隨著資通訊科技的進步及電子醫療器材的大量出現，越來越多種類的醫療紀錄需要被儲存與處理。基於方便性與可用性，現今資料幾乎都以數位形式進行儲存，而如何運用這些資料來輔助醫療行為，如臨床決策支持系統，則成為一門重要的研究議題。在臨床決策支持系統之中，預測模型是一個關鍵的技術議題，藉由分析醫療紀錄資料，可以對不同醫療指標做預測，例如:疾病或併發症早期發生的可能性、疾病發生相關特徵的評估、醫療效果預估以輔助醫療人員判斷等。這些預測模型訓練方式有些也分成one-size-fit-all模型和個人化模型，可以針對不同應用做取捨。在醫療領域中，一個療程通常都有其對應的指標去判斷療程的結果好壞，這些指標往往是國際醫療組織所定義，且這種指標通常都是療程結束後醫療機構抽血測量取得。雖然醫療指標可以輔助醫師與病患了解醫療效果，卻沒有辦法於當次療程提供即時改善的依據。因此，如何在當次的醫療行為過程中準確的預測醫療指標以輔助醫療人員優化療程，是有其必要性且具有價值的研究議題。
在本論文中，我們結合醫療設備機台上傳的機台設定與療程中監測的病人資訊以及療程進行前所做的實驗室化驗資料來開發一個醫療指標預估方法，可以在療程過程中就對療程結束後再進行化驗所取得的醫療指標進行達標與否的預測，以提供現場醫護人員進行療程優化的判斷依據，讓他們可以在療程中決定是否要調整機台參數來讓最後的指標結果有機會改善。實驗結果顯示，在不同的評估方法上，我們的方法跟其他傳統方法在達標及未達標的預測結果上較為精確且有大幅的提升，除了可以解決在醫療上實驗室化驗資料的缺失問題，也可以處理因為對該群有高相似度使得權重較高，但該群內部分佈不一致的問題，藉此降低誤判情況的產生。

With the rapid growth of the information and communication technologies and the emergence of a large number of electronic medical devices, more and more types of medical records need to be stored and processed. Based on the convenience and availability, most of the data is stored in electronic form, and how to make good use of these data to assist the medical behavior, such as clinical decision support systems, has become an important research issue. In the clinical decision support system, the prediction model is a key technical issue. By analyzing medical records, it is possible to predict some medical indicators, such as the possibility of diseases early detection and complications, the relevant feature assessment of disease occurrence and the medical effect estimation for supporting the healthcare professional to make a decision. Those models roughly could be trained as one-size-fit-all or personal ones, it depends on the different application scenarios.
In the medical field, a treatment usually has its corresponding indicator to judge whether the result is good or bad. These indicators are usually defined by the international medical organizations and usually obtained after the end of the treatment by doing a blood test for the patient. Although medical indicators can help physicians and patients understand the medical effects, there is no way to provide immediate improvement on the current treatment. Therefore, how to accurately predict medical indicators during the treatment process for assisting healthcare professionals improve the treatment is a necessary and valuable research topic.
In this thesis, we develop a medical indicator estimation method, called treatment indicator orientation prediction (TIOP), based on the machine setting uploaded by the medical equipment and the patient information monitored during the treatment and the laboratory test data before the treatment. During the treatment, it could predict the treatment indicator obtained after the end of the treatment and see whether it reaches the standard or not. Hence, they could decide whether to adjust the machine parameter, and the indicator result may improve after the adjustment. Experiments results show that the proposed method could deal with the missing data problem and dominating problem in the medical research field, and our approach compared with other methods not only is practicable, but also has much improvement and better performance in different kinds of evaluation methods.

[BEL 11] R. Bellazzi, F. Ferrazzi and L. Sacchi, “Predictive data mining in clinical medicine: a focus on selected methods and applications,” WIREs Data Mining Knowl Discov 1(5):416–430. doi: 10.1002/widm.23, 2011 Feb 11
[CDS 17] Centers for Disease Control and Prevention. National Chronic Kidney Disease Fact Sheet [CDC], 2017. Atlanta, GA: US Department of Health and Human Services, Centers for Disease Control and Prevention, 2017.
[CHE 17] M. Chen, Y. Hao, K. Hwang, Wang L and L. Wang. “Disease prediction by machine learning over big data from healthcare communities,” IEEE Access, vol. 5, No. 1, pp.8869–8879, 2017
[ECH 12] Echouffo-Tcheugui JB and Kengne AP, “Risk models to predict chronic kidney disease and its progression: a systematic review,” PLoS Med, vol. 9, No.11, 2012
[FON 17] V. Fonti, and E. Belitser, “Feature Selection using LASSO,” VU Amsterdam
Research Paper in Business Analytics, 2017.
