Diploma thesis is about creating server application that process and interprets data from the database. Main aim of such application is able to process a large number of database requests in real-time environment. Provided database contains records of measuring download speed and quality of mobile connection via different radio technology from various providers. Those measured data are sent from users all around the world and amount of data collected is still growing. Therefore created server application can adapt to increasing size of database thanks to aggregation. This method of aggregation and use of index in database tables are further discussed in the theoretical part. Mainly putting indexes in tables produce significant acceleration of processing database requests. Final product of this thesis is an application that consist of three components: a server application running aggregation, website that interprets measured data and back-end interface providing measured data as well. Data at the website are presented in form of graphs for different countries and used radio technologies. Web address and user manual for finished applications are provided in the fourth chapter of diploma thesis. In the last part of thesis are performed various speed tests of programmed application that confirm the effectiveness of selected and described methods to accelerate work with the database.

資料發掘目前在傳統關聯式資料庫相關議題上已有不少研究，但如果能再整合空間和時間要素進來，將可從資料中發掘出更明確、更具體的知識。以往常使用統計分析方法來分析空間資料，不幸的是，統計分析方法仍有許多問題亟待解決。而Han等人利用概念樹發掘「多層次關聯規則」的技術已相當成熟，值得學習。在時間方面，另外有學者提出「週期關聯規則」的觀念。於是本研究便想結合以上研究的優點，希望能創造出新的應用。 本研究試著將「空間特性」和「週期關聯規則」結合，提出「空間性週期關聯規則」的想法。首先從相關文獻中分別瞭解目前空間、時間資料發掘領域的研究現況，從而整合相關研究，提出研究架構。再以動態網頁技術配合假想的台北市便利商店交易資料庫，發展出一套雛型系統（目前只能作單一項目之間的關聯），以驗證本架構的可行性。最後提出進一步的研究建議，以供後續研究參考。 / There have been a lot of research about data mining in relational database. We can mine more specific and concrete knowledge in transaction databases by further considering spatial and temporal dimension. Until now the statistical spatial analysis has been one common technique for analyzing spatial data. However , there are still many remaining problems. Han et al. used concept hierarchies to mine multiple-level association rules. Their ideas are great and worth our learning. On the other hand , some scholars proposed the notion of cyclic association rules. Therefore , we combine the merits of these researches to discover more meaningful knowledge. In this research , we try to integrate the ideas of spatial associations with cyclic association and propose the idea of spatial cyclic association rules. First , we survey these researches in the fields of spatial and temporal data mining. A framework is then proposed. Finally , we implement a prototype system in WWW ( 1-itemset and 2-itemset only now).