利用分類技術分析產品項目最適性之行銷組合

Using Classification Techniques to Analyze the Most Adaptive Marketing Mix of Products

在本篇論文中，我們以消智者之交易資料為探勘的資料來源，每一筆交易資料包含有消費者曾經購買過的產品項目與其數量，以某些產品項目為探勘的目標，並視其他的產品為分類的屬性項目，分別從以下兩方面來發掘這些產品項目最適性的行銷組合：首先，我們只考量產品項目是否出現在交易資料中，在探勘的過程中，若交易資料包含有這些產品項目的比例值達到最小關聯度，則設定交易資料與這些產品項目之間的關聯性為「高」，否則設定為「低」。我們將交易資料進行分類分析以建構出一決策樹，從所建構出的決策樹中，可以找出那些屬性項目與這些產品項目之間的關聯性高，藉此做為發掘這些產品項目最適性之行銷組合的依據。再者，我們增加考量產品項目的購買數量，在探勘的過程中，若交易資料包含有這些產品項目的比例值達到最小數量關聯度，則設定交易資料與這些產品項目之間的關聯性為「高」，否則設定為「低」。我們提出一個方法來將包含有項目數量之交易資料進行分類分析以建構出一決策樹，從所建構出的決策樹中，可以找出那些屬性項目與這些產品項目之間的關聯性高，藉此做為發掘包含有項目數量之這些產品項目最適性的行銷組合的依據。此探勘結果，對企業在擬訂產品之行銷組合的策略時，將可以提供非常有用的參考資訊。

In this paper, we use consumers' transaction data as the source data of mining. Each transaction data contains a consumer ever bought product items with quantity. We let some product items as the target of mining, and regard other products as attribute items for classification. We discover the most adaptive marketing mix of the product items from two aspects. First, we only consider product items whether they are contained in transaction data or not. In the mining process, we compute the percentage of transaction data contains the product items, and assign the association degree between the transaction data and the product items to be "high" if the percentage satisfies the minimum association threshold. Otherwise, it is "low". We classify the transaction data to construct a decision tree, and find out the attribute items that have the association degree to be high with the product items according to the decision tree. It is the basis to discover the most adaptive marketing mix of the product items. Moreover, we extra consider product items with quantity in the transaction data. In the mining process, we compute the percentage of transaction data contains the product items, and assign the association degree between the transaction data and the product items to be "high" if the percentage satisfies the minimum quantitative association threshold. Otherwise, it is "low". We propose a method to classify the transaction data with quantitative items for constructing a decision tree, and find out the attribute items that have the association degree to be high with the product items according to the decision tree. It is the basis to discover the most adaptive marketing mix with quantitative items of the product items. The results of the mining can provide very useful information to plan the strategy of marketing mix of products for the business.