Computing top-k results matching XML queries is gaining importance due to the increasing of large XML repositories. In this paper, we propose a novel two-layer-based index construction and associated algorithms for efficiently computing top-k results for SLCA-based XML keyword search. We have conducted expensive experiments and the results show great advantage on efficiency compared with existing approaches.

Mobile communication data analysis has been often used as a background application to motivate many data mining problems. However, very few data mining researchers have a chance to see a working data mining system on real mobile communication data. In this demo, we showcase our new system MobileMiner on a real mobile communication data set, which presents a case study of business solutions using state-of-the-art data mining techniques. MobileMiner adaptively profiles users' behavior from their calling and moving record streams. Customer segmentation and social community analysis can be conducted based on user profiles. We show how data mining techniques can help in mobile communication data analysis. Moreover, we also show some interesting observations which still cannot be mined by the current techniques, and thus may motivate new research and development.

In this paper, we present a useful data modeling methodology in data warehousing which integrates three existing approaches normally used in isolation: goal-driven, data-driven and user-driven. It comprises of four stages. Goal-driven stage produces subjects and KPIs(Key Performance Indicators) of main business fields. Data-driven stage produces subject oriented enterprise data schema. User-driven stage yields analytical requirements represented by measures and dimensions of each subject. Combination stage combines the triple-driven results. By triple-driven, we can get a more complete, more structured and more layered data model of a data warehouse. We illustrate each stage step by step using examples in our case study.

No abstract available.

Integrating relational database technologies into Web information retrieval enables users to ask complex queries beyond traditional keyword searches over Web pages. One approach to this integration is to have a software layer on top of an information retrieval (IR) system and an RDBMS (relational database management system). A core operation in this top layer is to join the intermediate results from the two underlying systems (called the IR results and the DB results correspondingly) in order to produce the final ranked results for each query. Unfortunately, most conventional join algorithms are inefficient for this operation. In this paper, we propose one simple join algorithm called binary search join (BSJ) for the operation of joining the IR results and the DB results. This algorithm takes advantage of the fact that the IR results are already ranked by relevance and that the DB results are already sorted by the join attribute. It scans the IR results and for each IR result tuple performs a binary search over the DB results. We analytically and empirically study the performance of BSJ in comparison with several conventional join algorithms on a repository of Chinese news Web pages. The experiment results prove that BSJ works best in most cases

Spatial data mining is a promising technique that deals with extraction of implicit knowledge or other interesting patterns from large amount of spatial data. Though most data mining systems work with data stored in flat files or operational database, it has been recognized that mining in a data warehouse usually result in better information. Because data are usually cleansed before they are stored into data warehouse. Furthermore, data warehouse provides data with different levels of summarization for the clients, which will lead to fruitful data mining. However, current techniques of data warehouse can not handle spatial data well. Both dimensions and measures in the data model of data warehouse are nonspatial data. In this paper, we propose a new data model called spatial data cube for data warehouse. Spatial data cube supports both spatial and nonspatial data. We also introduce how to construct a spatial data cube that can answer queries efficiently by selective materialization. We believe that the spatial data cube can provide better support for spatial data mining.

As WWW becomes more and more popular and powerful, how to search information on the web in database way becomes an important research topic. COMMIX, which is developed in the DB group in Peking University (China), is a system towards building very large database using data from the Web for information extraction, integration and query answering. COMMIX has some innovative features, such as ontology-based wrapper generation, XML-based information integration, view-based query answering, and QBE-style XML query interface.

In this paper, a novel algorithm is proposed to tackle multi-class classification problems. For a K-class classification task, an array of K optimal pairwise coupling classifiers (O-PWC) is constructed, each of which is optimal to the corresponding class and provides a reliable probability estimation for that class. The classification accuracy rate is improved while the computational cost does not increase too much. At the same time, a more accurate estimation of posterior probabilities for a given pattern can be acquired. This algorithm is applied to face recognition on an ORL face database. Experimental results reveal that our method is effective and efficient.

The classification of underwater acoustic signals is one of the important fields of pattern recognition. Inspired by the experience of training human experts in sonar, we propose a two-phase training algorithm to exploit association rules to reveal understandable intrinsic rules which contribute to correct classification in known mis-classification data sets. Preliminary experimental results demonstrate the potential of these classification association rules to enhance the classification accuracy of underwater acoustic signals.

Methods for efficient mining of frequent patterns have been studied extensively by many researchers. However, the previously proposed methods still encounter some performance bottlenecks when mining databases with different data characteristics, such as dense vs. sparse, long vs. short patterns, memory-based vs. disk-based, etc. In this study, we propose a simple and novel hyper-linked data structure, H-struct and a new mining algorithm, H-mine, which takes advantage of this data structure and dynamically adjusts links in the mining process. A distinct feature of this method is that it has very limited and precisely predictable space overhead and runs really fast in memory-based setting. Moreover it can be scaled up to very large databases by database partitioning, and when the data set becomes dense, (conditional) FP-trees can be constructed dynamically as part of the mining process. Our study shows that H-mine has high performance in various kinds of data, outperforms the previously developed algorithms in different settings, and is highly scalable in mining large databases. This study also proposes a new data mining methodology, space-preserving mining, which may have strong impact in the future development of efficient and scalable data mining methods.