A WebGIS Based Data Mining System For Contraceptives

Data mining is the extraction of implicit, previsously unkonown, and potentially usefulinformation from massive data. As a burgeoning interdisciplinary research area, it is graduallydeepening application that plays an increasingly important role in all walks of life, especiallyin terms of decision supporting. As we know, contraceptives contribute to the conscientiousimplementation of the Family Planning Policy of China; meanwhile it helps to guarantee abalanced long-term development of our population. Up to now, the database of contraceptivesmanagement and service platform has accumulated a lot of data characterized with timelinessand localization. Under such circumstances, it is terribly urgent to discover intelligently thepotential valuable information hidden behind the data, in order to duly provide thegovernment with useful information that is regular, instructive and predictive. Theinformation would help the government with its decision-making and improve its servicelevel. As a result, the management of contraceptives in family planning would have a moresustainable, healthy, efficient and coordinated development.This paper centers on the research and implementation of JSMiner, a data mining systemfor contraceptives based on WebGIS. The main contributions are depicted as follows:1. Proposed a data mining process model that is based on the guided visualization. Webuilt a visualization guided data mining process model, which underlines the importance ofvisualization technology in the whole processing, with the analysis and researchs of theclassical data mining models. This idea is running through the JSMiner Project.2. This paper presents an improved algorithm which is based on the Maximal Clique forcredible association rules (here is called MaxC-SQL). In the traditional algorithms, it is fardifficult to determine the support rating because of the skewed distribution datasets, whilehere we introduce another novel conception which is called credible association rules, inwhich each item has the same support level. The confidence directly reflects the credibledegree of the rule. Combined with the advantage of SQL in the relational database, animproved algorithm based on maximal clique named MaxC-SQL is presented, and integrated to JSMiner.3. Multi-perspective and Multi-scaled visualization techniques based on GIS andthematic map are introduced. On account of the characteristics of the data’s gradual increasing,data information’s diversity and complexity, strongly regional relevance, etc. visualizationtechniques based on GIS and thematic map are introduced here, which creates a new methodto support decision-making with more intuitive and diverse presentation means, rather thanthe boring and rigid interface in the traditional text-based systems.4. A stateful dual-mode data mining system called JSMiner is developed. The wholesystem framework consists of three modules, namely, User Authentication Module, WebGISModule and Data Mining Module. RBAC-based authentication ensures the system’s securityand integrity of user information; the design of WebGIS module makes full use of the clusterservices and caching mechanism, etc. which greatly improves the response rate andconcurrency of the system; the standard PMML language as well as the ideas of algorithmmodularization and Agent are applied in the data mining module, which largely boosts thecompatibility and expansibility of the system. At the same time, the system provides twomining interfaces, the common and the professional, which can meet the needs of some userswho have no any data mining background knowledge and those who expert to data mining.Therefore, the application scope of this system is greatly broadened. And the built-in miningtask reasoning mechanism has largely improved the intellectuality of the system. Besides,based on the design of the system, the realization of the system is carried out on thedevelopment platform of VS2008. What’s more, this paper introduces the overalldevelopment plan, the related details of implementation, integrated deployment andoptimization, and so on.5. By introducing some mining results that are gotten in JSMiner’s test running, it canverify that JSMiner is of the feasibility and effectiveness.