心血管疾病是全球最常見的死亡原因之一,不僅心血管疾病檢查成本高昂,管理心血管疾病病患也對全世界的醫療系統造成極大的負擔。因此心血管疾病一直都是醫療界積極研究的目標疾病之一,其中也有許多研究提出診斷支援系統來協助預測心血管疾病病患,但這些決策支援系統都有個共通點,便是都需要透過該領域專業人員來建立專家知識庫,在分析結果的解釋與理解上並不是這麼容易,因此本研究希望透過結合資料探勘技術與鏈結開放資料來建立一套通用型決策諮詢架構,透過決策樹演算法來針對UCI心臟病資料集建立心臟病疾病預測模型,確立資料集當中對於心臟病疾病最有影響力的風險因子欄位,並且利用網路上存放鏈結開放資料的Endpoint進行SPARQL查詢,查詢並抓取與該風險因子相關的鏈結開放資料,最後將其整合並利用關聯規則演算法進行分析,分析出有用的規則或模式用以解釋或補充決策樹模型所預測之結果,並提供相關知識以協助使用者進行決策。最終實驗結果證明本研究所提出之決策諮詢架構的可行性且令使用者能夠於易了解或解釋決策樹模型的分析結果。由於建置知識庫並不需要領域專業人員,因此可以透過更換主分析資料集,便可將該架構應用於不同領域當中,並且利用Endpoint的SPARQL查詢抓取與該領域相關之鏈結開放資料分析建置知識庫,且可根據發佈之鏈結開放資料的變化進行動態更新。
Cardiovascular disease is one of the most common causes of death in the world. Not only is the cost of cardiovascular disease examinations high, but the management of cardiovascular diseases also places a huge burden on the medical system worldwide. Therefore, cardiovascular disease has always been one of the target diseases actively studied by the medical community. Many of them have proposed diagnostic support systems to help predict cardiovascular diseases. However, these decision support systems have a common point, they all need to establish an expert knowledge base through professionals in the field. Furthermore, the analysis results are not easy to explain and understand. Therefore, this study hopes to establish a general decision-making consultation framework by combining data mining technology and link open data. Through the decision tree algorithm to establish a predictive model of heart disease for the UCI heart disease dataset, establish the most influential risk factor field for heart disease in the dataset, and use the Endpoint on the network which store open data links to do the SPARQL query. Query and capture the linked open data related to the risk factor, and finally integrate it and analyze it by using the association rule algorithm and analyze useful rules or patterns to explain or supplement the predicted results of the decision tree model and provide relevant knowledge to assist users in making decisions. The final experimental results prove the feasibility of the decision-making consulting framework proposed by this study and enable the user to easily understand or interpret the analysis results of the decision tree model. Since the establishment of the knowledge base does not require domain professionals, the architecture can be applied to different fields by replacing the main analysis data set, and Endpoint's SPARQL query is used to capture the open data analysis of the links associated with the domain. The knowledge base is built and dynamically updated based on changes in the published open data.