根據衛福部統計處最新的老人狀況調查指出，在台灣現今有81%的老年人患有至少1種慢性病。罹患慢性病的患者須長期使用藥物控制病情，而用藥錯誤以及忘記服藥是目前最為嚴重的兩個問題，本論文將提出結合機器學習與語意推論於雲端運算架構(Integrating Machine Learning and Semantic into Cloud Computing Framework, IMLSCCF)來解決上述的問題，並開發出雲端用藥安全平台(Safety Medication Cloud Platform, SMCP)來驗證IMLSCCF的可行性。本研究使用健康存摺及開放資料(Open Data)當作資料來源，並藉由Fuzzy C-means(FCM)演算法、物聯網(Internet of Things)、語意網(Semantic Web)以及雲端運算(Cloud Computing)技術建構出系統基礎架構，物聯網部分是由手機App、智慧藥盒以及雲端管理平台所構成，可在需服藥時提醒銀髮族服藥以及提供藥物的相關資訊。由於藥物經常會與其他藥物或食物產生交互作用，因此本論文將透過語意網來推論出必須避免的藥物及食物，而食物中通常包含各種成分，論文中使用了結巴(Jieba)套件過濾並統計重要詞彙後，再透過機器學習(Machine Learning)中的FCM演算法做分類，本論文將會使用食品相關的開放資料做為測試資料的來源。由於資料會隨著使用者與食品種類的增加而增加，為了應付日積月累下來的龐大資料量，本論文將會使用Spark將語意網以及機器學習的部分利用雲端運算來實現，並透過Docker套件來達到虛擬化的目的，同時也比較了不同雲端運算環境之間的效能差異。透過SMCP所提供的功能可驗證IMLSCCF的可行性，並解決忘記服藥與用藥錯誤的問題。

As shown in the latest senior citizens survey of the Department of Statistics, Ministry of Health and Welfare, 81% of the senior citizens in Taiwan suffer from at least one chronic disease currently. The patients with chronic disease must take pills for long terms to control the illness condition. Taking wrong pills and forgetting to take pills are the most serious problems currently. This study will propose Integrating Machine Learning and Semantic into Cloud Computing Framework (IMLSCCF) to resolve these two problems. Moreover, it develops the Safety Medication Cloud Platform(SMCP) to validate the feasibility of IMLSCCF. The study takes the Health Bank and the Open Data as the data source, and uses Fuzzy C-means(FCM) algorithm, Internet of Things (IoT), Semantic Web and Cloud Computing technology to build the basic system architecture. In the part of IoT, it consists of the mobile App, smart pillbox and cloud management platform. When time is up, it can remind the patients to take pills and provide the related information of pills. As pill often generates reaction with other medicine or food, this study will apply the Semantic Web to infer the medicine and food types that must be avoided. Moreover, as the food contains various ingredients, this paper uses Jieba suite to filter and count the important vocabularies, and then make classification through FCM algorithm. It will take the open data related to food as the source of test data. The data will grow with the number of users and food types. To avoid the big data accumulated in long times, this paper will use Spark to realize the Semantic Web and the machine learning part through cloud computing. Moreover, it achieves the virtualization purpose through Docker suite. This makes the distribution of cloud computation resources more flexible. With the functions provided by SMCP, it can validate the feasibility of IMLSCCF and resolve the problems of forgetting to take pills and take wrong pills.

