Catálogo de publicaciones - libros

In this paper, an algorithm, called VB-FT-Mine (ectors-ased ault–olerant frequent patterns ), is proposed for mining fault-tolerant frequent patterns efficiently. In this approach, fault–tolerant appearing vectors are designed to represent the distribution that the candidate patterns contained in data sets with fault-tolerance. VB-FT-Mine algorithm applies depth-first pattern growing method to generate candidate patterns. The fault-tolerant appearing vectors of candidates are obtained systematically, and the algorithm decides whether a candidate is a fault-tolerant frequent pattern quickly by performing vector operations on bit vectors. The experimental results show that VB-FT-Mine algorithm has better performance on execution time significantly than FT-Apriori algorithm proposed previously.

Medicine Paring Analysis is one of the most important tasks in the research of Traditional Chinese Medicine Prescriptions. The most essential and difficult step is to mine associations between different medicine items. This paper proposes an effective approach in solving this problem. The main contributions include: (1) proposing a novel data structure called indexed frequent pattern tree (IFPT) to maintain the mined frequent patterns (2) presenting an efficient algorithm called Nearest Neighbor First (NNF) to mine association rules from IFPT (3) designing and implementing two optimization strategies that avoid the examinations of a lot of subsets of that can’t be the left part of any association rule of the form – and thus achieving a wonderful performance and (4) conducting extensive experiments which show that NNF runs far faster than Apriori algorithm and has better scalability. And finally we demonstrate the effectiveness of this method in Medicine Paring Analysis.

The present web is existing in the HTML and XML formats for persons to browse. Recently there is a trend towards the semantic web where the information can be can be processed and understood by agents. Most of the present research works focus on the translation from HTML to semantic web, but seldom on XML. In this paper, we design a method to translate XML to semantic web. It is known that ontologies play an important role in the semantic web, therefore firstly, we propose a genetic model to organize ontologies, and based on this model, we use three steps, viz. semantic translation, structural translation and schematic translation, to translate XML to semantic web. The aim of the paper is to facilitate the wide use of semantic web.

Web pages change frequently and thus crawlers have to download them often. Various policies have been proposed for refreshing local copies of web pages. In this paper, we introduce a new sampling method that excels over other change detection methods in experiment. Change Frequency (CF) is a method that predicts the change frequency of the pages and, in the long run, achieves an optimal efficiency in comparison with the sampling method. Here, we propose a new hybrid method that is a combination of our new sampling approach and CF and show how our hybrid method improves the efficiency of change detection.

We propose a new approach to discover and extract topic-specific hypertext resources from the WWW. The method, called schema driven and topical crawling, allows a user to define schema and extracting rules for a specific domain of interests. It supports automatically search and extract schema-relevant web pages from the web. Different from common approaches that surf solely on web pages, our approach supports crawler to surf on a virtual network composed by concept instances and relationships. To achieve such a goal, we design an architecture that integrates several techniques including web extractor, meta-search engine and query expansion, and provide a toolkit to support it.

In moving object databases, existing disk-based indexes are unable to keep up with the high update rate while providing speedy retrieval at the same time. However, efficient management of moving-object database can be achieved through aggressive use of main memory. In this paper, we propose an (IMPACT) framework where the moving object database is indexed by a pair of indexes based on the properties of the objects’ movement – a main-memory structure manages objects while a disk-based index handles objects. As objects become active (or inactive), they dynamically migrate from one structure to the other. Moreover, the main memory is also organized into two partitions – one for the main memory index, and the other as buffers for the frequently accessed nodes of the disk-based index. Our experimental study shows that the IMPACT framework provides superior performance.

Conventionally, the problem of location updates for moving objects has been addressed by adjusting the location reporting frequency or setting the uncertainty bound, according to object mobility patterns. This induces an obvious tradeoff between the communication cost and the uncertainty bound in querying moving object locations. Most existing works are focused on the object mobility pattern, without exploring the interdependency between queries and location updates. Furthermore, they take the precision of query results for granted as a result of a negotiated deviation threshold for reporting. The Aqua (daptive ery-ware) location updating scheme proposed in this paper exploits the interdependency between queries and updates. In particular, our scheme is adaptive to changes in both object mobility patterns and query characteristics, thereby resulting in significant performance improvement in terms of communication cost and query processing precision. We performed simulation studies and demonstrated that Aqua can produce desirable performance in most situations.

While the volumes of spatial data are tremendous and spatial operations are time-intensive, mobile devices own limited storages and low computational resources. Therefore, a spatial index for mobile map services should be small and efficiently filter out the candidate objects of a spatial operation as well. This paper proposes a spatial index called MHF(Multilevel Hashing File) for the mobile map service. The MHF has a simple structure for storage utilization and uses a hashing technique for search efficiency. This paper also designs a compression scheme of MBR(Minimum Bounding Rectangle) called HMBR. Although the HMBR scheme reduces the volume of MBR to almost a third, it still achieves a good filtering efficiency because of no information loss by quantization in case of small objects that occupy a major portion. Our experimental tests show that the proposed MHF with HMBR is appropriate for mobile devices in terms of the volume of index, the number of the MBR comparisons, the filtering efficiency and the execution time of spatial operations.

This paper introduces a new approach for efficiently indexing and querying constantly evolving data. Traditional data index structures suffer from frequent updating cost and result in unsatisfactory performance when data changes constantly. Existing approaches try to reduce index updating cost by using a simple linear or recursive function to define the data evolution, however, in many applications, the data evolution is far too complex to be accurately described by a simple function. We propose to take each constantly evolving data as a time series and use the ARIMA (Autoregressive Integrated Moving Average) methodology to analyze and model it. The model enables making effective forecasts for the data. The index is developed based on the forecasting intervals. As long as the data changes within its corresponding forecasting interval, only its current value in the leaf node needs to be updated and no further update needs to be done to the index structure. The model parameters and the index structure can be dynamically adjusted. Experiments show that the forecasting interval index (FI-Index) significantly outperforms traditional indexes in a high updating environment.

Spatio-temporal databases offer a rich repository and opportunities to develop techniques for discovering new types of spatio-temporal patterns. In this paper, we introduce a new class of spatio-temporal patterns, called the , to describe the repeated sequences of events that occur within small neighbourhoods. Such patterns are crucial to the understanding of habitual patterns. To discover this class of patterns, we develop an algorithm GenSTMiner based on the idea of pattern growth approach, and introduce some optimization techniques that are used to reduce the number of candidates generated and minimize the size of the projected databases. Our performance study indicates that GenSTMiner is highly efficient and outperforms PrefixSpan.