Search

Support vector clustering (SVC) is an important boundary-based clustering algorithm in multi applications. But SVC's popularity is degraded by its pricy computation and poor labeling performance. Different from existing modifications that only resolve one of two bottlenecks, this paper presents an improved SVC, iSVC, to address two bottlenecks simultaneously. iSVC's contributions are as follows: (1) It includes a reduction strategy that can help to develop clustering model on a qualified subset. The reduction strategy is based on the Schrodinger equation to find the crucial data towards model formulation. (2) The original objective is modified; it cooperates with the reduction strategy to produce the model with subtle loss of quality. (3) iSVC employs a new label approach to label data according to the geometric properties of feature space. The new approach labels data in a simple but effective way without suffering from the randomness originated in the old algorithm. (4) The geometric property is proofed to guarantee the new labeling approach's validation. Theoretical analysis and empirical evidence suggest that iSVC overcomes two bottlenecks well. And when compared with some common clustering methods, it does a good job in performance and efficiency, which opens a broad way of applications for SVC.

A novel learning schema SVCMR based on support vector is proposed in this paper to address M-class classification issue. It creates a tree-shaped decision frame where M/2 nodes are constructed with the three-separation model as the basic classifier. A class selection rule is defined to ensure basic classifiers be trained in turn on pair of classes with maximum feature distance. Class contours are extracted as data representatives to reduce training set size. Another point is that parameters involved in SVCMR are learned from data neighborhood, which brings adaptation to various datasets and avoids pricy cost spent on searching parameter spaces. Experiments on real datasets demonstrate the performance of SVCMR can be competitive to those state-of-the-art classifiers but with the higher effectiveness than them.

This paper presents a new neural network predictive model named PCA - DRKM - RBF, which combines Principal Component Analysis (PCA), Dynamic Rough set K-means (DRKM) and Radial Basis Function (RBF) neural network. The data processed by the principal component analysis is the neural network’s input, and the RBF neural network’s hidden nodes are the centers using DRKM. The paper forecasts the cyclodextrin closure constant, and the results indicate that the model has obviously improved the accuracy of prediction.

One of the main objectives in the analysis of microarray experiments is the identification of genes that are differentially expressed under two experimental conditions. Many analysis methods for identifying differentially expressed (DE) genes were developed. Here we compared three analysis methods which are the significance analysis of microarrays (SAM), the t-based meta-analysis (TB) and the rank product meta-analysis (RP). We found out strengths and weaknesses of the three analysis methods with respected to three measures, which we referred to as efficiency, stringency, and ability to handle heterogeneity. The TB is more efficient than the SAM and the RP. In addition, the RP is the best analysis method of the three with other two measures. We concluded that meta-analysis is a powerful tool for identifying differentially expressed genes.

Aiming at the prevalently concerned mining problem about constructing concept search in current Web search engine area, especially applying the Vector Space Model VSM to web search mining based on the association rules, this paper provides a highly efficient mining algorithm EARS. The EARS algorithm implements the association rules pruning based on VSM via constructing association library and computing similarity. The EARS stores the frequent itemsets via multi-dimensional linked lists and utilizes the recurrence relation among the frequent itemsets in order to effectively obtain the association rules. We verify the extended function of retrieving and querying on the web and the results of these experiments indicate that our method is effective and feasible.

Support vector regression is based on statistical learning theory under the framework of a new general-purpose machine learning method, which is a effective way to deal with nonlinear classification and nonlinear regression. Due to the comprehensive theoretical basis and excellent learning performance, The technology has become the current international machine learning research community hot spots, which can to better address the practical problem, such as the small sample and high dimension, nonlinear and local minima etc.. In the article, support vector regression (SVR) and the RBF neural network do function fitting tests, using simulation data, and the results are compared and evaluation. And use the SVR algorithm to solve practical problems in the area of real estate for predict housing values, with a view to consumers in the choice of housing to provide good guidance.

