Application of Data Mining Technique to Inventory classification management of Outpatient Services of Hospitals

[[abstract]]本研究探討醫療機構之藥品存貨數量管理。藥品存貨數量及其多品項因素而不易控管，對醫療人員及使用者而言，需要某種合理程度的存貨數量，不致有斷貨的現象而造成影響醫療業務執行及損及病患權益，也同時為不造成因避免藥品短缺的情形而過多的採購，進而增加成本，若是能在醫療物料進入醫療行為使用階段之前，以準確的存貨分類管理，將把醫療物料在醫療機構內各單位的實際使用品項及數量，做最佳的分配，可成為成功管控藥品存量的第一步驟。 在傳統思維中，存貨分類管理通常以成本做為決策變數，以各個物料在存貨管理規劃週期內（例如一季）之總使用金額為依據，計算出每一項目佔總存貨總數，以及總存貨金額的比例，將存貨加以分類，ABC分析法為其典型代表，在分類後進而對各類存貨項目訂定不同的存貨管理策略。由於醫療機構中藥品物料數往往有數百種甚至上千種之多，每分每秒使用醫療對象也不同，因此如果只以傳統成本為中心的分類方法來進行藥品存貨的分群，則會漏失許多在如此大量的交易資料中隱含著許多重要的資訊，且難察覺當中具特殊意義的各項關聯無法達到真正「最佳存量管制」之意義。進而，本研究希望能利用近年來所廣為討論的關聯法則（association rules）對於醫療交易資料庫進行資料探勘（Data mining）各項資料間的關聯性（Relationships），推導出一個能夠輔助單一法則存貨分類的關聯性分類方法，然後再對分類後提出與ABC分類存貨之間的關聯性分析評估，以作為決策的重要參考。 本研究利用個案醫院2010年1到3月的門診健保申報檔進行資料探勘，共計有476177筆申報資料，以月為階層單位，在多階層同一最小支持度概念下，使MIS=0 001％，MIC=50％，結果得到19條關聯法則（應該為15項，因有四項具有互相關聯影響），再對照門診存貨ABC分析表，可以得到從大量資料中探勘出從屬關係，來建立藥物使用模型和建立價值優先關係，因此，運用資料探勘技術，在此龐大資料庫中難以察覺的關係後，能夠提供管理人員適時調整存貨政策。[[abstract]]The purpose of the study is to explore the optimum medical articles inventory control in medical organization It is not easy to well control medical articles inventory due to multi-item factors and huge quantity in a medical organization Reasonable stock is a need for medical staffs and patients by keeping all medical articles available moreover to fulfill medical business implementation and not to affect the right of patient by shortage of medical articles Meanwhile It also has a benefit not to increase cost since the prediction of possible shortage cause over purchase The first step to succeed controlling the medical inventory control is the precious categorized of inventory and distribute right medical articles quantities to correct department accurately it will be the first step to do the successful stock control In traditional idea the cost is always be considered as decision variable Decision variable always considered the cost in traditional perception Based on the amount of cost of the inventory management in a planning cycle (eg a quarter) to calculating the total inventory items and the total amount of inventories ratio will be classified to make different inventory management strategies for each catalog ABC analysis is the typical method There are hundreds of variable pharmaceutical materials in medical institution hence large amount of transaction data and important information will be ignored if the inventory clustering is still decided by the traditional cost approach It is hard to detect all the special meaning relevant the "optimum inventory control" in the information Furthermore the study widely utilize the association rules on data mining of the medical transaction database in order to derive a single assist stock association rule on the classification categories and then submits the evaluation of ABC classification and the relationship on inventory analysis as an important reference for decision-making In this study 476 177 of data extracted from National Health Insurance files range from January to March 2010 the unit is monthly and in the multi-level concept MIS = 0 001% MIC = 50% the results are 19 association rules (actually 15 because there are four effects are interrelated) According to ABC analysis of inventory control table we can get a lot of data from a sub-model and set up a priority relationships In conclusion with techniques we will find large database- difficult to detect the relationship and it will provide management staff to make timely inventory management policies Many information can get from a subordinated model based on ABC analysis table of inventory to establish medical utilizing model and priority In conclusion apply data mining technique to find out the potential relationship in a huge database which can provide managing staff to adjust inventory policies timel