201040756 六、發明說明： f發明所屬之技術領域】 •銻本t日i係有關於一種疾病的中醫智能辨證處方 種能提供醫師或專家意見並學習的中 醫冬能辨證處方系統。 【先前技術】 Φ 案新顆性與進步性更形明確，謹就現有 中醫處方自動化之相關專利之缺點說明如下： 备L蔡^权，’中醫處方自動化之處理方法和 f 書字號：454135。中醫處方自動化之 '、統係依照該病患過去就診之處方給中醫師參 ί二旦遇到未有就診紀錄之病患或該病患之前無 =類似症狀時，則無法提供任何訊息給中醫師參 ^ =以其功能與一般電子病歷系統中常用的複製 刚一筆病患資料之功能無異。除此之外，該先前技 術之中醫處方自動化之處理系統所提供之自動學習 模組係將該病名與症狀之關聯性紀錄於資料庫中， 以作為下次該醫師如果遇到有相同症狀之病患時， 能順利^道為何種病名；但此功能與中醫的診療過 程是相違背的。因為，中醫師是依據病患之症狀辨 證，在決定其究竟為何種證型，然後依此證型決定 應該使用哪些方劑。所以，同一種西醫的疾病對中 醫師而言有可能由不同的證型所產生。其次，縱使 病患是來自於同一種證型’但碟不一定會出現相同 的症狀，故一定要出現相通症狀之自動學習模組對 中醫而言幫助甚小。最後，此自動學習模組並無篩 選功能，將會造成病患越多時，其要選擇的項‘目'也 變多。 2.吳致元、游俊樺，2005 ’ 一種運用階層式架構 201040756 的中醫處方資料處理系統， 1227842。運用階層式架構的 書字號： 的缺點如下：該系統係透過脈 類方式建立階層式查詢所需1 、健保等分 :因為查詢之階層過深造成 以=耗= 自定義、修改及查詢慣動態 學習產生中醫師用藥習慣，故中，法/由土統自我 時間於資料庫建立中。 醫$必須花費許多201040756 VI. Description of the invention: The technical field to which the invention belongs 】 锑 t i i 有 有 有 有 有 i i i i 中 中 中 中 中 中 中 中 中 中 中 中 中 中 中 中 中 中 中 中 中 中 中 中 中 中 中 中 中 中[Prior Art] Φ The new and progressive nature of the case is more clear. I would like to explain the shortcomings of the existing patents related to the automation of traditional Chinese medicine prescriptions as follows: Prepare L Cai ^ right, 'Processing method for Chinese medicine prescription automation and f book font size: 454135. "Traditional Chinese medicine prescriptions", according to the patient's past visits to the Chinese medicine practitioners who met the patient's record without the patient's record or the patient's previous symptoms without the same symptoms, can not provide any information to Chinese medicine practitioners Shishen ^ = with its function is similar to the function commonly used in the general electronic medical record system to copy just a patient's data. In addition, the automatic learning module provided by the prior art TCM prescription automation processing system records the association between the disease name and the symptom in the database as the next time the physician encounters the same symptom. When the patient is sick, what kind of disease name can be successfully passed; but this function is contrary to the diagnosis and treatment process of Chinese medicine. Because the Chinese medicine practitioner is based on the symptoms of the patient to determine the type of syndrome, and then determine which prescription should be used according to the type of syndrome. Therefore, the same Western medicine disease may be produced by different syndromes for Chinese medicine practitioners. Secondly, even if the patient is from the same syndrome, but the disc does not necessarily have the same symptoms, the automatic learning module that must have the same symptoms is of little help to Chinese medicine. Finally, this automatic learning module does not have a screening function, which will result in more items to be selected. 2. Wu Zhiyuan, You Junhua, 2005 ’ A Chinese medicine prescription data processing system using the hierarchical structure 201040756, 1227842. The disadvantages of using the hierarchical font size of the book are as follows: The system is required to establish a hierarchical query through the pulse type. Health insurance equals: because the level of the query is too deep, the consumption is too low = custom = custom, modified and query Learning to produce Chinese medicine practitioners' habits, so the Chinese/French/Traditional Self-Time is established in the database. Medical doctor must spend a lot
