Deren Li

Data Field for Hierarchical Clustering

In this paper, data field is proposed to group data objects via simulating their mutual interactions and opposite movements for hierarchical clustering. Enlightened by the field in physical space, data field to simulate nuclear field is presented to illuminate the interaction between objects in data space. In the data field, the self-organized process of equipotential lines on many data objects discovers their hierarchical clustering-characteristics. During the clustering process, a random sample is first generated to optimize the impact factor. The masses of data objects are then estimated to select core data object with nonzero masses. Taking the core data objects as the initial clusters, the clusters are iteratively merged hierarchy by hierarchy with good performance. The results of a case study show that the data field is capable of hierarchical clustering on objects varying size, shape or granularity without user-specified parameters, as well as considering the object features inside the clusters and removing the outliers from noisy data. The comparisons illustrate that the data field clustering performs better than K-means, BIRCH, CURE, and CHAMELEON.

Cloud Model-Based Spatial Data Mining

Abstract In spatial data mining, we have to deal with uncertainties in data and mining process. The nature of the uncertainties can be, for example, fuzziness and randomness. This paper proposed a cloud model-based data mining method that may simultaneously deal with randomness and fuzziness. First, cloud model is presented, which is described by using three numerical characteristics. Ex, En and He. Furthermore, three visualization methods on cloud model are further proposed, which can be produced by the cloud generators. Second, cloud model-based knowledge discovery is further developed. In cloud model context, spatial data preprocessing pays more attention to data cleaning, transform between qualitative concepts and quantitative data, data reduction, and data discretization. Spatial uncertain reasoning is in the form of linguistic antecedents and linguistic consequences, both of which are implemented by X-conditional and Y-conditional cloud generators. Spatial knowledge is represented with qualitative concepts from large amounts of data, and also the cloud model. Finally, as an example, these methods are applied to mine Baota landslide monitoring database. The experimental results show that the cloud model can not only reduce the task complexity, and improve the operational efficiency, but also enhance the comprehension of the discovered knowledge.

Partially supervised classification – based on weighted unlabeled samples support vector machine

This paper addresses a new classification technique: partially supervised classification (PSC), which is used to identify a specific land-cover class of interest from a remotely sensed image by using unique training samples belong to a specifically selected class. This paper also presents and discusses a novel Support Vector Machine (SVM) algorithm for PSC. Its training set includes labeled samples belong to the class of interest and unlabeled samples of all classes randomly selected from a remotely sensed image. Moreover, all unlabeled samples are assumed to be training samples of other classes and each of them is assigned a weighting factor indicating the likelihood of this assumption; hence, the algorithm is so-called ‘Weighted Unlabeled Sample SVM (WUS-SVM). Experimental results with both simulated and real data sets indicate that the proposed PSC method is more robust than 1-SVM and has comparable accuracy to a standard SVM.

An algebraic algorithm for point inclusion query

Abstract Linking number is a basic invariant of a closed curve in topology. In this paper, linking number is generalized in a computational way to reflect the spatial relationship between a point and a line segment, an arc, a plane subdivision, a three-dimensional shape, a three-dimensional object and a three-dimensional space subdivision. An algebraic algorithm for the point-in-polygon test is devised on the base of the generalized linking number. The algorithm is characterized by the algebraic property that the linking number of a point to a polygon is merely the arithmetic sum of the linking numbers of the point to all the line segments of the polygon. The algorithm takes intersection into consideration, but no intersection is actually computed. It calculates the linking number, but no trigonometric function is processed. Detailed time complexity analysis shows that this algorithm greatly improves currently existing algorithms for the point-in-polygon test. Another achievement of this paper is the consistent and successful extension of the algebraic algorithm for the point-in-polygon test to point inclusion queries in two-dimensional objects, plane subdivisions, three-dimensional shapes, three-dimensional objects and three-dimensional space subdivisions. All these algorithms retain the algebraic character.

