Peng Wu

Small-Time Scale Network Traffic Prediction Using Complex Network Models

The self-similar and nonlinear nature of network traffic makes high accurate prediction difficult. Various technology, including Autoregressive Integrated Moving Average (ARIMA), Local Approximation (LA), Neural Network (NN) etc., have been applied to internet traffic prediction. In this paper, Complex Network based on genetic programming and particle swarm optimization is proposed to predict the time series of internet traffic.We propose an automatic method for constructing and evolving our complex network model. The structure of complex network is evolved using genetic programming, and the fine tuning of the parameters encoded in the structure is accomplished using particle swarm optimization algorithm. The relative performances of our model are reported. The results show that our model has high prediction accuracy and can characterize real network traffic well.

Function sequence genetic programming

Genetic Programming(GP) can obtain a program structure to solve complex problem. This paper presents a new form of Genetic Programming, Function Sequence Genetic Programming (FSGP). We adopt function set like Genetic Programming, and define data set corresponding to its terminal set. Besides of input data and constants, data set include medium variables which are used not only as arguments of functions, but also as temporary variables to store function return value. The program individual is given as a function sequence instead of tree and graph. All functions run orderly. The result of executed program is the return value of the last function in the function sequences. This presentation is closer to real handwriting program. Moreover it has an advantage that the genetic operations are easy implemented since the function sequence is linear. We apply FSGP to factorial problem and stock index prediction. The initial simulation results indicate that the FSGP is more powerful than the conventional genetic programming both in implementation time and solution accuracy.

MENN Method Applications for Stock Market Forecasting

A new approach for forecasting stock index based on Multi Expression Neural Network (MENN) is proposed in this paper. The approach employs the multi expression programming (MEP) to evolve the architecture of the MENN and the particle swarm optimization (PSO) to optimize the parameters encoded in the MENN. This framework allows input variables selection, over-layer connections for the various nodes involved. The performance and effectiveness of the proposed method are evaluated using stock market forecasting problems and compared with the related methods.

The dynamic mechanism of a novel stochastic neural firing pattern observed in a real biological system

Abstract In the experimental neural pacemaker of a rat, a novel firing pattern has been discovered. This pattern was generated between the period 2 firing pattern and the period 3 firing pattern during the periodic adding bifurcation and inverse periodic adding bifurcation. The pattern was observed and analyzed in the present investigation. The composition of this novel firing pattern could be regarded as a transition between a string of period 2 burst and a string of period 3 burst without single period 2 or period 3 burst, which was different from those chaotic and stochastic neural firing patterns in previous reports. It was identified to be stochastic by the inter-event intervals (IEIs) analysis, although it exhibited chaos-like characteristics with the results of the inter-spike intervals (ISIs) analysis. The numerical simulation suggested that the new pattern observed in the real biological system could be simulated in the stochastic Chay model but not in the deterministic model. With the signal to noise ratio (SNR) analysis and bifurcation analysis, this novel firing pattern was considered to be generated by stochastic resonance under the influence of noise near the periodic adding (inverse) bifurcation point. The probability analysis of transformed binary chain further confirmed that the origin of stochastic and chaos (deterministic)-like characteristics of this novel firing pattern.

Building Ensemble Classifier Based on Complex Network for Predicting Protein Structural Class

In recent years, complex network models were developed to solve classification and time series prediction problems. In this paper, ensemble classifier based on complex network (mainly scale-free network) is firstly used to predict protein structural class. For the classifier design, genetic programming and particle swarm optimization algorithm are used alternately to evolve the structure and encoding parameters. The experimental results validate the good performance of the proposed method.

Using the Wide and Deep Flexible Neural Tree to Forecast the Exchange Rate

Forecasting exchange rate plays an important role in the financial market. It has become a hot research topic and many methods have been proposed. In this paper, a wide and deep flexible neural tree (FNT) is proposed to forecast the exchange rate. The wide component has the function to memorize the original input features, while the deep component can automatically extract unseen features. By balancing the width and depth of flexible neural tree, the structure of FNT is optimized from the experiments to forecast the exchange rate. Experiments have been conducted on four different kinds of exchange rate daily data to check the performance of the FNT. The architecture of the wide and deep FNT is developed by grammar guided genetic programming (GGGP) and the parameters are optimized by the particle swarm optimization algorithm (PSO). Proposed method performs well as compared to the autoregressive moving average model and neural networks.

Dynamical Analysis of a Stochastic Neuron Spiking Activity in the Biological Experiment and Its Simulation by I Na , P + I K Model

An irregular on-off like spiking activity is observed in the rat neural pacemaker experiment with the changes of extracellular calcium concentration. The spiking activity is simulated using a minimal model, the stochastic INa,P + I K model. The nonlinear time series analysis on ISI series shows the similar stochastic dynamical features of both experimental and simulated results. The power spectrum and SNR analysis suggests that this spiking activity is the autonomous stochastic resonance induced by noise near a subcritical Hopf bifurcation. Thus, it becomes easy for us to compare different stochastic firing patterns observed in the same experiment and stimulated under one same model. Besides, some deterministic-like characteristics by the analysis results on ISI series were also explained in this paper.

The Wide and Deep Flexible Neural Tree and Its Ensemble in Predicting Long Non-coding RNA Subcellular Localization

The long non-coding RNA (lncRNA) is a hot research topic among researchers in the field of biology. Recent studies have illustrated that the subcellular localizations carry salient information to understand the complex biological functions. However, the experimental setup cost and the computational cost to identify the subcellular localization of lncRNA is too high. Therefore, there is a need of some efficient and effective methods to predict the lncRNA subcellular locations. In this paper, a wide and deep flexible neural tree (FNT) is proposed to predict the subcellular localization of lncRNA. The wide component has ability to memorize the original input features, while the deep component has ability to automatically extract hidden features. To fully exploit lncRNA sequence information, we have extracted seven features which are further fed to four wide and deep FNT classifiers respectively. By ensemble four classifiers, it can predict 5 subcellular localizations of lncRNA, including cytoplasm, nucleus, cytosol, ribosome and exosome.

Predicting Multisite Protein Sub-cellular Locations Based on Correlation Coefficient

With the development of proteomics and cell biology, protein sub-cellular location has become a hot topic in bioinformatics. As the time goes on, more and more researchers make great efforts on studying protein sub-cellular location. But they only do research on single-site protein sub-cellular location. However, some proteins can belong to two or more sub-cellulars. So, we should transfer the line of sight to multisite protein sub-cellular location. In this article, we use Virus-mPLoc data set and choose pseudo amino acid composition and correlation coefficient two effective feature extraction methods. Then, putting these features into multi-label k-nearest neighbor classifier to predict protein sub-cellular location. The experiment proves that this method is reasonable and the precision reached 68.65% through the Jack-knife test.

Using a Hierarchical Classification Model to Predict Protein Tertiary Structure

To predict protein tertiary structure accurately is helpful for understanding the functions of proteins. In this study, a hierarchical classification method based on flexible neural tree was proposed to predict the structures, in which the tier classifiers were flexible neural trees due to their excellent performances. In order to classify the structures, three types of feature are used, i.e. the tripeptide composed of dimension reduction, the pseudo amino acid composition and the position information of amino acid residues. To evaluate our method, the 640 data set was used in this investigation. The experimental results suggest that our method overwhelms several representative approaches to predicting protein tertiary structure.