Construct Of J2EE-Based Data Mining System And Research On Clustering Technology

With the development of database technology and the enhancement of storage capacity, more and more data has been stored up, it is particularly important to find out the knowledge people real need from these massive data. As a multidisciplinary interdisciplinarity, data mining can find out useful models and rules from massive data and it's an important mean to get knowledge from data. As a research focus of data mining, clustering not only can find data structure as an independent tool but also can be treated as pretreatment of other methods to get better clusters.This paper has designed a web-based data mining system with J2EE technology, and has researched K-Means and DBSCAN clustering algorithm. Here is the main achievement: To meet the needs of B/S-based data mining system, designed a web-based data mining system with handy man-machine interface, which has used popular and useful technology and framework such as Spring Framework. Hibernate and JSF; Researched and implemented K-means and DBSCAN algorithm, and also improved K-means algorithm to correct local convergence when the original algorithm random choose initial central data object; Realized the visualization of data objects and data mining results, which can give users a intuitive impression of data objects and results, and make users convenient to get domain knowledge and be better to understand the results.This system can run in any platform and can be deployed in any J2EE container, it's has good extensibility, easy to use and maintain, and it's very safe. The system can get useful information and knowledge implied in massive data which is incomplete, noisy, fuzzy and random.