醫ί療目資前醫醫療資訊系統還是沿用西 式’主要是利用疾病分類、國 依：用次數將常用的處方集出現在前5。；是 上if η式严不太適用在中醫醫療資訊系統 级期丨雙主？原因疋.相同之疾病分類碼下，中醫的 類卻有非常多種，再加上不同之中醫理論就 。所以，造成中醫師在使用中醫 ^療貝訊系統時，必須花費很多的時間在鍵入用藥 >訊上，進而減少與病患面對面問診之時間。 故，如何將上述之問題加以摒除，實為本案發明人 所欲解決之技術困難點之所在。 【發明内容】 次本發明之發明人有鑒於上述缺失，乃搜集相關 資料，經由多方評估及考量，並以此領域累積之多 年經驗，經由不斷試作及修改，始成功研發出本發 明之中醫智能處方集系統，係包含： 一種中醫智能處方集系統，用於提供醫師或專家意見，係包含: 一中醫四診資料輸入系統； 一中醫辅助辨證系統； 一以資料探勘處方知識庫為主之中醫自動處方集系統； 5 201040756 一以中醫師處方專家庫為主之中醫 一中醫自動處方隼李蛴之眘祖^ 動處方集系統， 統； 力杲系統之貝枓探勘處方知識庫產生系 - 自動處方集系統之中醫師 一中醫辨證編碼與處方的比針描 f豕庫產生系統， 證編碼與處方之比；㈣賴組，_料算中醫辨 一J料探勘辨證知識庫或—中醫師辨. :資料探勘處方知識庫或一中醫師處 ： f係經由該中醫四診資料輸入系 辅助辨證系統依據使用者所㈣病患症狀中醫 處方知識庫或中醫師處方專H:料探勘 =的比對模組將中醫 ，及相似性量數，自動辨識出合適之處方集輸出至中醫 ▲自動處方及開方系統，而後中醫師可依照系^统 議處方集進行修改’最後可將修改資^於病； 〇 2透過資料探勘程序將醫師辨證及處方 知哄儲存於資料探勘辨證知識庫及資料探勘處方 ΐ Ξ ί t,以1气續令醫自動處方及開方系統使用。而 二能處方集系，统亦可選擇將病患病μ資料庫之透 豕知識呈現程序將醫師辨證及處方過程分別儲 ，中醫師辨證專家庫及中醫師處方專家庫中，供給中醫 自動處方及開方系統使用。 【實施方式】 為達成本發明上述之各項目的以下謹配合圖式， 舉一較佳實施例如下： (一）前言： 201040756 本發明之中醫智能處方集系統係運用資料探勘程 配合「中醫標準化症狀分類」及「中醫辨證編碼系統 (B_Code)」，藉由資料探勘程序中的數個演算法對 辨證及中醫臨床處方進行資料探勘，並藉此找到推颊 =並儲存於資料探勘辨證知識庫及資料探勘處方知 士中。此外，本發明之中醫智能處方集系統亦可運用^ 二合「、中醫標準化症狀分類」及「中醫 ^聯程度對中醫辨證及中醫臨床處方進行計算，並= 規則並儲存於中醫師辨證專家庫及中醫師處 用明之分析結果實證後可做為中醫臨床使 方集推論引擎做為建立中醫自動處方集 =2庫：===45、以資料探勘處 方專家庫為主之= 120、以中醫師處 ❹ Γ:處ΛΓ對模Γι1。二 之說(明二:)中醫四診資料輸人系統⑽（請參閱：圖2) - 藉固由：病患自覺症狀勾選輸入介面」200(請 同時提供舌吟入病患自我症狀，此輸入介面並 面j例如舌沴儀）及脈診（或脈診儀）輸入之介 步驟2-1 ”病患擁有·項強，目艮睛乾溫，羞明，眼睛酸溫流 7 201040756 涕，鼻窒，清涕，鼻燥，腦、耳鳴，齦腫痛咽喉痛口乾， 胸悶，咳嗽不暢，白痰，腹脹，腹痛，嘔吐，肩頸痠痛，肌膚 發癢，等幾項症狀，則醫師或病患可藉由「病患自覺症 狀勾選輸入介面」200輪入病患自我症狀，或透過「病 患自覺症狀文字輸入介面」21〇輸入。（請參閱：圖 步驟2-2 ’ 某位病患擁有··舌色紅，白苔，苔薄，舌色紫無苔，澀 脈一左關，弦脈—左關，澀脈_右關，弦脈—右關，則’可°藉由 本系統提供舌診（或舌診儀輔助介面22〇)及脈診（或脈 診儀辅助介面230)輸入之介面，將醫師、舌診儀. 診儀判斷之結果藉由「病患自覺症狀勾選輸入介 m 輸入病患自我症狀。（請參閱：圖2-1) 步驟2-3 當患者或醫師覺得藉由「病患自覺症狀勾選輸入介 面」200之標準化症狀勾選無法完全將患者病情很完整 的描述，則可將其症狀描述於「其他症狀」這個攔位 (請參閱：圖2-1) 而系統會主動透過「中醫標準化症狀同義詞庫」24〇 將「其他症狀」攔位中之主訴轉換成電腦可以處理之症 狀編碼，此步驟為將病患及醫師輸入非標準化症狀透過 症狀同義詞庫轉換成標準症狀sl〇1，並記入病患病歷 資料庫260中。例如··在「其他症狀」攔位中輸入「咽 乾」則系統會主動轉為「口乾」這個症狀編碼，以讓系 統在症狀之處理有相同用語’圖2_2即為一例示。 (三）第3圖為中醫輔助辨證系統實施例： 中醫辅助辨證系統包含病患病歷資料庫3〇〇、相異 性量數與相似性量數310、線性迴歸、L〇GISTIC迴歸、 ，類樹320 (上述演算分析法將於下文詳述）、資料探 勘辨證知識庫330、中醫師辨證專家庫34〇、中醫辨證 201040756 編碼B_Code輸出並供給中醫自動處方集系統使用 350，以及中醫師判斷中醫辨證證型是否正'確 一 ^病患症狀輸入之後，則「中醫輔助辨證 經由中醫師專家庫34〇或資料探勘知識庫33〇 '提供「 證結果及病患食療建議」（請參閱：圖3 ^。 ’、 产:t ί位主H Γ項強’，睛乾激，羞明，眼睛酸溫， ❹Medical treatment of medical information systems is still in the Western style. The main use of disease classification, national dependence: the number of commonly used prescriptions appear in the top 5. Is Yes on η-type strict not applicable in the Chinese medical information system? The reason is that under the same disease classification code, there are many kinds of Chinese medicine, plus different TCM theories. Therefore, it is necessary for Chinese medicine practitioners to spend a lot of time typing in