ERA is the acronym of environment, reactive rules and agent; the model is a new effective multi-agent cooperation framework, which has been successfully applied in a wide range of areas. Multi-port Container Stowage (MPCS) is a more practical problem. However, multi-port makes the container stowage problem more difficult. This paper presents a MPCS model based on time series and a multi-agent ERA model based on belief to solve the multi-port container stowage problem. The algorithm introduces the belief concept into agents and creates a multi-agent ERA model. The searching direction and strength of the agents are determined by the learning machine using belief parameter. Agent has the ability to evaluate its searching path. At last, the feasibility and efficiency of the model are validated by the experimental result.

Invariance transformation (IT) is a rewarding technique to facilitate classification. But it is often difficult to derive its definition. This paper derives a local invariance transformation definition from SVM decision function. The corresponding IT-distance definition is consequently designed in both input space and feature space. And a classification algorithm based on IT and Nearest Neighbor rule is proposed, named as ITNN. ITNN exploits hyper sphere centers as class prototypes and labels data using a weighted voting strategy. ITNN is of computational ease brought by training dataset reduction and hyper parameter self-tuning. We describe experimental evidence of classification performance improved by ITNN on real datasets over state of the arts.

The major problem existed in current E-Learning is the difficulty of resources sharing and searching, which is mostly caused by the weak-semantic learning resources and the poor sharing mechanism. By applying ontology and Web services technology to E-Learning, this paper proposes a sharing individualized E-Learning architecture, where the meaning of learning resources with well-defined is presented by ontology and information exchanging is supported by Web services based on SOA.

This paper presents a Multi-Classification Schema (MCS) which combines Weighted SVMs (WSVM) and Spectrum-based kNN (SkNN). Basic SVM is equipped with belief coefficients to reveal its capacity in identifying classes. And basic SVM is built in individual feature space to bring adaptation to diverse training data context. Coupled with a weighted voting strategy and a local informative metric, SkNN is used to address the case rejected by all basic classifiers. The local metric is derived from most discriminant directions carried by data spectrum information. Two strategies of MCS benefit computational cost: training dataset reduction, and pre-specification of SkNN working set. Experiments on real datasets show MCS improves classification accuracy with moderate cost compared with the state of the art.

A Spectrum-based Support Vector Algorithm (SSVA) to resolve semi-supervised classification for relational data is presented in this paper. SSVA extracts data representatives and groups them with spectral analysis. Label assignment is done according to affinities between data and data representatives. The Kernel function encoded in SSVA is defined to rear to relational version and parameterized by supervisory information. Another point is the self-tuning of penalty coefficient and Kernel scale parameter to eliminate the need of searching parameter spaces. Experiments on real datasets demonstrate the performance and efficiency of SSVA.

A Parameterized Semi-supervised Classification algorithm based on Support Vector (PSCSV) for multi-relational data is presented in this paper. PSCSV produces class contours with support vectors, and further extracts center information of classes. Data is labeled according to its affinity to class centers. A novel Kernel function encoded in PSCSV is defined for multi-relational version and parameterized by supervisory information. Another point is the self learning of penalty parameter and Kernel scale parameter in the support-vector-based procedures, which eliminates the need to search parameter spaces. Experiments on real datasets demonstrate performance and efficiency of PSCSV.

The microeconomic framework for data mining assumes that an enterprise chooses a decision maximizing the overall utility over all customers. In item selection problem, the store wants to select J item set S that maximizes the overall profit. Based on the microeconomic view, we propose a novel algorithm ItemRank to solve the problem of item selection with the consideration of cross-selling effect which has two major contributions. First, we propose customer behavior model, and demonstrate it with the data of customer-oriented business. Second, we propose the novel algorithm ItemRank which is implemented on the basis of customer behavior model. According to the cross-selling effect and the self-profit of items, ItemRank algorithm could solve the problem of item order objectively and mechanically. We conduct detailed experiments to evaluate our proposed algorithm and experiment results confirm that the new methods have an excellent ability for profit mining and the performance meets the condition which requires better quality and efficiency