研究論文有以專案成本來探討專案的績效，或是探討人力資源或人力派遣在專案上的最佳化運用，也有專門探討專案總成本的預測等；但少有研究論文直接以工時成本、毛利及總成本來探討專案執行績效；也鮮少有研究論文是以此方式得出的專案執行績效來探討專案特性之間的關聯性。 利用資料探勘技術的概念，並根據所處理問題之特性，選用適當的演算技術，進行智慧型資料分析，尋找大量資料內隱含之邏輯規則與資訊，挖掘出有價值的知識；利用Mircosoft資料採礦軟體之關聯分析演算法作為研究方法。本研究以歷史工程案例作為研究的目標，將每個案例中所預設的欄位名稱如已花工時、預算工時、合約金額、總成本、預算毛利設定為專案輸入參數，結案毛利設定為可預測項目，另以專案代號做為索引條件，將所有的參數的區間分佈情形做一統計，從其中找出有關聯的規則並探討之。 藉由關聯規則輸出，嘗試找出參數間的規則，並與工程案例的毛利值做一分析比較，進而歸類出專案毛利在14%是個明顯的參考指標，也會有因不同性質的專案工程，而會有不同的評價標準；利用此規則來深入探討公共工程類、石化工程類及高科技產業等的專案執行績效分布情形。 In a recessional market, small sized projects become one of main incomes for construction related companies. As known from literature,project feature involving complexity and uniqueness certainly affects project performance. In the view of project profit earned, man-hour spent,and cost accruals, this thesis examines the relationship between project performance and feature using data mining, generating association rules that can be used for practitioners to improve their performance results.Data collection aims at relatively small sized construction projects. A total of 202 projects are collected and as such 81% of their project size is less than NT$10 million. Two-third of the total spent less than 1000 man-hours.Using the Apriori algorithm, numerous association rules are discovered. The major findings conclude that (1) budgeting a small project with less than 14% profit margin has relatively high chances to win a bid; (2) performing relatively small construction projects has 55% chance of experiencing cost overrun; and (3) projects are more controllable if their contract prices are less than NT$8 million.