medications when using Chinese medicine to cure the Beixun system, thereby reducing the time for face-to-face consultation with patients. Therefore, how to eliminate the above problems is the technical difficulty point that the inventor of the case wants to solve. SUMMARY OF THE INVENTION In view of the above-mentioned deficiencies, the inventors of the present invention have collected relevant materials, and through various evaluations and considerations, and accumulated many years of experience in this field, through continuous trial and modification, successfully developed the TCM intelligence of the present invention. The formulary system includes: a TCM intelligent formulary system for providing physician or expert opinions, including: a Chinese medicine four diagnostic data input system; a Chinese medicine assisted syndrome differentiation system; a data exploration prescription knowledge base mainly for Chinese medicine practitioners Automatic formulary system; 5 201040756 A Chinese medicine practitioner's prescription expert library mainly Chinese medicine one Chinese medicine automatic prescription 隼 Li Weizhi Shenzu ^ dynamic formula system, system; 杲 杲 之 枓 枓 枓 枓 枓 - - - - - - - In the system, the ratio of the code of the doctor to the TCM syndrome code and the prescription is the ratio of the code to the prescription, and the ratio of the code to the prescription; (4) Lai group, _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ Data exploration prescription knowledge base or a Chinese medicine practitioner: f is based on the use of the Chinese medicine four diagnostic data input system to assist the syndrome system (4) Symptoms of TCM TCM Prescription Knowledge Base or TCM Prescription Special H: Material Exploration = Comparison Module will automatically identify the appropriate set of TCM and similarity quantities to Chinese medicine ▲ Automated prescription and prescription system The Chinese medicine practitioner can then modify the prescription according to the system. 'The last modification can be used for the disease; 〇2 The doctor's syndrome and prescription knowledge are stored in the data exploration and knowledge base and the data exploration prescription through the data exploration program. ί t, with 1 gas continued to use the automatic prescription and prescription system. The second prescription system can also choose to store the doctor's syndrome differentiation and prescription process separately through the knowledge disclosure program of the patient's disease database. The Chinese medicine doctor's syndrome expert library and the Chinese medicine practitioner prescription expert library are provided to the Chinese medicine automatic prescription. And the use of the system. [Embodiment] In order to achieve the above-mentioned items of the present invention, a preferred embodiment is as follows: (1) Foreword: 201040756 The TCM intelligent prescription system of the present invention uses a data exploration process to cooperate with "Chinese medicine standardization" "Symptom