A Component-Based Coordination Visual Analysis System For Spatiotemporal Data

Nowadays, massive sets of data are collected and stored. Effective using these data andmining interesting knowledge is becoming urgent needs. However, facing these large andcomplex data set, traditional data analysis method seems powerless. Visual analytics is thescience of analytical reasoning facilitated by interactive human-machine interfaces, thatcombining the analytic capabilities of the computer with the abilities of the human perceptionand cognition. It has become a new research focus, because of this new data analysis way.In the paper, a component-based visual analytics system with multiple coordinated views ispresented, for exploring and analyzing complex spatiotemporal data, and finding hiddenspatio-temporal and multivariate patterns. Hangzhou real estate dataset is chosen as an exampledataset for analytics. This dataset is a typical spatiotemporal data, with characteristics such ashouse spatial distribution, average price and number of sales changing pattern, correlation ofhouse attributes, etc. Our system provides an effective and interactive analytics method, usingvisualization.Four core visual analytics components are developed: GeoMap, PCP (Parallel CoordinatePlot), TimeSeries and Treemap. GeoMap provides spatial view; PCP provides multi-dimensionalview; TimeSeries provides temporal view; and Treemap provides synthetical view.In implementation, each component smoothly blends clustering, interaction, color scheme,and a variety of ways, for enhancing visual analytics ability. GeoMap is builded by tile Pyramid,and we introduce map marker clustering algorithm avoiding visual overload. PCP is used tovisualise high-dimensional data that more than four dimensions. This component enhancestraditional defined parallel coordinates, by implementing brushing, histogram, axes controlling,color mapping. TimeSeries is implemented with dual-scale navigation in Overview+Detail way.This component supports multiple granularity of time aggregation and a variety of chart types.Treemap, combining with the four layout algorithms, is implemented with interactiveconfiguration visualization parameters. This component can comprehensively reflect themultidimensional and spatial compositions of spatio-temporal data.In order to avoid components being isolated each other in function, all components are efficiently linked in an extensible and customizable way, by implementing efficient multiplecoordinated views. These components complement each other in function, and the whole systemanalytics ability is enhanced.