Advanced Materials Research Vols. 989-994

Abstract: In order to obtain more accurate parameters of support vector machine model, using genetic algorithm to optimize the parameters is an effective method. This paper analyzes the principle of support vector machine for regression, support vector machine kernel function selection, kernel parameters, penalty factor selection and adjustment methods, taking into account genetic algorithm is effective in solving optimization problems, proposed a method using genetic algorithm to optimize the parameters of support vector machine, which uses genetic algorithms to make cross-validation error minimized. The simulation results demonstrate the effectiveness of this method.

Abstract: The problem of robustly globally exponential stability in the mean square is investigated for stochastic uncertain discrete-time bidirectional associative memory (BAM) neural networks with time-varying delays and Markovian jumping parameters. The uncertainties are assumed to be the linear fractional form. By using Lyapunov-Krasovskii functional (LKF) method and some novel technique, a delay-dependent exponential stability criterion is established in terms of linear matrix inequalities (LMIs). A numerical example is provided to show the effectiveness and the improvement of the proposed methods.

Abstract: the correlation coefficient estimation algorithm based on subspace decomposition is presented in this paper. The correlation coefficient between the signals is obtained by getting eigen-value decomposition of the data covariance matrix, and deriving the relationship between signal subspace and noise subspace. Simulation results verify that this algorithm can be realized to get the correlation coefficient between two incident signals whose DOA are known and the effect of the correlation coefficient estimation is made by different signal direction angular intervals.

Abstract: To improve E-Commerce recommendation efficiency for multi-consumer population, coordinative optimization problem for multi-consumer purchase is studied. Multi-objective coordinative optimization model for multi-consumer purchase in E-commerce (MCOMMPE) is established, the optimization objectives consist of maximizing purchase probability, profit, green product orientation; minimizing lost consumer number; keeping equalization among varied purchase schemes. To solve easily, MCOMMPE is transformed into single objective optimization model. Cultural Immune Algorithm (CIA) is studied, affinity degree calculating, clone selecting, antibody population space and belief space are studied in detail, and the basic steps of CIA are designed. The validity of MCOMMPE and its CIA are validated.

Abstract: Parcel Sea Map mapping is currently implemented by manual, which causes the task tedious and error prone. To overcome this problem, this paper proposes an automatic mapping method for Parcel Sea Map. The boundary point data are basic mapping elements, which are obtained from sea cadastral survey. The mapping result can be demonstrated by two aspects: convex polygon and concave polygon. For convex polygon, well known left-turn and convex hull algorithms are adopted. For concave polygon, two algorithms are proposed. One is enumeration and the other is cutting algorithm. The above two mapping results can finally be determined by the inspector from sea cadastral survey. Experiments show that Parcel Sea Map automatic mapping can efficiently reduce the manual burden of present mapping.

Abstract: The traditional network traffic classification methods have many shortcomings, the classification accuracy is not high, the encrypted traffic cannot be analyzed, and the computational burden is usually large. To overcome above problems, this paper presents a new network traffic classification method based on optimized Hadamard matrix and ECOC. Through restructuring the Hadamard matrix and erasing the interference rows and columns, the ECOC table is optimized while eliminating SVM sample imbalance, and the error correcting ability for classification is reserved. The experiments results show that the proposed method outperform in network traffic classification and improve the classification accuracy.

Abstract: Most existing watermark extraction algorithms were dependent on prior knowledge. This paper proposed a blind extraction method without relying on prior knowledge. According to constructing new observation based on nonsubsampled contourlet transform, which utilizes low frequency and directional components of watermarked image, more independent components are generated. We involve these components into watermarked image and resort this solution to multichannel blind source separation. Estimated watermark is recovered by ICA algorithm. Experiment results indicate that the proposed method can achieve better results in contrast with two existing algorithms.

Abstract: With the development of Radio Frequency Identification (RFID) technology, it also exposed some problems in practical applications. Especially in multi-tag anti-collision problem, how to quickly identify multiple tags in the scope of the reader has become the key to solve the problem. On the basis of the Binary Search Algorithm (BSA) and Dynamic Binary Search Algorithm (DBSA), the paper proposed an Improved Dynamic Binary Search Algorithm. The simulation result proves that the improved method in multi-tag anti-collision is more efficiency than the previous two, which is a good solution to the shortcomings of the first two algorithms.

Abstract: The interaction of two four-petal Gaussian beams propagating in strongly nonlocal nonlinear media is investigated. The results show that the intensity evolution of two beams during their propagation is periodical, which is similar to that of a single beam propagation. However the combined optical field of two beams during propagation is more complicated than that of a single beam. During propagation, the two beams are attracted each other, and at the superposed region of two optical fields, the interference fringes appear. The influences of different beam orders and different input powers on the propagation properties are also discussed.

Abstract: With the rapid development of network, texts which contain position, views and opinions of events are exploding. Texts of review contain author’s feelings, views and tendencies the author wants to express. People need to analyze these texts automatically to acquire sentiment tendency of the author. This paper presents a model for automatic text analysis about sentiment tendency on comment text. The model combines algorithms based on emotional dictionary and Support Vector Machine learning algorithm together, which takes advantage of both algorithms.