應用資料探勘技術於動物醫院病歷資料之研究

由於動物醫療系統發展越來越成熟，傳統紙本病歷已被電子病歷取代，動物就醫的各種診斷資料被完整的紀錄，使用動物電子病歷來做資料探勘分析的學者也越來越多了。隨著醫學發展時代的演進，醫療系統走向專科化，來提供寵物更專業合適的治療，但就診科別越分越細，往往使寵物雇主陷入該看什麼科別或是看錯科別的窘境，而這些動物病歷如果能適當加以分析是可以找出一些影響疾病的關鍵因素或關聯的，藉由動物就醫資料來找出疾病與症狀之間的關聯性，來給予醫療診斷的資訊參考，且可降低診斷過程中的疏忽。因此，本研究是針對動物醫院就診病歷資料來進行探勘，利用分群演算法與關聯法則兩種方法，以不事先設定的作法探討動物醫院就診病歷資料中，與泌尿科、皮膚科或內科有高度相關的可能症狀有哪些，希望能找到寵物最常看診的症狀疾病關聯。先以HTCA(Hierarchical Thresholding Clustering Algorithm)分群法找出各症狀所歸屬的群集，再使用關聯法則來找出影響分類的屬性，進而找出症狀歸類的原因並且了解動物就診情況。

關鍵字

資料探勘；分群化；關聯法則；電子病歷

並列摘要

In this paper, we try to find the relations by applying data mining technology to the diagnostic data. The correlations between the treatments and the symptoms of sick pets are collected to reduce the negligence of the diagnostic processes. This research uses medical records of an animal hospital as the input of mining process. The clustering algorithms and association rules use two methods to investigate the practice. The original medical records have been pre-processed as a useful table. Since urology, dermatology or internal medicine are highly correlated with possible symptoms, we hope the conclusion of this paper may help doctors to find the association rules between symptoms ad diseases of pets. Two steps are applied in the experiments. We first find the proper clustering by applying HTCA clustering method, and then use the association rule to find the impact of divisions of property. After identifying the impact classification attributes, and then veterinarians may understand treatment situation of the sick animals.