRs and of the v-SVR over the e-SVR algorithms. We note that the performance of all models varies and considerably deteriorates in the peak of the debt crisis. In terms of the kernels, our results do not confirm the belief that the Radial Basis Function is the optimum choice for financial series.