[JAM 15] E. Jamoom and N. Yang, “National Electronic Health Records Survey: 2015 State and National Electronic Health Record Adoption Summary Tables,” Table of Electronic Health Record Adoption and Use among Office-based Physicians in the U.S., by State: 2015 National Electronic Health Records Survey, 2016.
[JUN 10] HS. Jung et al., “Analysis of hemodialysis‐associated hypoglycemia in patients with type 2 diabetes using a continuous glucose monitoring system”. Diabetes Technol Ther 12: 801–7, Oct. 2010.
[KUN 16] V. Kunwar, K. Chandel, A. S. Sabitha and A. Bansal, “Chronic Kidney Disease analysis using data mining classification techniques,” in Cloud System and Big Data Engineering (Confluence), 6th IEEE International Conference, pp. 300-305, 2016
[MAL 17] B. Malmir, M. Amini, and S. I. Chang, “A medical decision support system for disease diagnosis under uncertainty,” Expert Systems with Applications, vol.88, pp.95–108, 2017.
[NG 15] K. Ng, J. Sun, J. Hu and F. Wang, “Personalized Predictive Modeling and Risk Factor Identification using Patient Similarity,” AMIA Jt Summits Transl Sci Proc, pp.132 – 136, 2015.
[NKF 06] National Kidney Foundation [NKF/DOQI], “Clinical practice guidelines and clinical practice recommendations,” 2006 updates hemodialysis adequacy, peritoneal dialysis adequacy, vascular access. Am J Kidney Dis, 2006.
[PAR 11] G. Parthiban, A. Rajesh and S. K. Srivatsa, “Diagnosis of Heart Disease for Diabetic Patients using Naive Bayes Method,” International Journal of Computer Applications, vol. 24, No. 3, pp.7–11, June. 2011.
[PER 15] A. Perotte, R. Ranganath, J. S. Hirsch, D. Blei and N. Elhadad, “Risk prediction for chronic kidney disease progression using heterogeneous electronic health record data and time series analysis”. J. Am. Med. Inform. Assoc, vol. 22, No. 5 pp. 872–880, 2015.
[RAJ 18] A. Rajkomar, et al., “Scalable and accurate deep learning for electronic health records”. npj Digital Medicinevolume, vol. 1, No.18, 2018.
[RAV 14] A. V. Ravindr, N. Sriraam and M. Geetha, “Discovery of significant parameters in kidney dialysis data sets by K-means algorithm,” 2014 International Conference on Circuits Communication Control and Computing, Nov. 2014.
[SAN 15] T. Santhanam and M.S. Padmavathi, “Application of K-Means and Genetic Algorithms for Dimension Reduction by Integrating SVM for Diabetes Diagnosis”, Procedia Computer Science, vol. 47, pp. 76-83, 2015.
[SAY 17] A.K.Sayana, K.P. Sreelakshmi and T. Vanitha, “A comparative study of KNN and SVM data classification algorithms in chronic kidney disease,” International Journal of Latest Trends in Engineering and Technology Special Issue SACAIM, pp. 426-429, 2017.
[SHA 17] A. Sharafoddini, J.A. Dubin and J. Lee, “Patient similarity in prediction models based on health data: a scoping review,” JMIR medical informatics, 2017.
[SMO 04] A. J. Smola and B. Schӧlkopf, “A tutorial on support vector regression”. Statistics and computing, vol. 14, No.3, 2004.
[SUO 17] Q. Suo, F. Ma, Y. Yuan, M. Huai, W. Zhong, A. Zhang and J. Gao, “Personalized disease prediction using a CNN-based similarity learning method,” Bioinformatics and Biomedicine (BIBM), 2017 IEEE International Conference on, 2017.
[TAN 11] N. Tangri, et al. “A predictive model for progression of chronic kidney disease to kidney failure,” JAMA, vol. 305 pp. 1553-9, 2011.
[TAZ 16] N. Tazin, S.A. Saba and M.T. Chowdhur, “Diagnosis of Chronic Kidney Disease using effective classification and feature selection technique,” International Conference on Medical Engineering Health Informatics and Technology (MediTec), 2016.
[TIK 18] P. Tikariha and P. Richhariya, “Comparative Study of Chronic Kidney Disease Prediction using Different Classification Techniques,” Proceedings of International Conference on Recent Advancement on Computer and Communication. Lecture Notes in Networks and Systems, vol. 34, 2018.