Qiliang Liu

An adaptive spatial clustering algorithm based on delaunay triangulation

In this paper, an adaptive spatial clustering algorithm based on Delaunay triangulation (ASCDT for short) is proposed. The ASCDT algorithm employs both statistical features of the edges of Delaunay triangulation and a novel spatial proximity definition based upon Delaunay triangulation to detect spatial clusters. Normally, this algorithm can automatically discover clusters of complicated shapes, and non-homogeneous densities in a spatial database, without the need to set parameters or prior knowledge. The user can also modify the parameter to fit with special applications. In addition, the algorithm is robust to noise. Experiments on both simulated and real-world spatial databases (i.e. an earthquake dataset in China) are utilized to demonstrate the effectiveness and advantages of the ASCDT algorithm.

A general method of spatio-temporal clustering analysis

Spatio-temporal clustering has been a hot topic in the field of spatio-temporal data mining and knowledge discovery. It can be employed to uncover and interpret developmental trends of geographic phenomenon in the real world. However, existing spatio-temporal clustering methods seldom consider both spatiotemporal autocorrelations and heterogeneities among spatio-temporal entities, and the coupling in space and time has not been well highlighted. In this paper, a unified framework for the clustering analysis of spatio-temporal data is proposed, and a novel spatio-temporal clustering algorithm is developed by means of a spatio-temporal statistics methodology and intelligence computation technology. Our method is applied successfully to finding spatio-temporal cluster in China’s annual temperature database for the period 1951–1992.

Multi-level method for discovery of regional co-location patterns

ABSTRACT Regional co-location patterns represent subsets of feature types that are frequently located together in sub-regions in a study area. These sub-regions are unknown a priori, and instances of these co-location patterns are usually unevenly distributed across a study area. Regional co-location patterns remain challenging to discover. This study developed a multi-level method to identify regional co-location patterns in two steps. First, global co-location patterns were detected, and other non-prevalent co-location patterns were identified as candidates for regional co-location patterns. Second, an adaptive spatial clustering method was applied to detect the sub-regions where regional co-location patterns are prevalent. To improve computational efficiency, an overlap method was developed to deduce the sub-regions of (k + 1)-size co-location patterns from the sub-regions of k-size co-location patterns. Experiments based on both synthetic and ecological data sets showed that the proposed method is effective in the detection of regional co-location patterns.

Adaptive detection of spatial point event outliers using multilevel constrained Delaunay triangulation

Abstract Spatial outlier detection is a research hot spot in the field of spatial data mining. Because of the lack of specific research on spatial point events, this study presents an adaptive approach for s patial p oint e vents o utlier d etection (SPEOD) using multilevel constrained Delaunay triangulation. First, the spatial proximity relationships between spatial point events are roughly captured by Delaunay triangulation. Then, three-level constraints are described and used to refine spatial proximity relationships with the consideration of statistical characteristics. Finally, those spatial point events connected by remaining edges are gathered to form a series of subgraphs. Those subgraphs containing very few point events are regarded as spatial outliers. Experiments on both synthetic and real-world spatial data sets are used to show that the proposed SPEOD algorithm can detect various types of spatial point event outliers with high efficiency. Moreover, there is no need to input any parameter in SPEOD.

Modeling the effect of scale on clustering of spatial points

Abstract It has been established that spatial clustering patterns are scale-dependent. However, scale is still not explicitly handled in existing methods to detect clusters in spatial points; thus, users are often puzzled by the varied clustering results obtained by different spatial clustering methods and/or parameters. To handle the effect of scale on the cluster detection of spatial points, two kinds of scales are first specified in this study: scale of data and scale of analysis. These two kinds of scales are embodied by a set of three indictors: data resolution, spatial extent, and analysis resolution. Further, a scale-driven clustering model with these three scale indicators as parameters is statistically constructed based on the Natural Principle and graph theory. A comparative study of this scale-driven clustering model with existing methods is carried out with a simulated spatial dataset. It is found that only this new method is able to discover the multi-scale spatial clustering patterns defined in the benchmarks. Further, Carex lasiocarpa population data is used to illustrate the practical value of the proposed scale-driven clustering model.

Recognizing building groups for generalization: a comparative study

ABSTRACT Recognition of building groups is a critical step in building generalization. To find building groups, various approaches have been developed based on the principles of grouping (or the Gestalt laws of grouping), and the effectiveness of these approaches needs to be evaluated. This study presents a comparative analysis of nine typical such approaches, including three approaches that only consider proximity principle and six approaches that consider multiple grouping principles. Real-life dataset at 1:5000, 1:10,000, and 1:50,000 scales provided by National Geomatics Center of China is used to evaluate the performance of these approaches. Buildings at smaller scales are used to construct the benchmarks to test the grouping results at larger scales, and the adjusted Rand index is adopted to indicate the accuracy of the detected groups. Significant tests (Friedman test and Wilcoxon signed-rank test) are also performed to provide both the overall and pairwise comparisons of these approaches. The results show that (1) the average accuracy of most existing approaches is between 0.3 and 0.5, and the performances of these approaches are significantly different; (2) when only proximity is considered, the buffer analysis approach performs significantly better than other approaches; (3) when multiple grouping principles are considered, the local constraint-based approach usually performs better than other approaches; (4) existing approaches that consider similarity and/or continuity seldom improve the performance of building grouping.

