在資料探勘的技術中,關聯規則是最常被用來描述消費者的消費習慣。利用關聯分析找尋規則時,影響規則有效性的主要因素來自於最小支持度,若給定的最小支持度值過高時,則有些重要的訊息將會被掩蓋掉,相反地,若給定的支持度太低則會挖掘出過多無意義的規則,反而造成太多雜訊的困擾。過往,在最小支持度之訂定均以決策者之主觀決定,因此,在本研究中將藉由成本之觀念客觀地建立最小支持度。在本論文中,採用BSM演算法建構出主要支持度之關聯規則,再根據所得到的關聯規則評估其對商品數量、管理、行銷等方面可能產生之影響。將其受影響之參數值代入研究中所建立之成本評估模式,計算其支持度下的成本、利潤。將各支持度所得到的成本、利潤繪製一趨勢圖,以此找出適宜之支持度。商品之採購量與折扣率為模式中考慮之變數,探討對於成本之影響。在實驗驗證下發現,改變每條規則所負擔的成本或是改變商品損壞率亦或是改變兩者,對於最小支持度訂定並無明顯差別。但當採購品發生採購頻率較小者若採用低的最小支持度則所得到的利潤愈大。而採購頻率較大者,所採用較高之最小支持度,所得到的利潤值較大;再者,考慮改變可享有折扣之採購量或享有之折扣亦或是同時改變兩者之情況時,對於選取最小支持度並無太大地改變。在不同的折扣率之採購數量中,可發現當該數量較小時,對採購品之採購頻率為低者之影響最大,相對地,當該數量較大時,對採購品之採購頻率較高者之影響最大。
The association rule is one of the technologies in data mining which is used to describe the consumptive habits of the consumers. When using association rule to seek the consumptive habits, the minimum supportability is a major factor to affect the decision rules. A larger minimum supportability is chosen, the less important information is extracted. On the contrary, a smaller minimum supportability is preferred, the plenty useless rules are shown. The decision will be disordered when there are plenty of noise rules appeared. In this thesis, we propose a novel approach to determine a suitable value of minimum supportability by evaluating the cost model. In this study, the BSM algorithm is used to find out the association rules based on the discussed supportability, and then based on these association rules, we evaluate the impact on the number of the products, management and sales, finally, compute the cost by plugging the suitable values of variables and parameters into the cost model. The minimum support value can be solved by minimizing a cost function of supportability. Simulation and BSM algorithm are used to evaluate the performance of our model. Experimental results show that there is no significant difference in the determination of minimum supportability when the bearing cost of each rule or the damage rate of products damage changes. Also, discount rate has little influence on the value of supportability. Numerical analysis reveals that profit has relationship with the purchasing frequency and the value of supportability. Higher purchasing frequency with larger supportability will result a higher profit, vice verso, lower purchasing frequency with smaller supportability will result a lower profit.