在台灣活絡的股票市場下，所有投資者皆期待能在眾多類股之中，挑選出值得投資的類股，就股票投資決策而言，投資者會從許多訊息中，找尋可能達到投資獲利的機會。如何利用許多公開資料，轉換成投資時有用的資訊，是市場所有投資者追求的最高目標，因此，很多研究便利用各式方法，例如基本分析法或技術分析法等財務上的研究工具，以期預測與分析股價。 以往許多研究選擇以財務方法作為研究方法，本研究則是運用資料探勘中關聯性法則技術，探討台灣集中市場各類股間關聯性、與國際股價指數之間的關聯性，以及與總體經濟變數之之關聯性，並且運用集群分析方法，將股票市場分成兩群，探討關聯法則與集群分析是否能夠提供相互驗證、或具有哪些互補之處。首先蒐集公開股價指數資料，並建立資料庫，再以資料探勘技術找尋關聯性規則，以建議投資者進行投資可採用投資哪些類股。在資料探勘技術中，本研究選擇關聯法則中的Apriori演算法，以最小支持度、最小可靠度與增益值最為門檻值，尋找關聯法則，佐以集群分析結果相互驗證，希冀能為市場投資者做出更臻完善之建議。分析的結果分為幾大區塊，分別是台灣集中市場各類股之間與加權股價指數之關聯、台灣加權股價指數和各類股與國際股價指數之關聯、集中市場各類股與匯率之關聯，以及運用集群分析下整體市場分群情況。本研究並針對每區塊分析之研究結果，繪出整體概括性的類股地圖，提供股票市場投資者，在選擇投資類股上的參考依據。 What all of the investors expect is to search a worthy investment in the stock market. Investors always want to get all kinds of messages to make decisions of investing and look forward to getting profit. Their final target is to pursue how to transfer the public data into useful information. Many researchers used various kinds of financial means such as basic analysis, technology analysis, and so on, to analyze the stock price. Many of the previous researches used financial tools as research ways over the years. This study used the association rule on Taiwan centralization stock market, international stock index and macroeconomics factor. The research adopted clustering to separate market into two groups. The public data, such as stock indexes, were collected to establish data house. Besides, the technology of data mining was used to find association rule. Then, the researcher tried to find if there are likeness or differences between the clustering way and the association. The Apriori algorithm was adopted to look for the association rule and verified by the clustering analysis. Moreover, the researcher expected the results could offer all investors useful suggestions in the stock market. The results of this study could be divided into several blocks, such as stock indexes in Taiwan centralization stock market, the association of stock indexes between domestic and foreign stock market, the association between Taiwan stock indexes and exchange rate, and the situation of whole market clustering. Finally, the results of the research may provide an overall reference for investors in the stock market.
