Financial volatility trading using recurrent neural networks

We simulate daily trading of straddles on financial indexes. The straddles are traded based on predictions of daily volatility differences in the indexes. The main predictive models studied are recurrent neural nets (RNN). Such applications have often been studied in isolation. However, due to the special character of daily financial time-series, it is difficult to make full use of RNN representational power. Recurrent networks either tend to overestimate noisy data, or behave like finite-memory sources with shallow memory; they hardly beat classical fixed-order Markov models. To overcome data nonstationarity, we use a special technique that combines sophisticated models fitted on a larger data set, with a fixed set of simple-minded symbolic predictors using only recent inputs. Finally, we compare our predictors with the GARCH family of econometric models designed to capture time-dependent volatility structure in financial returns. GARCH models have been used to trade volatility. Experimental results show that while GARCH models cannot generate any significantly positive profit, by careful use of recurrent networks or Markov models, the market makers can generate a statistically significant excess profit, but then there is no reason to prefer RNN over much more simple and straightforward Markov models. We argue that any report containing RNN results on financial tasks should be accompanied by results achieved by simple finite-memory sources combined with simple techniques to fight nonstationarity in the data.     

A hybrid option pricing model using a neural network for estimating volatility

The Black–Scholes (BS) model is the standard approach used for pricing financial options. However, although being theoretically strong, option prices valued by the model often differ from the prices observed in the financial markets. This paper applies a hybrid neural network which preprocesses financial input data for improving the estimation of option market prices. This model is comprised of two parts. The first part is a neural network developed to estimate volatility. The second part is an additional neural network developed to value the difference between the BS model results and the actual market option prices. The resulting option price is then a summation between the BS model and the network response. The hybrid system with a neural network for estimating volatility provides better performance in terms of pricing accuracy than either the BS model with historical volatility (HV), or the BS model with volatility valued by the neural network.     

Exchange rate forecasting using a combined parametric and nonparametric self-organising modelling approach

Financial prediction has attracted a lot of interest due to the financial implications that the accurate prediction of financial markets can have. A variety of data driven modelling approaches have been applied but their performance has produced mixed results. In this study we apply both parametric (neural networks with active neurons) and nonparametric (analog complexing) self-organising modelling methods for the daily prediction of the exchange rate market. We also propose a combined approach where the parametric and nonparametric self-organising methods are combined sequentially, exploiting the advantages of the individual methods with the aim of improving their performance. The combined method is found to produce promising results and to outperform the individual methods when tested with two exchange rates: the American Dollar and the Deutche Mark against the British Pound.     

Financial prediction and trading strategies using neurofuzzyapproaches

Neurofuzzy approaches for predicting financial time series are investigated and shown to perform well in the context of various trading strategies involving stocks and options. The horizon of prediction is typically a few days and trading strategies are examined using historical data. Two methodologies are presented wherein neural predictors are used to anticipate the general behavior of financial indexes (moving up, down, or staying constant) in the context of stocks and options trading. The methodologies are tested with actual financial data and show considerable promise as a decision making and planning tool.     

Fast semantic object search and detection for vegetable trading information using Steiner tree

We propose an approach to speed up the semantic object search and detection for vegetable trading information using Steiner Tree. Through analysis, comparing the relevant ontology construction method, we present a set of ontology construction methods based on domain ontology for vegetables transaction information. With Jena2 provides rule-based reasoning engine, More related information could be searched with the help of ontology database and ontology reasoning, query expansion is to achieve sub-vocabulary of user input, the parent class of words, equivalence class of extensions, and use of ontology reasoning to get some hidden information to use of these technologies, we design and implementation of ontology-based semantic vegetables transaction information retrieval system, and through compare to keyword-based matching of large-scale vegetable trading site retrieval systems, the results show that the recall and precision rate of ontology-based information retrieval system much better than keyword-based information retrieval system, and has some practical value.     

Mean shifts detection and classification in multivariate process: a neural-fuzzy approach

For monitoring multivariate quality control process, traditional multivariate control charts have been proposed to detect mean shifts. However, a persistent problem is that such charts are unable to provide any shift-related information when mean shifts occur in the process. In fact, the immediate classification of the magnitude of mean shifts can greatly narrow down the set of possible assignable causes, hence facilitating quick analysis and corrective action by the technician before many nonconforming units are manufactured. In this paper, we propose a neural-fuzzy model for detecting mean shifts and classifying their magnitude in multivariate process. This model is divided into training and classifying modules. In the training module, a neural network (NN) model is trained to detect various mean shifts for multivariate process. Then, in the classifying module, the outputs of NN are classified into various decision intervals by using a fuzzy classifier and an additional two-point-in-an-interval decision rule to determine shift status. An example is presented to illustrate the application of the proposed model. Simulation results show that it outperforms the multivariate T²control chart in terms of out-of-control average run length under fixed type I error. In addition, the correct classification percentages are also studied and the general guidelines are given for the proper use of the proposed model.     

