Big label: categorizing the Web efficiently and accurately

Searching on the Web has never been an easy task. Even if semantic information is successfully inferred from a user query, how can we benefit from it? The most popular remedy today is to categorize the Web in advance. By gathering similar Web resources into a group, the search performance should increase even though search engines still have little idea about the semantics part. To categorize a set of Web resources according to meta-information associated with them, at first, one has to analyze the relationships between meta-information and Web resources. However, the result will be severely affected by the ambiguous nature of the Web. As a result, the goal of this research is to propose a new labeling method to enhance both the efficiency and accuracy of Web resources categorization.     

Applying hierarchical grey relation clustering analysis to geographical information systems - A case study of the hospitals in Taipei City

Deng proposed grey clustering analysis (GCA) in 1987. Later, Jin presented a new method in 1993, called grey relational clustering (GRC) method that combined grey relational analysis with clustering. However, the GRC method cannot use a tree diagram to make appropriate classification decisions without re-computation. This study thus attempts to combine GRC and hierarchical clustering analysis. Given the existence of an excess of medical resources in the Taipei area, this study attempts to understand the degree of concentration of medical resources in this area. Specifically, this study applies a geographical information system (GIS) to present the geographical distribution of hospitals in Taipei. Additionally, a new-type of cluster analysis, known as hierarchical grey relation clustering analysis, is used to analyze the distribution of hospitals and understand how they compete with one another. The analytical results demonstrate that hierarchical grey relation clustering analysis is a suitable method of analyzing geographical position. Tree diagrams can help policymakers make appropriate classification decisions without re-computation. The study results can inform hospitals of their competitors and help them to develop appropriate responses. Additionally, the analytical results can also provide a reference to government or hospital policymakers to help them position hospitals in areas, thus achieving a better distribution of medical resources in Taipei.     

分布式工作流管理系统中的一种资源搜索策略

针对分布式工作流管理系统中的资源搜索问题，提出一种新的搜索策略。该方法利用聚类算法将网分簇管理，使得查找过程中的逻辑拓扑和物理拓扑相互匹配，在此基础上采用结合了兴趣度的蚁群算法提供查找线索。实验表明，该方法提高了工作流引擎服务的整体响应时间，在查找成功率、平均响应时间和平均消息数上较泛洪查找和传统蚁群算法都更具优势。     

高光谱图像深度模糊核聚类的洋底锰结核识别

目的 海洋资源调查是海洋科学研究的重要组成部分，对于开发利用海洋资源、保护海洋环境有重要意义。将深海高光谱图像用于调查洋底锰结核资源，相比传统图像方法信息更为全面，识别更为准确。但是高光谱方法中用于有监督分类识别的分类器需要人工标定的标签，这在深海环境中是较为困难的。针对这一局限性，本文提出了一种融合深度网络与模糊核聚类的深度模糊核聚类（deep kernel fuzzy C-means，DKFCM）算法，实现对洋底锰结核高光谱图像的无监督聚类。方法 DKFCM由随机深度卷积网络（Rdnet）及改进的模糊核聚类算法两大模块组成。Rdnet通过降维、随机图块卷积及非线性激活操作的循环，实现对高光谱图像浅层及深层特征的提取，融合这两类特征作为后续聚类识别的输入。改进的模糊核聚类算法先用欧氏距离计算初始聚类中心，再用模糊核聚类的方法以实现海洋资源的准确分类。结果 实验结果表明，DKFCM无监督聚类能有效聚类洋底资源，对锰结核的聚类准确率达到76.59%，相比单用K-means聚类提高了20.99%，相比经Rdnet提取特征后再用K-means聚类提高了13.76%，对比实验表明DKFCM算法在无标签数据的情况下也能达到良好的准确率。结论 本文所提的高光谱深度模糊核聚类方法，实现了深海锰结核的无监督聚类，可以用于海洋资源量的评估。     

Classification: Methods for the Exploratory Analythi of Multivariate Data

A fuzzy clustering strategy is used to identify subsets of influential observations in regression. As part of the fuzzy clustering strategy, the analyst can explore the uniqueness of selected subsets and the degree of belonging of observations to selected subsets. This is accomplished by either varying a fuzzy parameter or the number of clusters. Once the observations or subsets have been identified, the analyst can then compute regression diagnostics to confirm their degree of influence in regression. Bootstrapping and high-breakdown procedures were used to confirm the influence of the previously identified subsets. This fuzzy clustering strategy is applied to the modified data on wood-specific gravity and an augmented production dataset. Both datasets have been previously presented in the literature.     

