透過資料倉儲挖掘台灣颱風與大宗花卉變化的關聯性法則

Abstract

本文提出使用資料倉儲於台灣颱風與大宗花卉之間的關聯性法則探勘。每年颱風導致台灣的花卉產業重大的損失。然而，在花卉方面的損失與颱風之間的關聯性並不是如此明顯。主要的原因是為了分析這層關聯背後所要準確蒐集花卉資料的困雖度，因此包括了許多高成本的風險，例如如何從不同的來源去蒐集花卉資料；如何判斷正確的花卉資料；如何將有用的花卉資料正確分類使其能夠進行統計分析…等。這份論文中採用資料倉儲系統解決上述艱難問題。換句話說，倉儲中的花卉資料來源於五個作為當地花卉批發市場交易資料目的地的資料超市。因此，資料蒐集的過程僅僅只需一個線上分析處理資料下載程序。接著運用Apriori演算法於台灣颱風與大宗花卉之間的關聯性法則探勘。並發現有58條種關聯性法則在其中。

Data warehouse has been proposed for mining association rules between typhoons and main flowers in Taiwan. Typhoons bring huge damages in flower industry in Taiwan every year. However, the relation between flower damages and typhoons is not so clear. One major reason is the difficulty of collecting the right flower data for analyzing the relation because it includes many high cost tasks such as how to collect the flower data from the various sources, how to justify the correct flower data, how to sort out the valid flower data into the right category for statistical use, etc. In this thesis, a data warehouse was implemented for solving this hard challenge. In other words, the flower data in the data warehouse come from the five data marts which are distributed data achieves for storing transactional data in local flower wholesale markets. As a result, the data collection process becomes as a simple On-Line Analytical Processing (OLAP) data download procedure. Then Apriori algorithm was applied to mining the association rules between typhoons and major flowers. Also 58 association rules are found.