摘要 ................................................. i
Abstract ............................................ ii
誌謝 ................................................. iii
目錄 ..................................................iv
表目錄 .............................................. vi
圖目錄 ............................................. vii
第一章 簡介 ........................................... 1
1.1 研究背景 .......................................... 1
1.2 研究動機 .......................................... 2
1.3 研究目的 .......................................... 2
第二章 研究技術探討 ................................... 3
2.1 開放資料 .......................................... 3
2.2 健康存摺 .......................................... 3
2.3 物聯網的應用 ...................................... 4
2.3.1 物聯網 .......................................... 4
2.3.2 Arduino Ameba ................................... 5
2.4 結巴斷詞 .......................................... 6
2.5 模糊C-means分群法 ................................. 7
2.6 語意網(Semantic Web) .............................. 9
2.6.1 RDF ............................................. 9
2.6.2 OWL ............................................. 9
2.6.3 Rules .......................................... 10
2.6.4 Jena ........................................... 10
2.7 Hadoop ........................................... 11
2.7.1 Hadoop分散式檔案系統 ........................... 12
2.7.2 Hadoop YARN .................................... 13
2.8 Spark ............................................ 14
2.8.1 彈性分散式資料集 ............................... 15
2.8.2 Spark叢集管理器 ................................ 16
2.9 Docker ........................................... 17
2.10 相關研究 ........................................ 18
第三章 系統設計與研究步驟 ............................ 19
3.1 系統設計 ......................................... 19
3.1.1 結合機器學習與語意推論於雲端運算架構 ........... 19
3.1.2 雲端用藥安全平台 ............................... 21
3.1.3 開發環境 ....................................... 22
3.2 研究方法與流程 ................................... 23
3.2.1 匯入開放資料 ................................... 23
3.2.2 物聯網架構 ..................................... 24
3.2.3 結巴斷詞 ....................................... 25
3.2.4 Fuzzy C-Means演算法 ............................ 26
3.2.5 制定Ontology與Rule推薦.......................... 30
3.2.6 結合雲端運算 ................................... 42
3.2.7 整合機器學習以及語意網 ......................... 44
第四章 系統實作 ...................................... 45
4.1 應用場景 ......................................... 45
4.2 系統特色 ......................................... 46
4.3 雲端管理平台 ..................................... 47
4.4 App行動裝置 ...................................... 53
4.5 智慧藥盒 ......................................... 56
第五章 系統實驗與評估 ................................ 58
5.1 雲端運算環境架構 ................................. 59
5.1.1 測試環境一 ..................................... 59
5.1.2 測試環境二 ..................................... 60
5.1.3 測試環境三 ..................................... 61
5.1.4 雲端運算叢集模式 ............................... 63
5.2 雲端運算效能測試 ................................. 65
5.2.1 測試一 ......................................... 65
5.2.2 測試二 ......................................... 67
5.2.3 測試一與測試二比較 ............................. 69
5.2.4 測試三 ......................................... 70
5.2.5 測試四 ......................................... 72
5.2.6 測試三與測試四比較 ............................. 74
5.3 測試總結 ......................................... 75
5.4 FCM演算法準確性評估 .............................. 76
第六章 結論 .......................................... 77
參考文獻 ............................................. 78
Extended Abstract .................................... 81
簡歷(CV) ............................................. 96

[1]行政院衛生福利部, “中華民國102年老人狀況調查報告,” 行政院衛生福利部, p. 500, 2014.
[2]Wikipedia, “Open data.” [Online]. Available: https://en.wikipedia.org/wiki/Open_data.
[3] C.Yeh, P.Wang, Y.Pan, and M.Kao, “A Scalable Privacy Preserving System for Open Data,” 2016 Int. Comput. Symp., pp. 312–317, 2016.
[4]中央健保局, “健康存摺.” [Online]. Available: https://myhealthbank.nhi.gov.tw/IHKE0002/IHKE0002S01.ASPX.
[5]Wikipedia, “IoT.” [Online]. Available: https://en.wikipedia.org/wiki/Internet_of_things.
[6]A.Giri, S.Dutta, and S.Neogy, “Enabling Agricultural Automation to Optimize Utilization of Water, Fertilizer and Insecticides by Implementing Internet of Things (IoT),” 2016 Int. Conf. Inf. Technol. InCITe 2016 - Next Gener. IT Summit Theme - Internet Things Connect your Worlds, pp. 125–131, 2017.
[7] T.Vall, K.Sedki, S.Despres, M.Jaulant, K.Tabia, and A.Ugon, “On Personalization in IoT,” 2016 Int. Conf. Comput. Sci. Comput. Intell., pp. 186–191, 2016.
[8]P.Dhar, P. P.Gupta, and G. H. R.College, “Intelligent Parking Cloud Services based on IoT using MQTT Protocol,” Int. J. Eng. Res., vol. 5013, no. 5, pp. 457–461, 2016.
