Search

Summary jsEvRBF is a time-series forecasting method based on genetic algorithm and neural nets. Written in JavaScript language, can be executed in most web browsers. Consequently, everybody can participate in the experiments, and scientists can take advantage of nowadays available browsers and devices as computation environments. This is also a great challenge as the language support and performance varies from one browser to another. In this paper, jsEvRBF has been tested in a volunteer computing experiment, and also in a single-browser one. Both experiments are related to forecasting currencies exchange, and the results show the viability of the proposal.

This paper describes the coevolutionary algorithm L-Co-R (Lags COevolving with Radial Basis Function Neural Networks - RBFNs), and analyzes its performance in the forecasting of time series in the short, medium and long terms. The method allows the coevolution, in a single process, of the RBFNs as the time series models, as well as the set of lags to be used for predictions, integrating two genetic algorithms with real and binary codification, respectively. The individuals of one population are radial basis neural networks (used as models), while sets of candidate lags are individuals of the second population. In order to test the behavior of the algorithm in a new context of a variable horizon, 5 different measures have been analyzed, for more than 30 different databases, comparing this algorithm against six existing algorithms and for seven different prediction horizons. Statistical analysis of the results shows that L-Co-R outperforms other methods, regardless of the horizon, and is capable of predicting short, medium or long horizons using real known values.

This paper introduces Lags COevolving with Rbfns (L-Co-R), a coevolutionary method developed to face time-series forecasting problems. L-Co-R simultaneously evolves the model that provides the forecasted values and the set of time lags the model must use in the prediction process. Coevolution takes place by means of two populations that evolve at the same time, cooperating between them; the first population is composed of radial basis function neural networks; the second one contains the individuals representing the sets of lags. Thus, the final solution provided by the method comprises both the neural net and the set of lags that better approximate the time series. The method has been tested across 34 different time series datasets, and the results compared to 6 different methods referenced in literature, and with respect to 4 different error measures. The results show that L-Co-R outperforms the rest of methods, as the statistical analysis carried out indicates.

This paper presents the implementation of a time series forecasting algorithm, jsEvRBF, that uses genetic algorithm and neural nets in a way that can be run in must modern web browsers. Using browsers to run forecasting algorithms is a challenge, since language support and performance varies across implementations of the JavaScript virtual machine and vendor. However, their use will provide a boost in the number of platforms available for scientists. jsEvRBF is written in JavaScript, so that it can be easily delivered to and executed by any device containing a web-browser just accessing an URL. The experiments show the results yielded by the algorithm over a data set related to currencies exchange. Best results achieved can be effectively compared against previous results in literature, though robustness of the new algorithm has to be improved.

This work analyzes the behavior and effectiveness of the L-Co-R method using a growing horizon to predict. This algorithm performs a double goal, on the one hand, it builds the architecture of the net with a set of RBFNs, and on the other hand, it sets a group of time lags in order to forecast future values of a time series given. For that, it has been used a set of 20 time series, 6 different methods found in the literature, 4 distinct forecast horizons, and 3 distinct quality measures have been utilized for checking the results. In addition, a statistical study has been done to confirms the good results of the method L-Co-R.

This paper shows how E-tsRBF deals with time-series prediction in a changing horizon environment. E-tsRBF is a meta-evolutionary algorithm that simultaneously evolves both the neural networks and the set of lags needed to forecast time series. The method uses radial basis function neural networks, a kind of net that has been successfully applied to time series prediction in literature. Frequently, methods to build and train these networks must be given the past periods or lags to be used in order to create patterns and forecast any time series. Up to twenty-one time series are evaluated in this work, showing the behaviour of the new method.

In this paper a study of two approaches of a meta-algorithm, Meta_CHC_RBF, is presented. The main goal of this algorithm is to automatically design Radial Basis Function Networks (RBFNs) finding a suitable configuration of parameters (automatically adapted to every problem) necessary for the algorithm EvRBF, an evolutionary algorithm for the automatic design of asymmetric RBFNs. The principal difference between two proposals is the type of codification, in the fist one, the meta-algorithm uses binary codification, while in the second one, it implements real codification; affecting this influence of the codification kind in the carried out experimentation. Finally, results show that the first approach yields good marks reducing the computation time, with respect the second one.

This work introduces SymbPar, a parallel meta-evolutionary algorithm designed to build Radial Basis Function Networks minimizing the number of parameters needed to be set by hand. Parallelization is implemented using independent agents to evaluate every individual. Experiments over classifications problems show that the new method drastically reduces the time took by sequential algorithms, while maintaining the generalization capabilities and sizes of the nets it builds.

Radial basis function neural networks have been successfully applied to time series prediction in literature. Frequently, methods to build and train these networks must be given the past periods or lags to be used in order to create patterns and forecast any time series. This paper introduces E-tsRBF, a meta-evolutionary algorithm that evolves both the neural networks and the set of lags needed to forecast time series at the same time. Up to twenty-one time series are evaluated in this work, showing the behavior of the new method.

In this work SymbPar, a parallel co-evolutionary algorithm for automatically design the Radial Basis Function Networks, is proposed. It tries to solve the problem of huge execution time of Symbiotic_CHC_RBF, in which method are based. Thus, the main goal of SymbPar is to automatically design RBF neural networks reducing the computation cost and keeping good results with respect to the percentage of classification and net size. This new algorithm parallelizes the evaluation of the individuals using independent agents for every individual who should be evaluated, allowing to approach in a future bigger size problems reducing significantly the necessary time to obtain the results. SymbPar yields good results regarding the percentage of correct classified patterns and the size of nets, reducing drastically the execution time.

This paper introduces Symbiotic_CHC_RBF, a co-evolutionary algorithm intended to automatically establish the parameters needed to design models for classification problems. Co-evolution involves two populations, which evolve together by means of a symbiotic relationship. One of the populations is the method EvRBF, which provides the design of radial basis function neural nets by means of evolutionary algorithms. The second population evolves sets of parameters for the method EvRBF, being every individual of the population a configuration of parameters for the method. Results show that Symbiotic_CHC_RBF can be effectively used to obtain good models, while reducing significantly the number of parameters to be fixed by hand.

Radial basis function networks (RBFNs) have shown their capability to be used in classification problems, so that many data mining algorithms have been developed to configure RBFNs. These algorithms need to be given a suitable set of parameters for every problem they face, thus methods to automatically search the values of these parameters are required. This paper shows the robustness of a meta-algorithm developed to automatically establish the parameters needed to design RBFNs. Results show that this new method can be effectively used, not only to obtain good models, but also to find a stable set of parameters, available to be used on many different problems.

Increasing the usability of traditional methods is one of the key issues on future trends in data mining. Nevertheless, most data mining algorithms need to be given a suitable set of parameters for every problem they face, thus methods to automatically search the values of these parameters are required. This paper introduces two co-evolutionary algorithms intended to automatically establish the parameters needed to design radial basis function neural networks. Results show that both algorithms can be effectively used to obtain good models, while reducing significantly the number of parameters to be fixed at hand.

One of the most important issues that must be taken in mind to optimize the design and the generalization abilities of trained artificial neural networks (ANN) is the architecture of the net. In this paper Symbiotic_RBF is proposed, a method to do automatically the process to design models for classification using symbiosis. For it, there are two populations who evolve together by means of coevolution. One of the populations is the method EvRBF, which provides the design of radial basis function neural nets by means of evolutionary algorithms. The second population evolves sets of parameters for the method EvRBF, being every individual of the population a configuration of parameters for the method. Thus, the main goal of Symbiotic_RBF is to find a suitable configuration of parameters necessary for the method EvRBF, which is adapted automatically to every problem.