Mining in Spatio-Temporal Databases

Recent interest in spatio-temporal applications has been fueled by the need to discover and predict complex patterns that occur when we observe the behavior of objects in the three-dimensional space of time and spatial coordinates. Although the complex and intrinsic relationships among the spatio-temporal data limit the usefulness of conventional data mining techniques to discover the patterns in the spatio-temporal databases, they also lead to opportunities for mining new classes of patterns in spatio-temporal databases. This chapter provides a survey of the work done for mining patterns in spatial databases and temporal databases, and the preliminary work for mining patterns in spatio-temporal databases. We highlight the unique challenges of mining interesting patterns in spatio-temporal databases. We also describe two special types of spatio-temporal patterns: location-sensitive sequence patterns and geographical features for location-based service patterns.

Applications of Moving Objects Databases

Miniaturization of computing devices and advances in wireless communication and sensor technology are some of the forces propagating computing from the stationary desktop to the mobile outdoors. Some important classes of new applications that will be enabled by this revolutionary development include location-based services, tourist services, mobile electronic commerce and digital battlefield. Some existing application classes that will benefit from the development include transportation and air traffic control, weather forecasting, emergency response, mobile resource management and mobile workforce. Location management, that is, the management of transient location information, is an enabling technology for all these applications. Location management is also a fundamental component of other technologies, such as fly-through visualization, context awareness, augmented reality, cellular communication and dynamic resource discovery. Moving Objects Databases (MODs) store and manage the location as well as other dynamic information about moving objects. In this chapter we will present the applications of MODs and their functionality. The target readership is researchers and engineers working in databases and mobile computing.

Simple and Incremental Nearest-Neighbor Search in Spatio-Temporal Databases

The efficient processing of nearest-neighbor queries in databases of moving objects is considered very important for applications such as fleet management, traffic control, digital battlefields and more. Such applications have been rapidly spread due to the fact that mobile computing and wireless technologies nowadays are ubiquitous. This chapter presents important aspects towards simple and incremental nearest-neighbor search for spatio-temporal databases. More specifically, we describe the algorithms that have already been proposed for simple and incremental nearest neighbor queries and present a new algorithm regarding that issue. Finally, we study the problem of keeping a query consistent in the presence of insertions, deletions and updates of moving objects.

Indexing Multi-Dimensional Trajectories for Similarity Queries

With the abundance of low-cost storage devices, a plethora of applications that store and manage very large multi-dimensional trajectories (or time-series) datasets have emerged recently. Examples include traffic supervision systems, video surveillance applications, meteorology and more. Thus, it is becoming essential to provide a robust trajectory indexing framework designed especially for performing similarity queries in such applications. In this regard, this chapter presents an indexing scheme that can support a wide variety of (user-customizable) distance measures while, at the same time, it guarantees retrieval of similar trajectories with accuracy and efficiency.

Spatio-Temporal Prediction Using Data Mining Tools

The spatio-temporal prediction problem requires that one or more future values be predicted for time series input data obtained from sensors at multiple physical locations. Examples of this type of problem include weather prediction, flood prediction, network traffic flow, and so forth. In this chapter we provide an overview of this problem, highlighting the principles and issues that come to play in spatio-temporal prediction problems. We describe some recent work in the area of flood prediction to illustrate the use of sophisticated data mining techniques that have been examined as possible solutions. We argue the need for further data mining research to attack this difficult problem. This chapter is directed toward professionals and researchers who may wish to engage in spatio-temporal prediction.

Quadtree-based Image Representation and Retrieval

This chapter is a survey of quadtree uses in the image domain, from image representation to image storage and content-based retrieval. A quadtree is a spatial data structure built by a recursive decomposition of space into quadrants. Applied to images, it allows representing image content, compacting or compressing image information, and querying images. For 13 years, numerous image-based approaches have used this structure. In this chapter, the authors underline the contribution of quadtree in image applications.

Integrating Web Data and Geographic Knowledge into Spatial Databases

With the phenomenal growth of the World Wide Web, rich data sources on many different subjects have become available online. Some of these sources store daily facts that often involve textual geographic descriptions. These descriptions can be perceived as indirectly georeferenced data – for example, addresses, telephone numbers, zip codes and place names. In this chapter we focus on using the Web as an important source of urban geographic information and propose to enhance urban Geographic Information Systems (GIS) using indirectly georeferenced data extracted from the Web. We describe an environment that allows the extraction of geospatial data from Web pages, converts them to XML format and uploads the converted data into spatial databases for later use in urban GIS. The effectiveness of our approach is demonstrated by a real urban GIS application that uses street addresses as the basis for integrating data from different Web sources, combining the data with high-resolution imagery.

Management of Large Moving Objects Databases

This chapter deals with several important issues pertaining to the management of moving objects datasets in databases. The design of representative benchmarks is closely related to the formal characterization of the properties (that is, distribution, speed, nature of movement) of these datasets; uncertainty is another important aspect that conditions the accuracy of the representation and therefore the confidence in query results; finally, efficient index structures, along with their compatibility with existing softwares, is a crucial requirement for spatio-temporal databases, as it is for any other kind of data.

Spatial Joins

This chapter describes algorithms, cost models and optimization techniques for spatial joins. Joins are among the most common queries in Spatial Database Management Systems. Due to their importance and high processing cost, a number of algorithms have been proposed covering all possible cases of indexed and non-indexed inputs. We first describe some popular methods for processing binary spatial joins and provide models for selectivity and cost estimation. Then, we discuss evaluation of multiway spatial joins by integrating binary algorithms and synchronous tree traversal. Going one step further, we show how analytical models can be used to combine the various join operators in optimal evaluation plans. The chapter can serve as a comprehensive reference text to the researcher who wants to learn about this important spatial query operator and to the developer who wants to include spatial query processing modules in a Database System.

Approximate Computation of Distance-Based Queries

In spatial database applications the similarity or dissimilarity of complex objects is examined by performing distance-based queries (DBQs) on data of high dimensionality (a generalization of spatial data). The R-tree and its variations are commonly cited as multidimensional access methods that can be used for answering such queries. Although the related algorithms work well for low-dimensional data spaces, their performance degrades as the number of dimensions increases (dimensionality curse). To obtain acceptable response time in high-dimensional data spaces, algorithms that obtain approximate solutions can be used. In this chapter, we review the most important approximation techniques for reporting sufficiently good results quickly. We focus on the design choices of efficient approximate DBQ algorithms that minimize the response time and the number of I/O operations over tree-like structures. The chapter concludes with possible future research trends in the approximate computation of DBQs.