Classification" and "Chinese Medicine Dialectical Coding System (B_Code)", through a number of algorithms in the data exploration program to conduct data exploration of syndrome differentiation and TCM clinical prescriptions, and find the cheeks = and store in the data mining knowledge base And data exploration prescriptions. In addition, the TCM intelligent formulary system of the present invention can also use the "two-in-one", "standardized symptom classification of Chinese medicine" and "the degree of TCM" to calculate the TCM syndrome differentiation and the clinical prescription of Chinese medicine, and the rule is stored in the TCM syndrome expert database. And the Chinese medicine practitioners can use the analysis results of the analysis to be the TCM clinical remedy inference set engine as the establishment of TCM automatic prescription set = 2 library: ===45, based on the data exploration prescription expert library = 120, to Chinese medicine师处❹ Γ: ΛΓ ΛΓ ΛΓ Γ 1 1 。 。 。 。 。 Γ ( ( ( ( ( Γ Γ Γ Γ Γ Γ Γ Γ Γ Γ Γ Γ Γ Γ 明 明 明 明 明 明 明 明 明 明 明 明 明 明 明 明 明 明 明 明 明 明 明 明 明Please also provide self-symptoms of the tongue into the patient, this input interface is j-parallel, for example, the tongue-and-mouth device, and the pulse diagnosis (or pulse diagnosis instrument) is introduced. Step 2-1 "The patient has a strong, eye-catching Warm, shy, eye acid temperature flow 7 201040756 涕, nose, clear throat, nasal dryness, brain, tinnitus, phlegm, sore throat, dry mouth, chest tightness, poor cough, white diarrhea, abdominal distension, abdominal pain, vomiting, shoulder and neck Sore, itchy skin, and several other symptoms, then the physician Patients can by "conscious patients symptoms tick input interface" 200 into the patient self symptoms, or "suffering from disease symptoms text input interface" through 21〇 input. (See: Figure 2-2 'A patient has a red tongue, white fur, thin fur, purple tongue, no moss, a vein left, a pulse, a left pulse, a vein, a right , chord pulse - right off, then 'by the system to provide a tongue diagnosis (or tongue diagnosis device auxiliary interface 22 〇) and pulse diagnosis (or pulse diagnosis device auxiliary interface 230) input interface, the physician, tongue diagnosis. The result of the diagnosis is to enter the patient's self-symptom by using the patient's self-consciousness check box. (See: Figure 2-1) Step 2-3 When the patient or physician feels that the patient's self-conscious symptoms are selected The standardized symptom of the input interface 200 can not completely describe the patient's condition, and the symptoms can be described in the "other symptoms" block (see: Figure 2-1) and the system will actively adopt the "Chinese medicine standardization". The Symptoms Synonym "24" converts the complaint in the "Other Symptoms" block into a symptom code that can be processed by the computer. This step is to convert the patient and the physician into non-standardized symptoms through the symptom thesaurus into the standard symptom sl1, and Recorded in the patient medical