Partitioning road networks using density peak graphs: Efficiency vs. accuracy

Road traffic networks are rapidly growing in size with increasing complexities. To simplify their analysis in order to maintain smooth traffic, a large urban road network can be considered as a set of small sub-networks, which exhibit distinctive traffic flow patterns. In this paper, we propose a robust framework for spatial partitioning of large urban road networks based on traffic measures. For a given urban road network, we aim to identify the different sub-networks or partitions that exhibit homogeneous traffic patterns internally, but heterogeneous patterns to others externally. To this end, we develop a two-stage algorithm (referred as FaDSPa) within our framework. It first transforms the large road graph into a well-structured and condensed density peak graph (DPG) via density based clustering and link aggregation using traffic density and adjacency connectivity, respectively. Thereafter we apply our spectral theory based graph cut (referred as α-Cut) to partition the DPG and obtain the different sub-networks. Thus the framework applies the locally distributed computations of density based clustering to improve efficiency and the centralized global computations of spectral clustering to improve accuracy. We perform extensive experiments on real as well as synthetic datasets, and compare its performance with that of an existing road network partitioning method. Our results show that the proposed method outperforms the existing normalized cut based method for small road networks and provides impressive results for much larger networks, where other methods may face serious problems of time and space complexities.     

Multivariate methods using mixtures: Correspondence analysis,scaling and pattern-detection

Matrices of binary or count data are modelled under a unified statistical framework using finite mixtures to group the rows and/or columns. These likelihood-based one-mode and two-mode fuzzy clusterings provide maximum likelihood estimation of parameters and the options of using likelihood ratio tests or information criteria for model comparison. Geometric developments focused on pattern detection give likelihood-based analogues of various techniques in multivariate analysis, including multidimensional scaling, association analysis, ordination, correspondence analysis, and the construction of biplots. Illustrative examples demonstrate the effectiveness of these visualisations for identifying patterns of ecological significance (e.g. abrupt versus slow species turnover).     

Computing point estimates from a non-Gaussian posterior distribution using a probabilistic k-means clustering approach

The Kalman filter algorithm gives an analytical expression for the point estimates of the state estimates, which is the mean of their posterior distribution. Conventional Bayesian state estimators have been developed under the assumption that the mean of the posterior of the states is the ‘best estimate’. While this may hold true in cases where the posterior can be adequately approximated as a Gaussian distribution, in general it may not hold true when the posterior is non-Gaussian. The posterior distribution, however, contains far more information about the states, regardless of its Gaussian or non-Gaussian nature. In this study, the information contained in the posterior distribution is explored and extracted to come up with meaningful estimates of the states. The need for combining Bayesian state estimation with extracting information from the distribution is demonstrated in this work.     

Agriculture satellite image segmentation using a modified artificial Hopfield neural network

Beekeeping plays an important role in increasing and diversifying the incomes of many rural communities in Kingdom of Saudi Arabia. However, despite the region’s relatively good rainfall, which results in better forage conditions, bees and beekeepers are greatly affected by seasonal shortages of bee forage. Because of these shortages, beekeepers must continually move their colonies in search of better forage. The aim of this paper is to determine the actual bee forage areas with specific characteristics like population density, ecological distribution, flowering phenology based on color satellite image segmentation. Satellite images are currently used as an efficient tool for agricultural management and monitoring. It is also one of the most difficult image segmentation problems due to factors like environmental conditions, poor resolution and poor illumination. Pixel clustering is a popular way of determining the homogeneous image regions, corresponding to the different land cover types, based on their spectral properties. In this paper Hopfield neural network (HNN) is introduced as Pixel clustering based segmentation method for agriculture satellite images.