A Framework for Discovering Evolving Domain Related Spatio-Temporal Patterns in Twitter

In massive Twitter datasets, tweets deriving from different domains, e.g., civil unrest, can be extracted to constitute spatio-temporal Twitter events for spatio-temporal distribution pattern detection. Existing algorithms generally employ scan statistics to detect spatio-temporal hotspots from Twitter events and do not consider the spatio-temporal evolving process of Twitter events. In this paper, a framework is proposed to discover evolving domain related spatio-temporal patterns from Twitter data. Given a target domain, a dynamic query expansion is employed to extract related tweets to form spatio-temporal Twitter events. The new spatial clustering approach proposed here is based on the use of multi-level constrained Delaunay triangulation to capture the spatial distribution patterns of Twitter events. An additional spatio-temporal clustering process is then performed to reveal spatio-temporal clusters and outliers that are evolving into spatial distribution patterns. Extensive experiments on Twitter datasets related to an outbreak of civil unrest in Mexico demonstrate the effectiveness and practicability of the new method. The proposed method will be helpful to accurately predict the spatio-temporal evolution process of Twitter events, which belongs to a deeper geographical analysis of spatio-temporal Big Data.

A Spatial Anomaly Points and Regions Detection Method Using Multi-Constrained Graphs and Local Density

Spatial anomalies may be single points or small regions whose non-spatial attribute values are significantly inconsistent with those of their spatial neighborhoods. In this article, a Spatial Anomaly Points and Regions Detection method using multi-constrained graphs and local density (SAPRD for short) is proposed. The SAPRD algorithm first models spatial proximity relationships between spatial entities by constructing a Delaunay triangulation, the edges of which provide certain statistical characteristics. By considering the difference in non-spatial attributes of adjacent spatial entities, two levels of non-spatial attribute distance constraints are imposed to improve the proximity graph. This produces a series of sub-graphs, and those with very few entities are identified as candidate spatial anomalies. Moreover, the spatial anomaly degree of each entity is calculated based on the local density. A spatial interpolation surface of the spatial anomaly degree is generated using the inverse distance weight, and this is utilized to reveal potential spatial anomalies and reflect their whole areal distribution. Experiments on both simulated and real-life spatial databases demonstrate the effectiveness and practicability of the SAPRD algorithm.

An adaptive spatial clustering algorithm based on the minimum spanning tree-like

Spatial clustering is an important means for spatial data mining and spatial analysis, and it can be used to discover the potential rules and outliers among the spatial data. Most existing spatial clustering methods cannot deal with the uneven density of the data and usually require predefined parameters which are hard to justify. In order to overcome such limitations, we firstly propose the concept of edge variation factor based upon the definition of distance variation among the entities in the spatial neighborhood. Then, an approach is presented to construct the minimum spanning tree-like (MST-L). Further, an adaptive MST-L based spatial clustering algorithm (AMSTLSC) is developed in this paper. The spatial clustering algorithm only involves the setting of the threshold of edge variation factor as an input parameter, which is easily made with the support of little priori information. Through this parameter, a series of MST-L can be automatically generated from the high-density region to the low-density one, where each MST-L represents a cluster. As a result, the algorithm proposed in this paper can adapt to the change of local density among spatial points. This property is also called the adaptiveness. Finally, two tests are implemented to demonstrate that the AMSTLSC algorithm is very robust and suitable to find the clusters with different shapes. Especially the algorithm has good adaptiveness. A comparative test is made to further prove the AMSTLSC algorithm better than classic DBSCAN algorithm.

Multi-scale approach to mining significant spatial co-location patterns

Spatial co-location pattern mining aims to discover a collection of Boolean spatial features, which are frequently located in close geographic proximity to each other. Existing methods for identifying spatial co-location patterns usually require users to specify two thresholds, i.e. the prevalence threshold for measuring the prevalence of candidate co-location patterns and distance threshold to search the spatial co-location patterns. However, these two thresholds are difficult to determine in practice, and improper thresholds may lead to the misidentification of useful patterns and the incorrect reporting of meaningless patterns. The multi-scale approach proposed in this study overcomes this limitation. Initially, the prevalence of candidate co-location patterns is measured statistically by using a significance test, and a non-parametric model is developed to construct the null distribution of features with the consideration of spatial auto-correlation. Next, the spatial co-location patterns are explored at multi-scales instead of single scale (or distance threshold) discovery. The validity of the co-location patterns is evaluated based on the concept of lifetime. Experiments on both synthetic and ecological datasets show that spatial co-location patterns are discovered correctly and completely by using the proposed method; on the other hand, the subjectivity in discovery of spatial co-location patterns is reduced significantly.