record database 260. For example, In other symptoms, if you enter "stomach dry" in the block, the system will actively switch to the "dry mouth" symptom code, so that the system has the same language in the treatment of symptoms. Figure 2_2 shows an example. (3) Figure 3 shows Examples of TCM-assisted syndrome differentiation system: TCM-assisted syndrome differentiation system includes patient medical record database 3, dissimilarity quantity and similarity quantity 310, linear regression, L〇GISTIC regression, and class tree 320 (the above calculation analysis method will Detailed description below), data exploration and syndrome knowledge base 330, TCM syndrome expert library 34〇, TCM syndrome 201040756 code B_Code output and supply to the Chinese medicine automatic formulary system use 350, and Chinese medicine doctors judge whether the TCM syndrome type is positive ^ After the patient's symptoms are entered, the "Traditional Chinese Medicine Assisted Syndrome provides the "Certificate Results and Patient Diet Recommendations" via the Physician's Expert Library 34 or the Data Exploration Knowledge Base 33" (see: Figure 3 ^. ', Production: t ί主主H Γ项强', the eyes are dry, shy, eyes are sour, ❹
22’22，：痰，腹脹’腹痛’喂吐，肩頸痠痛，肌1’ 發癢，等幾項症狀；舌診儀判斷為舌色紅，白苔，苔薄專 舌色紫，無苔；脈診儀判斷左右關皆為澀、。 步驟3-2 中醫輔助辨證系統會依據醫師所選擇之知識 料探勘辨證知識庫330或中醫師辨證專家庫34〇)判 病患之證型，並經由相異性量數與相似性量數31〇 ; 性迴歸、LOGISTIC迴歸、分類樹32〇依照演算法判斷， 例如：當醫師選擇資料探勘辨證知識庫33〇時，系 動判斷病患證型為：企虛、陰虚、濕。（請參 步驟3-3 1； 當中醫輔助辨證系統自動判斷病患證型後會主動 病患證型輸入於「中醫辨證輸入系統」中，並主動 醫辨證編碼B—Code為：第一碼：濕、第二碼：無: ，三碼··血、陰、第四碼：虛（虧）。（請參二 步驟3-4 醫師可於「中醫辨證輸入系統」判斷病患證型及 醫辨證編碼B_Code是否正確（步驟S3〇1)，以符合 醫師臨床辨證判斷，並可做為日後資料探勘知識庫 =練之樣本，確認後輸出至r中醫辨證編碼B—c〇de 」，並供給中醫自動處方集使用』350。例如：醫師辦 減後之第一碼：濕，熱、第二碼：脾，肝腎、第三碼：血曰 9 201040756 氣陰、第四碼：虛（虧），痹，瘀。（請參閱：圖3-2) 上述資料探勘辨證知識庫330所使用之線性迴歸、 LOGISTIC迴歸、分類樹320等運用方式說明如下： 1. 線性迴歸 資料探勘知識庫系統可以利用「線性迴歸」來 估計線性方程式的係數，此線性方程式跟一個或多 個自變數有關（這些自變數可以正確地預測依變數 值）。例如’可以根據患者症狀、舌診、脈診之類的 自變數’來預測中醫辨證編碼B_Code (依變數）是 否可能出現。 2. Logistic 迴歸 中醫辨證編瑪B_C〇de 或「不出現」時，Log 資料探勘知識庫系統亦可以利用rL〇gistic迴 歸」來估計線性方程式的係數，當使用者希望能夠 根據患者症狀、舌診、脈診之類的自變數，來預測 ode (依變數）是否可能「出現」 Logistic迴歸分析就报有用。它 和線性迴歸模式很相似，但是更適合二分依變數的 模式。Logistic迴歸係數可以用來估計模式中每一 個自變數的odds比率，藉此odds比率可以判斷 是否會出現。這是本發㈣統的預設值。 分類樹 類樹」會建立樹狀結構的分類模式，它會 根據患者症狀、舌診、脈診之類的自#赵技跑资括22'22,: 痰, bloating 'abdominal pain', vomiting, shoulder and neck pain, muscle 1' itching, and several other symptoms; tongue diagnosis is judged to be tongue red, white moss, moss thin tongue purple, no moss The pulse diagnosis device judges that the left and right sides are all 涩. Step 3-2 The TCM-assisted syndrome differentiation system will judge the syndrome type of the patient according to the knowledge material exploration knowledge base 330 selected by the physician or the TCM syndrome syndrome library 34〇, and pass the dissimilarity quantity and similarity quantity 31〇 Sexual regression, LOGISTIC regression, and classification tree 32〇 are judged according to the algorithm. For example, when the physician chooses the data exploration and identification knowledge base 33〇, the system determines that the patient's syndrome type