Catálogo de publicaciones - libros

In order to mine one typical non-sphere cluster, the linear belts in a spatial database, a mathematical morphological operator based method is proposed in this paper. The method can be divided into two basic steps: firstly, the most suitable re-segmenting scale is found by our clustering algorithm MSCMO which is based on mathematical morphological scale space; secondly, the segmented result at this scale is re-segmented to obtain the final linear belts. This method is a robust mining method to semi-linear clusters and noises, which is validated by the successful extraction of seismic belts.

Driven by the issue of geo-spatial data mining under grid computing environment, a new representation method called spatial information multi-grid (SIMG) for depicting spatial data and spatial information is presented in this paper. A strategy of dividing the globe with multi-level spatial grid is proposed. And through further studying on the precision of description on feature of detail, this paper tries to divide the globe with SIMG and constructs the framework of SIMG in China. Based on SIMG this paper tries to realize data mining of different thematic information on the same geographical position, data mining of dynamic spatial-temporal information of the same thematic content, data mining and transforming of different coordinate systems and to make SIMG a fundamental research work for further developing spatial information sharing and data mining.

A uniform framework of 3D spatial data model (UF-3DSDM) is proposed in this paper. It is a union of various 3D spatial data models, and any actual 3D spatial data model is its subset. Based on the UF-3DSDM, the stratum modeling method of QPTV is further developed on the real borehole sample data. In the context of 3D QTPV model, a data mining method is given on surface DEM of 3D data.

Standard land price is an economical indicator for measuring land value. In this paper, we propose to use the tension spline interpolation function to mine standard land price. First, we extend the definition of standard land price, which is based on land region composed of several neighboring land parcels with the same or similar features. Second, the regional factors that affect the standard land price are classified into the geometric features of point, line and area according to the quantitative rules. Third, a tension spline interpolation function is proposed to mine standard land price, which is determined by the influential factors. Finally, as a case study, the proposed method is applied to mine land prices for Nanning City in China. The case study shows that the proposed method is a practical and satisfactory one.

We propose a novel approach for mining recent frequent itemsets. The approach has three key contributions. First, it is a single-scan algorithm which utilizes the special property of suffix-trees to guarantee that all frequent itemsets are mined. During the phase of itemset growth it is unnecessary to traverse the suffix-trees which are the data structure for storing the summary information of data. Second, our algorithm adopts a novel method for itemset growth which includes two special kinds of itemset growth operations to avoid generating any candidate itemset. Third, we devise a new regressive strategy from the attenuating phenomenon of radioelement in nature, and apply it into the algorithm to distinguish the influence of latest transactions from that of obsolete transactions. We conduct detailed experiments to evaluate the algorithm. It confirms that the new method has an excellent scalability and the performance illustrates better quality and efficiency.

Online mining of changes from data streams is an important problem in view of growing number of applications such as network flow analysis, e-business, stock market analysis etc. Monitoring of these changes is a challenging task because of the high speed, high volume, only-one-look characteristics of the data streams. User subjectivity in monitoring and modeling of the changes adds to the complexity of the problem.

This paper addresses the problem of i) capturing user subjectivity and ii) change modeling, in applications that monitor frequency behavior of item-sets. We propose a three stage strategy for focusing on item-sets, which are of current interest to the user and introduce metrics that model changes in their frequency (support) behavior.

The three main requirements for clustering data streams on-line are one pass over the data, high processing speed, and consuming a small amount of memory. We propose an algorithm that can fulfill these requirements by introducing an incremental grid data structure to summarize the data streams on-line. In order to deal with high-dimensional problems, the algorithm adopts a simple heuristic method to select a subset of dimensions on which all the operations for clustering are performed. Our performance study with a real network intrusion detection stream data set demonstrates the efficiency and effectiveness of our proposed algorithm.