Evaluation approach to stock trading system using evolutionary computation

The past researches emphasize merely the avoidance of over-learning at the system level and ignore the problem of over-learning at the model level, which lead to the poor performance of the evolutionary computation based stock trading decision-making system. This study presents a new evaluation approach to focus on evaluating the generalization capability at the model level. An empirical study was provided and the results reveal four important findings. First, the decision-making system generated at the model design stage outperforms the system generated at the model validation stage, which shows over-learning at the model level. Secondly, for the decision-making system generated either at the model design stage or at the model validation stage, the investment performance in the training period is much better than that in the testing period, exhibiting over-learning at the system level. Third, employing moving timeframe approach is unable to improve the investment performance at the model validation stage. Fourth, reducing the evolution generation and input variables are unable to avoid the over-learning at the model level. The major contribution of this study is to clarify the issue of over-learning at the model and the system level. For future research, this study developed a more reliable evaluation approach in examining the generalization capability of evolutionary computation based decision-making system.     

The dynamic relations among return volatility,trading imbalance,and trading volume in futures markets

Trading imbalances reflect the quality of market information and may contain more information than the number of trades or trading volume. In order to better understand how trading imbalances play a role different from traditional variables (i.e., number of trades and trading volume) in explaining volatility, we use intraday data to examine the dynamic relations among return volatility, trading imbalances, and traditional variables for E-mini S&P 500 futures and Japanese Yen futures contracts, respectively. The Granger-causality tests indicate strong feedback effects between volatility and trading variables, confirming the information-based and hedging-based trading. We also compare the results of the traditional volumes and trading imbalances through variance decomposition and impulse responses analysis. It is shown that the sequential arrival of private information through trading imbalance is more important in explaining return volatility than the traditional variables, which are a proxy for the public information.     

A genetic network programming with learning approach for enhanced stock trading model

In this paper, an enhancement of stock trading model using Genetic Network Programming (GNP) with Sarsa Learning is described. There are three important points in this paper: First, we use GNP with Sarsa Learning as the basic algorithm while both Technical Indices and Candlestick Charts are introduced for efficient stock trading decision-making. In order to create more efficient judgment functions to judge the current stock price appropriately, Importance Index (IMX) has been proposed to tell GNP the timing of buying and selling stocks. Second, to improve the performance of the proposed GNP-Sarsa algorithm, we proposed a new method that can learn the appropriate function describing the relation between the value of each technical index and the value of the IMX. This is an important point that devotes to the enhancement of the GNP-Sarsa algorithm. The third point is that in order to create more efficient judgment functions, sub-nodes are introduced in each node to select appropriate stock price information depending on the situations and to determine appropriate actions (buying/selling). To confirm the effectiveness of the proposed method, we carried out the simulation and compared the results of GNP-Sarsa with other methods like GNP with Actor Critic, GNP with Candlestick Chart, GA and Buy&Hold method. The results shows that the stock trading model using GNP-Sarsa outperforms all the other methods.     

Maximizing winning trades using a novel RSPOP fuzzy neural network intelligent stock trading system

The increasing reliance on Computational Intelligence techniques like Artificial Neural Networks and Genetic Algorithms to formulate trading decisions have sparked off a chain of research into financial forecasting and trading trend identifications. Many research efforts focused on enhancing predictive capability and identifying turning points. Few actually presented empirical results using live data and actual technical trading rules. This paper proposed a novel RSPOP Intelligent Stock Trading System, that combines the superior predictive capability of RSPOP FNN and the use of widely accepted Moving Average and Relative Strength Indicator Trading Rules. The system is demonstrated empirically using real live stock data to achieve significantly higher Multiplicative Returns than a conventional technical rule trading system. It is able to outperform the buy-and-hold strategy and generate several folds of dollar returns over an investment horizon of four years. The Percentage of Winning Trades was increased significantly from an average of 70% to more than 92% using the system as compared to the conventional trading system; demonstrating the system’s ability to filter out erroneous trading signals generated by technical rules and to preempt any losing trades. The system is designed based on the premise that it is possible to capitalize on the swings in a stock counter’s price, without a need for predicting target prices.     

Forecasting large scale conditional volatility and covariance using neural network on GPU

Forecasting volatility is an important issue in financial econometric analysis. This paper aims to seek a computationally feasible approach for predicting large scale conditional volatility and covariance of financial time series. In the case of multi-variant time series, the volatility is represented by a Conditional Covariance Matrix (CCM). Traditional models for predicting CCM such as GARCH models are incapable of dealing with high-dimensional cases as there are O(N ²) parameters to be estimated in the case of N-variant asset return, and it is difficult to accelerate the computation of estimating these parameters by utilizing modern multi-core architecture. These GARCH models also have difficulties in modeling non-linear properties. The widely used Restricted Boltzmann Machine (RBM) is an energy-based stochastic recurrent neural network and its extended model, Conditional RBM (CRBM), has shown its capability in modeling high-dimensional time series. In this paper, we first propose a CRBM-based approach to forecast CCM and show how to capture the long memory properties in volatility, and then we implement the proposed model on GPU by using CUDA and CUBLAS. Experiment results indicate that the proposed CRBM-based model obtains better forecasting accuracy for low-dimensional volatility and it also shows great potential in modeling for large-scale cases compared with traditional GARCH models.     

