運用資料倉儲與資料探勘技術設計與實作漁業資訊分享熱線

Abstract

漁業資訊分享熱線是以網際網路為基礎的資訊系統，其建構的目的是為了探索重要的漁業知識。漁業資訊分享熱線的建構流程，由本質面來看依序可分為三個階段：資料倉儲，線上分析處理與資料探勘。在第一個階段中，定期地將異質性來源的漁業資料，匯集並存入資料倉儲。接下來在線上分析處理的階段，透過表格與圖表的呈現，依需求瀏覽與篩選出特定的的訊息。第三個階段則是利用資料探勘技術，將前個階段所保留的訊息作進一步分析處理。在本論文中，共進行了三種資料探勘的案例。第一個案例使用序列模式分析方法，追蹤高漁獲量漁船的航行軌跡。第二個案例則在給定的準則下，利用二階段分群分析方法找出優良的漁船單位。第三個案則是藉由多變量技術進行漁區與月份對於漁獲情形關係的假設檢定與後續檢定。

An Internet-based software system entitled Fishery Information Sharing Hotline (FISH) is constructed for exploring important fishery knowledge. Essentially the FISH system consists of three sequential stages: data warehouse, On-Line Analytical Processing (OLAP), and data mining. At the first stage, the fishery data from various sources are collected into the data warehouse periodically. Then selected tabular messages or visual diagrams are scanned and screened as required at the OLAP stage. Also the qualified messages will be exported for more advanced processing using data mining techniques. In this thesis, three data mining cases are studied at the final stage. Case one is to trace the voyaging trails of interested vessels using a mining method for searching sequential patterns. For the given criteria, case two is to filter out outstanding vessel groups using a two-stage clustering method. Case three is to test some well-known hypotheses for fisheries production relationships based on fishing grounds and seasons. The research results show the feasibility of FISH approach.