[9]M.Mohsen, F.Nasaruddin, A.Gani, A.Karim, I.Hashem, A.Siddiqa, and I.Yaqoob, “Big IoT Data Analytics: Architecture, Opportunities, and Open Research Challenges,” IEEE Access, vol. PP, no. 99, p. 1, 2017.
[10]Wikipedia, “Arduino.” [Online]. Available: https://en.wikipedia.org/wiki/Arduino.
[11]S. T.Kebir, M.Bouhedda, S.Mekaoui, M.Guesmi, and A.Douakh, “Gesture Control of Mobile Robot Based Arduino,” 2016 8th Int. Conf. Model. Identif. Control, pp. 1081–1085, 2016.
[12]A.Nayyar and V.Puri, “A Review of Arduino Board’s, Lilypad’s & Arduino Shields,” 2016 Int. Conf. Comput. Sustain. Glob. Dev., pp. 1485–1492, 2016.
[13]Realtek, “Ameba.” [Online]. Available: https://www.amebaiot.com/.
[14]J. C.Bezdek, R.Ehrlich, and W.Full, “FCM: The Fuzzy C-means Clustering Algorithm,” Comput. Geosci., vol. 10, no. 2–3, pp. 191–203, 1984.
[15] I.Mirrezaei, “Information Extraction from Unstructured Data using RDF,” 2016 Int. Conf. ICT Bus. Ind. Gov., no. July, pp. 1–16, 2013.
[16]J.Xu, H.Wang, and H.Trimbach, “An OWL Ontology Representation for Machine-learned Functions using Linked Data,” Proc. - 2016 IEEE Int. Congr. Big Data, BigData Congr. 2016, pp. 319–322, 2016.
[17]A.Fobel and N.Subramanian, “Comparison of the Performance of Drools and Jena Rule-based Systems for Event Processing on the Semantic Web,” 2016 IEEE/ACIS 14th Int. Conf. Softw. Eng. Res. Manag. Appl. SERA 2016, pp. 24–30, 2016.
[18]A.Ameen, K.U.R.Khan, and B.P.Rani, “Extracting Knowledge from Ontology using Jena for Semantic Web,” 2014 Int. Conf. Converg. Technol. I2CT 2014, pp. 2–6, 2014.
[19]Q.Liu, Y.Fu, G.Ni, and J.Mei, “Big Data Management Performance Evaluation in Hadoop Ecosystem,” 2017 3rd Int. Conf. Big Data Comput. Commun., pp. 413–421, 2017.
[20]S.He, “Spatial Query Processing for Location Based Application on Hbase,” 2017 IEEE 2nd Int. Conf. Big Data Anal., pp. 110–114, 2017.
[21]R.Ilakiyaa and N. J.Nalini, “Supervised Learning Based HDFS Replication Management System,” 2017 Int. Conf. Tech. Adv. Comput. Commun., pp. 116–120, 2017.
[22]Y.Shao, C.Li, W.Dong, and Y.Liu, “Energy-Aware Dynamic Resource Allocation on Hadoop YARN Cluster,” 2016 IEEE 18th Int. Conf. High Perform. Comput. Commun. IEEE 14th Int. Conf. Smart City; IEEE 2nd Int. Conf. Data Sci. Syst., 2016.
[23]W.Huang, L.Meng, D.Zhang, and W.Zhang, “In-Memory Parallel Processing of Massive Remotely Sensed Data Using an Apache Spark on Hadoop YARN Model,” IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., vol. 10, no. 1, pp. 3–19, 2017.
[24]J.Chen, K.Li, S.Member, Z.Tang, and K.Bilal, “A Parallel Random Forest Algorithm for Big Data in a Spark Cloud Computing Environment,” IEEE Trans. Parallel Distrib. Syst., vol. 28, no. 4, pp. 919–933, 2017.
[25]W.Zhang, T.Mo, W.Li, H.Huang, and X.Tian, “The Comparison of Decision Tree Based Insurance Churn Prediction between Spark ML and SPSS,” 2016 9th Int. Conf. Serv. Sci., pp. 134–139, 2016.