is: deficiency, yin deficiency, and dampness. (Please refer to step 3-3 1; When the TCM-assisted syndrome system automatically judges the patient's syndrome type, the active patient syndrome type will be entered in the "Traditional Chinese Medicine Dialectical Input System", and the active medical syndrome code B-Code is: the first code : Wet, second code: none: , three yards · blood, yin, fourth code: virtual (loss). (Please refer to step 3-4. Physicians can judge the patient's syndrome type and "Traditional Chinese Medicine Dialectical Input System" Whether the medical syndrome code B_Code is correct (step S3〇1), in accordance with the doctor's clinical dialectical judgment, and can be used as a sample of the future data exploration knowledge base = training, and output to the Chinese medicine syndrome code B-c〇de after confirmation, and Supply TCM automatic prescription set to use 350. For example: the first code after the doctor reduced: wet, hot, second code: spleen, liver and kidney, third code: blood stasis 9 201040756 Qi Yin, fourth code: virtual (deficit ), 痹, 瘀. (See: Figure 3-2) The linear regression, LOGISTIC regression, and classification tree 320 used in the above-mentioned data exploration and knowledge base 330 are described as follows: 1. The linear regression data exploration knowledge base system can be Use Linear Regression to estimate linear squares The coefficient of the program, which is related to one or more independent variables (these independent variables can correctly predict the dependent value). For example, 'the syndrome can be predicted according to the patient's symptoms, tongue diagnosis, pulse diagnosis, etc.' Whether the code B_Code (depending on the variable) is possible. 2. When the Logistic regression TCM syndrome compilation B_C〇de or "does not appear", the Log data exploration knowledge base system can also use rL〇gistic regression to estimate the coefficient of the linear equation. The user wants to be able to predict whether an ode (variable) is likely to "appear" based on the patient's symptoms, tongue diagnosis, pulse diagnosis, etc. Logistic regression analysis is useful. It is similar to the linear regression model, but more suitable for binary According to the mode of the variable, the logistic regression coefficient can be used to estimate the odds ratio of each self-variable in the mode, and the odds ratio can be used to determine whether it will appear. This is the default value of the (4) system. The classification tree tree will be established. The classification pattern of the tree structure, which will be based on the patient's symptoms, tongue diagnosis, pulse diagnosis, etc.
