應用機器學習與模糊神經網路在健康照護之研究

The Study of Applying Machine Learning and Fuzzy Neural Network to Health Care

戴彬(DAI, BIN)

指導教授 ： 陳榮靜

朝陽科技大學/資訊學院/資訊管理系/博士(2019年)

摘要

醫療資訊資料的快速積累為人工智慧與健康領域的結合提供了基礎。人工智慧在網路結構和演算法的不斷發展與優化中，在醫療健康領域的應用愈發廣泛和深入。人工智慧在醫療領域推動著醫療診治模式、疾病的診斷與預防、藥物研發、成本控制和決策等方面展示了愈加重要的作用。人工智慧和機器學習方法在人類健康領域的應用正在改變著傳統的健康服務領域的模式。對於維護人體健康的方方面面，機器智慧方法已經可以提供越來越有價值的資訊，同時可以提供準確的輔助決策。對於人體健康的推薦包括日常健康意見、疾病狀態即時監控、醫療資料快速整合、治療方案推薦等。
本篇論文針對人體健康維護的相關問題，討論和研究了利用機器智慧方法如何更加簡潔、高效、準確地提供人體健康的推薦系統，包括利用模糊神經網路、機器學習方法在維護人體健康方面的架構設計、資料處理、深度網路模型的設計、結果視覺化等方面進行了研究和探討。同時，設計並驗證了利用多層結構的機器學習模型來進行樣本資料的分析與訓練，並取得了較好的結果。該模型以不同的角度和方法提出了除深度卷積神經網路之外的智慧資料處理方法。
論文的研究內容包括：
1.研究了與健康照護問題相關的機器智慧方法，包括模糊控制系統與神經網路結合所產生的建模方法，以及經典的機器學習演算法，如隨機森林、支援向量機等，並討論了在資料分析、規則推理、輸入參數調整、權重調整等方面的改進與實現。
2.針對現代網路環境中資料的特性與結構，引入了機器學習理論並探討與對比在現代控制方法中的應用與優勢，提出了在具體問題中實現方法；
3.提出了針對健康照護問題的機器學習的架構設計與思路，分析與探討各組成模組的相互關係與整體解決方案思路。
4.提出了一種多層結構的深度學習架構建立方法，利用卷積神經網路的方法將成熟的深度學習網路架構引入多層處理結構，利用其穩定的結構與權重設置，改變控制目的，實現在較少計算資源的條件下產生優化輸出。並量化評價其結果的準確性。

關鍵字

健康照護 ； 模糊神經網路 ； 機器學習 ； 支援向量機 ； 隨機森林 ； 深度學習

並列摘要

With the combination of artificial intelligence technology and medical data, the application in the field of medical health has become more extensive and in-depth. Artificial intelligence continues to accelerate the development of the medical field and has carried out a series of changes in the diagnosis and treatment of personal diseases, drug development, diagnosis and control of new diseases. The application of artificial intelligence and machine learning methods in human health is reshaping the statutes of the entire industry. For all aspects of human health, machine intelligence approaches can provide more and more valuable information while providing accurate decision making. Recommendations for human health include daily health advice, real-time monitoring of disease status, rapid integration of medical data, and recommendations for treatment options.
In the dissertation we discussed and studied the use of machine intelligence to provide a more concise, efficient and accurate recommendation system for human health, including the use of fuzzy neural networks and machine learning methods to maintain human health. Architecture design, data processing, design of deep network models, visualization of results have been studied and discussed. At the same time, the machine learning model with multi-layer structure was designed and verified to analyze and train the sample data, and good results were obtained. The model proposes intelligent data processing methods other than deep convolutional neural networks from different angles and methods. The research content of the dissertation r includes:
1. Researched the machine intelligence methods related to health care issues, including the modeling methods generated by the combination of fuzzy control systems and neural networks, as well as classical machine learning algorithms such as random forests, support vector machines, and discussed improvements and implementations in analysis, rule reasoning, input parameter adjustment, weight adjustment, etc.
2. Aiming at the characteristics and structure of data in a modern network environment, the machine learning theory is introduced, and the application and advantages of modern control methods are discussed and compared, and the realization method in specific problems is proposed.
3. The architecture design and ideas of machine learning for health care are put forward, and the relationship between each component module and the overall solution ideas are analyzed and discussed.
4. A method for building a deep learning architecture with multi-layer structure is proposed. The method of convolutional neural network is used to introduce the mature deep learning network architecture into the multi-layer processing structure. The stable structure and weight settings are utilized to change the control purpose and achieve optimal output under the condition of less computing resources. The accuracy of the results is evaluated quantitatively.

並列關鍵字

Health Care ； Fuzzy Neural Network ； Machine Learning ； SVM ； Random Forest ； Deep Learning

參考文獻

[1]G. A. Leonov and N. V. Kuznetsov(2011), "Algorithms for Searching for Hidden Oscillations in the Aizerman and Kalman Problems，" Doklady Mathematics，Vol.84, No.1,pp.475–481.

[2]R. C. Schank(1991), "Where's the AI," AI magazine. Vol.12, No.4,pp.38.

[3]J. McCarthy and P.J.Hayes (1981), "Some Philosophical Problems from the Standpoint of Artificial Intelligence," Readings in Artificial Intelligence,pp.431-450.

[4]C. M. Bishop(2006), "Pattern Recognition and Machine Learning," Springer.

[5]R. M. Bethea and B.S. SDuran and T. L. Boullion(1985), "Statistical Methods for Engineers and Scientists," New York: Marcel Dekker.