Rough Spatial Interpretation

Rough set is a new approach to uncertainties in spatial analysis. In this paper, we complete three works under the umbrella of rough space. First, a set of simplified rough symbols is extended on the basis of existing rough symbols. It is in terms of rough interpretation and specialized indication. Second, rough spatial entity is proposed to study the real world as it is, without forcing uncertainties to change into a crisp set. Third, rough spatial topological relationships are studied by using rough matrix and their figures. The relationships are divided into three types, crisp entity and crisp entity (CC), rough entity and crisp entity (RC) and rough entity and rough entity (RR). A universal intersected equation is further developed. Finally, rough membership function is further extended with the gray scale in our case study. And the maximum and minimum maps of river thematic classification are generated via the rough membership function and rough relationships.

[Agrave] Trous Wavelet Decomposition Applied to Image Edge Detection

Abstract Detecting image edge is considered as a key step in many complicated processing methods such as image segmentation, image recognition, and feature extraction. Many methods of detecting image edges have been developed, but almost every method has its restriction in application of image processing. In this paper, disadvantages and advantages of some classic methods for image edge detection are thoroughly discussed. Base on the analysis, a new À trous wavelet decomposition algorithm is applied to detecting image edge. From the experimental results, we can find that the edges detected by our À trous wavelet decomposition method are better than those processed by the classic Sobel, and Robert methods. In addition, when the original image is stained by noise, the new method almost is not disturbed, on the contrary, the classic algorithms are sensitive to noise. However, besides the advantages of the new method of detecting image edge, we also find out a shortcoming, the disadvantage need to further research for improving the result.

Virtual tang-style timber-frame building complex

Timber-frame building is one of gems of Chinese ancient buildings. Protecting and researching historic buildings by using computer technologies has become a hot topic and urgent task in the field of digital cultural heritage, with increasingly disappearing the timber-frame buildings and their architectural arts and crafts. This paper introduces constructing an information system based on virtual reality technology of timber-frame buildings, and addresses important technologies involved in reconstructing ancient buildings, such as data collection, 3D reconstruction, data organization and management, virtual assembly, and virtual roaming. An integrated solution is proposed and provides a brand-new method for 3D reconstruction, documentation, repair, research and protection of timber-frame buildings.

A Try for Handling Uncertainties in Spatial Data Mining

Uncertainties pervade spatial data mining. This paper proposes a method of spatial data mining handling randomness and fuzziness simultaneously. First, the uncertainties in spatial data mining are presented via characteristics, spatial data, knowledge discovery and knowledge representation. Second, the aspects of the uncertainties in spatial data mining are briefed. They often appear simultaneously, but most of the existing methods cannot deal with spatial data mining with more than one uncertainty. Third, cloud model is presented to mine spatial data with both randomness and fuzziness. It may also act as an uncertainty transition between a qualitative concept and its quantitative data, which is the basis of spatial data mining in the contexts of uncertainties. Finally, a case study on landslide-monitoring data mining is given. The results show that the proposed method can well deal with randomness and fuzziness during the process of spatial data mining.

Software and applications of spatial data mining

Most big data are spatially referenced, and spatial data mining (SDM) is the key to the value of big data. In this paper, SDM are overviewed in the aspects of software and application. First, spatial data are summarized on their rapid growth, distinct characteristics, and implicit values. Second, the principles of SDM are briefed with the descriptive definition, fundamental attributes, discovery mechanism, and usable methods. Third, SDM software is presented in the context of software components, developing methodology, typical software for geographical information system (GIS) data and remote sensing (RS) images, and software trend. Fourth, SDM applications are outlined on GIS data, RS image, and spatiotemporal video data. The final is the concluding remarks and perspectives. WIREs Data Mining Knowl Discov 2016, 6:84–114. doi: 10.1002/widm.1180

δ-Open set clustering-A new topological clustering method

Clustering is an unsupervised learning method widely used for identifying the inherent data structure and applied to various fields such as data mining, patter recognition, machine learning, and others. A new topological clustering method called δ‐open set clustering is proposed in this study. The key idea of this method is to determine δ‐open sets in data, for which each δ‐open set represents one specific category of data. It is shown that this method has robust performance even for complex data set. It can classify the complex type of data sets coming with diverse shapes, recognize noise and deal with data set of high dimensionality. This method is effective even when the distribution of data is unbalanced. In the clustering process, one requires a single input parameter, namely the value of δ. A face identification experiment on the Olivetti Face Database indicates that this method performs much more reliably than the peak clustering method. We also provide another improved δ‐open set clustering that makes δ‐open set clustering capable of handling clusters with extreme density difference.