The Application of DSS with the Assitant of Data Diffusion Models

研究領域:工業工程類 計畫編號:NSC98-2221-E273-003[[abstract]]決策支援系統在與商業範疇連結時，需要正確的資料來源作為決策背景之領域知識的基礎；傳統的企業智慧所牽涉之工具莫不需要大量的資料以驗證決策模式的穩定性，然而在系統建立初期，沒有充分的資料範例作為建模的依據，決策的品質將會受到影響；本研究計畫針對系統建立初期架構資料擴散模式，藉由此技術從現有少量資料產生出虛擬樣本，並據此建立決策模式，如此可加速決策支援系統的穩定性。因此，研究內容前半部著重於，資料擴散模式的建立以及與決策支援系統的結合，再者，針對商業上實例應用，分析並討論其成效與可行性。本計畫將針對三種類型的企業，分析資料擴散模式對其商業決策支援的可行性，包含兩個傳統產業：紡織業與染料業，以及一個汽油通路零售業；這些企業不若一般的資訊科技產業重視顧客關係管理，因此，在進行商業決策時，缺乏強而有利的決策支援系統，甚至希望系統建立初期就能馬上使用，本計畫將應用資料擴散模式，以小樣本資料分析的技術增加虛擬樣本，並結合常見之類神經網路、貝氏網路等工具，分別為各種案例建立特定問題之決策模式。並分析比較使用資料擴散模式的成效。 When applying Decision Support Systems (DSS) in business scope, it is necessary to construct the domain knowledge based on sufficient and correct data for decision-making. The traditional business intelligence (BI) tools need massive data to validate the stability of the decision models. However, in the early stage of system establishment, the decision quality would be fragile without enough data for the model construction. The research is aimed to establish the data diffusion model for the system’s early stage. Virtual samples will be generated by this technology, from fewer data. The resulted decision model would possess the better stability. The first part of this research focuses on the model construction and the system integration, and then the analysis of feasibility of business applications. In the applications, we focus on three types of industries when analyzing the feasibility. It includes two traditional industries: the spinning and the dyestuff industry, and a gas retailer. These industries pay less attention to their customer relationship management and are lack of the useful DSS for business decision-making. The project will apply the data diffusion model to produce virtual samples based on small data sets. With connecting to the usual tools, like Artificial Neural Networks (ANN) and Bayesian Network (BN) etc., we construct the specific decision model for each case and analyze the corresponding performance