Search

The support vector machine (SVM) is a class of powerful classification tools that have many successful applications. Their classification results usually belong to either one class or the other. But in many real-world applications, each data point no more exactly belongs to one of the two classes, it may 70% belong to one class and 30% to another. That is, there is a fuzzy membership associated with each data. In such an environment, fuzzy SVM (FSVM), which treats every sample as both positive and negative classes with the fuzzy membership, were introduced. In this way the FSVM will have more generalization ability, while preserving the merit of insensitive to outliers. Although the FSVM has good generalization capability, the computational complexity of the existing FSVM is rather large because the final solution is obtained from solving a quadratic programming (QP) problem. For reducing the complexity, this study proposes a least squares method to solve FSVM. In the proposed model, we consider equality constraints instead of inequalities for the classification problem with a formulation in a least squares sense. As a result the solutions follow directly from solving a set of linear equations instead of QP thus reducing the computational complexity greatly relative to the classical FSVM. For illustration purpose, a real-world credit risk assessment dataset is used to test the effectiveness of the LS-FSVM model. [ABSTRACT FROM AUTHOR]

We compare the predication performance of neural networks with the different frequencies of input data, namely daily data, weekly data, monthly data. In the 1 day and 1 week ahead prediction of foreign exchange rates forecasting, the neural networks with the weekly input data performs better than the random walk models. In the 1 month ahead prediction of foreign exchange rates forecasting, only the special neural networks with weekly input data perform better than the random walk models. Because the weekly data contain the appropriate fluctuation information of foreign exchange rates, it can balance the noise of daily data and losing information of monthly data. [ABSTRACT FROM AUTHOR]

In this study, a hybrid synergy model integrating exponential smoothing and neural network is proposed for financial time series prediction. The proposed model attempts to incorporate the linear characteristics of an exponential smoothing model and nonlinear patterns of neural network to create a "synergetic" model via the linear programming technique. For verification, two real-world financial time series are used for testing purpose. [ABSTRACT FROM AUTHOR]

This paper proposes a new method for crude oil price forecasting based on support vector machine (SVM). The procedure of developing a support vector machine model for time series forecasting involves data sampling, sample preprocessing, training & learning and out-of-sample forecasting. To evaluate the forecasting ability of SVM, we compare its performance with those of ARIMA and BPNN. The experiment results show that SVM outperforms the other two methods and is a fairly good candidate for the crude oil price prediction. [ABSTRACT FROM AUTHOR]

We propose a new computational method of input selection for stock market forecasting with neural networks. The method results from synthetically considering the special feature of input variables of neural networks and the special feature of stock market time series. We conduct the experiments to compare the prediction performance of the neural networks based on the different input variables by using the different input selection methods for forecasting S&P 500 and NIKKEI 225. The experiment results show that our method performs best in selecting the appropriate input variables of neural networks. [ABSTRACT FROM AUTHOR]

We represent research themes by the temporal topic profiles based on MeSH terms from MEDLINE citations. By comparing the differences of the temporal profiles for the same topic at the two different periods, we find that the temporal profiles for a topic at the new period may result from three kinds of concepts replacements of the temporal profiles at the old period, namely broad replacement, parallel replacement and narrow replacement. Such findings provide new ways to generate potential relationships from the current biomedical literature. [ABSTRACT FROM AUTHOR]

In this study, a multi-agent web text mining system on the grid is developed to support enterprise decision-making. First, an individual intelligent learning agent that learns about underlying text documents is presented to discover the useful knowledge for enterprise decision. In order to scale the individual intelligent agent with the large number of text documents on the web, we then provide a multi-agent web text mining system in a parallel way based upon grid technology. Finally, we discuss how the multi-agent web text mining system on the grid can be used to implement text mining services. [ABSTRACT FROM AUTHOR]

Many methodologies have been introduced to deal with project portfolio selection problem including some techniques that help to evaluate individual projects, or to select a portfolio among available projects. This paper, however, provides several XML-based schemes for building efficient and flexible project portfolio selection systems based on an integrated portfolio selection model proposed. In additional to provide an interactive and adaptable portfolio selection management framework for the organizations, we also stress the use of XML to standardize the data representation for developing Webbased decision support systems with heterogeneous data sources and multiple decision models. [ABSTRACT FROM AUTHOR]

This study presents a neural network & web-based decision support system (DSS) for foreign exchange (forex) forecasting and trading decision, which is adaptable to the needs of financial organizations and individual investors. In this study, we integrate the back-propagation neural network (BPNN)- based forex rolling forecasting system to accurately predict the change in direction of daily exchange rates, and the Web-based forex trading decision support system to obtain forecasting data and provide some investment decision suggestions for financial practitioners. This research reveals the structure of the DSS by the description of an integrated framework, and meantime we find that the DSS is integrated, user-oriented by its implementation, and practical applications reveal that this DSS demonstrates very high forecasting accuracy and its trading recommendations are reliable. [ABSTRACT FROM AUTHOR]

In this study, a novel hybrid AI system framework is developed by means of a systematic integration of artificial neural networks (ANN) and rulebased expert system (RES) with web-based text mining (WTM) techniques. Within the hybrid AI system framework, a fully novel hybrid AI forecasting approach with conditional judgment and correction is proposed for improving prediction performance. The proposed framework and approach are also illustrated with an example here. [ABSTRACT FROM AUTHOR]

The performance of financial forecasting with neural networks dependents on the particular training set. We design mean-change-point test to divide the original dataset into different training sets. The experiment results show that the larger training set does not necessarily produce better forecasting performance. Although the original datasets are different, the change-points to produce the optimal training sets are close to each other. We can select the suitable size of training set for financial forecasting with neural networks based on the mean-change-point test. [ABSTRACT FROM AUTHOR]

In this study, a hybrid intelligent data mining methodology, genetic algorithm based support vector machine (GASVM) model, is proposed to explore stock market tendency. In this hybrid data mining approach, GA is used for variable selection in order to reduce the model complexity of SVM and improve the speed of SVM, and then the SVM is used to identify stock market movement direction based on the historical data. To evaluate the forecasting ability of GASVM, we compare its performance with that of conventional methods (e.g., statistical models and time series models) and neural network models. The empirical results reveal that GASVM outperforms other forecasting models, implying that the proposed approach is a promising alternative to stock market tendency exploration. [ABSTRACT FROM AUTHOR]

Reserve price auctions are one of hot research topics in the traditional auction theory. Here we study the starting price in an online auction, counterpart of the public reserve price in a traditional auction. By considering three features of online auctions, the stochastic entry of bidders (subject to Poisson process), the insertion fee proportional to the starting price, and time discount, we have analyzed the properties of extremum points of the starting price for maximizing seller's expected revenue, and found that, under certain conditions, the optimal starting price should be at the lowest allowable level, which is contrary to results from the classic auction theory and finds its optimality in reality. We have also developed a general extended model of multistage auction and carried out analysis on its properties. At last, some directions for further research are also put forward. [ABSTRACT FROM AUTHOR]

Scientific bibliographies in online databases provide a rich source of information for scientists in support of their research. In this paper, we propose a new method to predict possible relations between a starting, known concept of interest and other concepts by mining scientific literature databases like Medline. The central novel feature of our method is to predict new relations based on finding brothers of middle concepts within the concept hierarchy. The method can help researchers explore new research directions from current scientific literature. [ABSTRACT FROM AUTHOR]