A feature-free search query classification approach using semantic distance

When classifying search queries into a set of target categories, machine learning based conventional approaches usually make use of external sources of information to obtain additional features for search queries and training data for target categories. Unfortunately, these approaches rely on large amount of training data for high classification precision. Moreover, they are known to suffer from inability to adapt to different target categories which may be caused by the dynamic changes observed in both Web topic taxonomy and Web content. In this paper, we propose a feature-free classification approach using semantic distance. We analyze queries and categories themselves and utilizes the number of Web pages containing both a query and a category as a semantic distance to determine their similarity. The most attractive feature of our approach is that it only utilizes the Web page counts estimated by a search engine to provide the search query classification with respectable accuracy. In addition, it can be easily adaptive to the changes in the target categories, since machine learning based approaches require extensive updating process, e.g., re-labeling outdated training data, re-training classifiers, to name a few, which is time consuming and high-cost. We conduct experimental study on the effectiveness of our approach using a set of rank measures and show that our approach performs competitively to some popular state-of-the-art solutions which, however, frequently use external sources and are inherently insufficient in flexibility.     

Applying option Greeks to directional forecasting of implied volatility in the options market: An intelligent approach

This paper examines movement in implied volatility with the goal of enhancing the methods of options investment in the derivatives market. Indeed, directional movement of implied volatility is forecasted by being modeled into a function of the option Greeks. The function is structured as a locally stationary model that employs a sliding window, which requires proper selection of window width and sliding width. An artificial neural network is employed for implementing and specifying our methodology. Empirical study in the Korean options market not only illustrates how our directional forecasting methodology is constructed but also shows that the methodology could yield a reasonably strong performance. Several interesting technical notes are discussed for directional forecasting.     

Overlay network resource allocation using a decentralized market-based approach

We present a decentralized market-based approach to resource allocation in a heterogeneous overlay network. This resource allocation strategy dynamically assigns resources in an overlay network to requests for service based on current system utilization, thus enabling the system to accommodate fluctuating demand for its resources. Our approach is based on a mathematical model of this resource allocation environment that treats the allocation of system resources as a constrained optimization problem. From the solution to the dual of this optimization problem, we derive a simple decentralized algorithm that is extremely efficient. Our results show the near optimality of the proposed approach through extensive simulation of this overlay network environment. The simulation study utilizes components taken from a real-world middleware application environment and clearly demonstrates the practicality of the approach in a realistic setting.     

Data-abstraction refinement: a game semantic approach

This paper presents a semantic framework for data abstraction and refinement for verifying safety properties of open programs with integer types. The presentation is focused on an Algol-like programming language that incorporates data abstraction in its type system. We use a fully abstract game semantics in the style of Hyland and Ong and a more intensional version of the model that tracks nondeterminism introduced by abstraction in order to detect false counterexamples. These theoretical developments are incorporated in a new model-checking tool, Mage, which implements efficiently the data-abstraction refinement procedure using symbolic and on-the-fly techniques.     

A self-organising mixture autoregressive network for FX time series modelling and prediction

Nowadays a great deal of effort has been made in order to gain advantages in foreign exchange (FX) rates predictions. However, most existing techniques seldom excel the simple random walk model in practical applications. This paper describes a self-organising network formed on the basis of a mixture of adaptive autoregressive models. The proposed network, termed self-organising mixture autoregressive (SOMAR) model, can be used to describe and model nonstationary, nonlinear time series by means of a number of underlying local regressive models. An autocorrelation coefficient-based measure is proposed as the similarity measure for assigning input samples to the underlying local models. Experiments on both benchmark time series and several FX rates have been conducted. The results show that the proposed method consistently outperforms other local time series modelling techniques on a range of performance measures including the mean-square-error, correct trend predication percentage, accumulated profit and model variance.     

A novel recurrent neural network-based prediction system for option trading and hedging

In order to reduce their exposure to the erratic fluctuations of the financial markets, traders are nowadays increasingly dealing with options and other derivative securities instead of directly trading in the underlying assets. This paradigm shift has attracted the attention of many researchers, and there has been a tremendous increase in the awareness and activities of derivative securities. In particular there is a need to devise new techniques to address the limitations of traditional parametric pricing methods, which rely on assumptions and approximations to capture the complex dynamics of pricing processes. This paper proposes a novel non-parametric method using an ad-hoc recurrent neural network for estimating the future prices of war commodities such as gold and crude oil as well as currencies, which are increasingly gaining importance in the financial markets. The price predictions from the network, shown to be accurate and computationally efficient, are used in a hedging system to avoid unnecessary risks. Experiments with actual gold and currency trading data show that our system using the proposed network and strategy can construct portfolios yielding a return on investment of about 4.76%.