應用NoSQL資料庫建置健保資料庫之巨量資料視覺化呈現

Abstract

國家衛生研究院所建置的健保資料庫，儲存了自健保開辦以來全台投保者的就醫資料，累積下來非常可觀的健保資料成為實證的基礎，期盼從這些資料當中分析出有價值、有意義的訊息。 為了建置1996年到2010年15年間台灣健保資料庫一百萬人抽樣歸人檔，本研究選擇使用文件型NoSQL資料庫 - MongoDB，發揮其處理非結構化資料的特色，將原本近兩億筆的大量就醫紀錄整合為百萬筆歸人資料，不僅降低了龐大的資料量，也更為便利地查詢使用者有興趣之就醫經歷的患者。 在資料的存放上透過分片(Sharding)技術，以患者出生年月為依據進行分片，並結合MapReduce來進行運算，可循序有效地取得目標患者的資料，提供使用者有良好的操作環境。 本研究的目標是從巨量的健保資料中，即時地取得符合查詢條件之就醫經歷的患者資料。而患者就醫紀錄可能相當繁多，為了清楚地呈現患者的狀況，系統提供一視覺化的介面，使用者能一目了然地查看患者就醫的經歷，並做進一步的探討。

National Health Research Institutes of Taiwan had build National Health Insurance Research Database, this database stored medical information since 1996 insured health care offered, it accumulated substantial health care data that has become the basis of many valuable medical studies. In order to build a more efficient 1996 to 2010, 15 years, Taiwan National Health Insurance Health Database of Longitudinal Health Insurance Database 2010, this study uses a NoSQL Database - MongoDB, to process characteristics of unstructured data, it transformed originally nearly 200 million medical records, into millions of personal data, not only reduces the amount of data, made them convenient for users to search particular medical information which they are interested. Using the sharding technique, individual date of birth was picked as the basis for sharding, combined with MapReduce operations user can efficiently acquire their target information. The object of this study is to arrange the healthcare information in a timely manner to provide an easy access medical data format to render the patient's condition, on a visual interface, users can view patient medical detail with few keyboard strokes.