[26]C.-S. N.Min-Seon Lee, Earl Kim, “Design of Educational Big Data Application using Spark,” 2017 19th Int. Conf. Adv. Commun. Technol., pp. 355–357, 2017.
[27]R.DelValle, P.Kaushik, A.Jain, J.Hartog, and M.Govindaraju, “Electron: Towards Efficient Resource Management on Heterogeneous Clusters with Apache Mesos,” 2017 IEEE 10th Int. Conf. Cloud Comput., pp. 262–269, 2017.
[28]A.Abu-Doleh and U.V.Catalyurek, “Spaler: Spark and GraphX based de novo genome assembler,” Proc. - 2015 IEEE Int. Conf. Big Data, IEEE Big Data 2015, no. C, pp. 1013–1018, 2015.
[29]B.Akgün and Ş. G.Öğüdücü, “Streaming Linear Regression on Spark MLlib and MOA,” 2015 IEEE/ACM Int. Conf. Adv. Soc. Networks Anal. Min., pp. 1244–1247, 2015.
[30]R.Joy and K. K.Sherly, “Parallel Frequent Itemset Mining with Spark RDD Framework for Disease Prediction,” Proc. IEEE Int. Conf. Circuit, Power Comput. Technol. ICCPCT 2016, 2016.
[31]N.Naik, “Docker Container-Based Big Data Processing System in Multiple Clouds for Everyone,” 2017 IEEE Int. Syst. Eng. Symp., pp. 1–7, 2017.
[32]P.Xu, S.Shi, and X.Chu, “Performance Evaluation of Deep Learning Tools in Docker Containers,” 2017 3rd Int. Conf. Big Data Comput. Commun., pp. 395–403, 2017.
[33]Q.Zhu, “Pharmacological Class Data Representation in the Web Ontology Language (OWL),” 2014 IEEE Int. Conf. Big Data (Big Data), pp. 77–84, 2014.
[34]T.Yu, M.Cui, and L.Jia, “Towards a TCM Domain Ontology : Standardization , Ontology Engineering , and Applications,” 2015 8th Int. Conf. Biomed. Eng. Informatics, no. Bmei, pp. 500–504, 2015.
[35]T.Horan, “Toward Global Health Interoperability : Ontology Development of Traditional Chinese Medicine Through Comparative Analyses,” 2015 48th Hawaii Int. Conf. Syst. Sci., no. Icd, 2015.
[36] R.Chen, Y.Lo, B.Liao, and C.Bau, “Knowledge Integration for Diabetes Drugs Ontology,” vol. 2, pp. 87–94, 2014.
[37]P.Partila, J.Tovarek, J.Frnda, M.Voznak, M.Penhaker, and T.Peterek, “Knowledge Integration for Diabetes Drugs Ontology, ” vol. 298. 2014.
[38]G.Jiang, H.Liu, and H. R.Solbrig, “Mining Severe Drug-Drug Interaction Adverse Events Using Semantic Web Technologies : A Case Study,” BioData Min., pp. 628–638, 2014.
[39]A.Ben, F.Mahbub, A.Karanasiou, Y.Mrabet, A.Lavelli, and P.Zweigenbaum, “Text Mining for Pharmacovigilance : Using Machine Learning for Drug Name Recognition and Drug – drug Interaction Extraction and Classification,” J. Biomed. Inform., vol. 58, pp. 122–132, 2015.
[40]S.Hunta, “Drug-Drug Interactions Prediction from Enzyme Action Crossing Through Machine Learning Approaches,” 2015 12th Int. Conf. Electr. Eng. Comput. Telecommun. Inf. Technol. pp. 1–4, 2015.
[41]K.Sid, “Big Data Analytics Techniques in Virtual Screening for Drug Discovery.” Proceeding BDCA’17 Proc. 2nd Int. Conf. Big Data, Cloud Appl., 2017.
[42]D.Harnie, A.Vapirev, J.Wegner, A.Gedich, M.Steijaert, R.Wuyts, and W.Meuter, “Scaling Machine Learning for Target Prediction in Drug Discovery using Apache Spark,” 2015 15th IEEE/ACM Int. Symp. Clust. Cloud Grid Comput., pp. 871–879, 2015.