各國政府為提高國際競爭優勢，紛紛積極推動「電子化政府」。我國電子化政府建設自八十六年起開始推動，迄今已經行政院擴大為e-Taiwan計畫。電子化政府推動之業務電腦化，帶動政府業務資訊系統的快速發展，其彙集而成之大型資料庫，為政府統計工作帶來莫大的發展契機。 本研究從電子化政府的過程、內政業務行政程序、知識挖掘及採勘方法，提出參考資料模型，可能的統計軟體工具以及電子化政府中知識發現的實驗架構。再者，本研究藉臺閩地區外籍與大陸配偶結婚登記資料集，運用各種群集分析如K-means、ANN、TwoStep等，並利用我國人口數時間序列採用多模式方法進行人口預測，並將前述分析結果回饋資料庫，最後，作者實現一個知識發現系統雛型，其中包含了前端資料庫、資料集、知識庫以及EIS使用介面。 本研究成果總結如下：(1)資料挖掘工作產出之知識，除真實呈現社會現象外，亦作為政府政策之指南；(2)在本研究發展之系統中，新興資料挖掘技術及傳統資料分析方法，二者相輔相成；(3)某些資料挖掘技術適合相符的資料型態，例如文中人口預測資料較適合指數平滑法勝於ANN，亦即，我們可以籍由多模式分析比較其結果，來達到更佳的效果；(4)藉由知識庫模型的建立達成知識創造、共享與管理的目標；(5)資料挖掘工作可以回饋改善資訊系統或業務缺失。 / In order to enhance international competitive advantages, most of the government authorities over the world are engaging in realizing their e-Government platforms. The ROC Government began to develope its e-Government- Infrastructure since 1997, and up-to-date is expanding the e-Taiwan Project as a whole by Executive Yuan. The computerization of administration processes within various government agencies push forward fast development of administration information systems via handling administrative works and lead to utmost opportunities for the government statistics by means of very large databases. Starting from a survey on developements of e-Government, administrative processes for interior affairs, and knowledge mining as well as discovery techniques, this study brings out reference data models, potential statistical softwaretools, and an experimental framework as a whole for knowledge discovery in the context of e-Government. In the next step, this study experiments with applying clustering techniques such as K-means, ANN, and Twostep on datamart regarding marriage of foreigners ( including citizens from Mainland China ) in Taiwan, and with employeeing multi-modes approach on population forecasting. The results of aforementioned analysises are feed into backend database. At last, this author carries out a prototype of knowledge discovery system which includes front-end data base, data marts, knowledge base and interfaces to EIS. The results of the research can be summarized as following: 1.Knowledge derived by means of data mining is capable to represent social events / affairs as well as to serve as a kind of guideline for developing government ploicies. 2. The modern data-ming techniques and classical data-analysis approaches complement with each other in the system developed in this research. 3. Certain mining technique is suitable of corresponding data pattern, for example, expotential smoothing is more suitable for our population data than ANN, which means that we may often achieve better result by multi-mode analysis and comprison with the outputs of different modes. 4. Knowledge creation, sharing, and management can be achieved by means of the knowledge discovery processes on the framework developed in this research. 5. We can figure out errorful raw data in the mining output and feedback to the data source to improve its quality.