201040756 l ·相異性量數 相異性量數可以用來衡量症狀與中醫辨證編碼 B—Code之相異狀況，如果相異狀況愈低則症狀盥 B—Code的關係就愈大。資料探勘知識庫33〇估' 下列相異性量數： a. 歐基里得直線距離。這是從四重表格中依 SQRT(b+c)所計算得來的值，其中b和c代表 對角線上’對應到觀察值出現在某一項目、但未 出現在另一項目的儲存格。 b. 歐基里得直線距離平方。計算得到的是不一致觀 ° 察值的數目，它的最小值為〇,但無最大值上限。 c. 形式差異係數。範圍介於〇到1之間的二元資 料相異性量數。這是從四重表格中依bc/(n料2) 所計算出來的係數，其中b和C代表對角線 上，對應到觀察值出現在某一項目、但未出現在 另一項目的儲存格，而n是觀察值的總個數。 d. 變異數。這是從四重表格中依（b+c)/4n所計算 出來的係數，其中b和c代表對角線上，對應 到觀察值出現在某一項目、但未出現在另一項目 ❹ 的儲存格，而11是觀察值總個數。其範圍介於〇 到1之間。 、 e. Lance與Williams係數。這是從四重表格中依 (b+c)/(2a+b+c)所計算出來的係數，其中a代 表對應至觀察值、且同時出現在兩個項目中的儲 存格，而b和c代表對角線上，對應到觀察值 出現在某一項目、但未出現在另一項目的儲存 格。這個量數的範圍介於〇到1之間。 2.相似性量數 相似性量數可以用來衡量症狀與中醫辨證編碼 B_Code之相似狀況，如果相似狀況愈高則症狀與 11 201040756 B—Code的關係就愈大。資料探勘知識庫330可使用 下列相似性量數： a. Russel與Rao。這是二元的内部（點）運算結 果。其中，相符和不相符的加權值相等。 b. 簡單配對係數。為相符者佔總值的比率。其中， 相符和不相符的加權值相等。這是本系統的'預設 值。 c. Jaccard相似性比。這是一個係數，執行時，該 係數就已經排除聯合不出現的情況。其中，相符 和不相符的加權值相等。這個選項也叫做相似性 比。 * ° a _ neath係數。此係數是以條件機率為 這個機率指的是’某個項目的特性正好符 二2項目中的數值。如果某個項目被當成其他 2 =預測值，那麼在使用者算出這些項目個別 f的平均數之後，就可以計算出這個數值。 數的計算方式是，將相符者個數減 二不C間再除以總項目個數。其範圍介 值^這個值是G〇〇dman和KruSkal的 ambda值。它會使用一個項目來一 方^預測），來對應到誤 (PRE)。其範圍介於〇到1之間。 g-^le’s γ係數。此索引為如0表格的交 Ϊ間而且它與邊際總和無關。其範圍介☆ -1到, (四）圖4為本發明圖中醫自 方知識庫產生系統147之自細動處=201040756 l · The number of dissimilarity The amount of dissimilarity can be used to measure the difference between the symptom and the T-code of the TCM syndrome. If the difference is different, the relationship between the symptom and the B-Code will be greater. Data Exploration Knowledge Base 33 estimates 'The following dissimilarity quantities: a. Euclid's linear distance. This is the value calculated from the quadruple table according to SQRT(b+c), where b and c represent the cells on the diagonal that correspond to the observed value appearing in one item but not appearing in another item. . b. Euclid's linear distance squared. The calculated number of inconsistent observations is 最小值, but there is no upper limit. c. Formal difference coefficient. The amount of binary data dissimilarity ranging from 〇 to 1. This is the coefficient calculated from the quadruple table according to bc/(n material 2), where b and C represent the diagonal, corresponding to the cell in which the observed value appears in one item but does not appear in another item. And n is the total number of observations. d. The number of variations. This is the coefficient calculated from (b+c)/4n in the quadruple table, where b and c represent the diagonal, corresponding to the observation that the observation appears in one item but does not appear in another item. Lattice, and 11 is the total number of observations. Its range is from 〇 to 1. , e. Lance and Williams coefficients. This is the coefficient calculated from (b+c)/(2a+b+c) from the quadruple table, where a represents the cell corresponding to the observed value and appears in both items simultaneously, and b and c represents the diagonal, corresponding to the cell where the observed value appears in one item but does not appear in another item. This quantity ranges from 〇 to 1. 2. Similarity measures The similarity measure can be used to measure the similarity between symptoms and T_ syndrome code B_Code. If the similarity is higher, the relationship between symptoms and 11 201040756 B-Code will be greater. The data mining knowledge base 330 can use the following similarity quantities: a. Russel and Rao. This is the result of a binary internal (dot) operation. Among them, the matching and non-conforming weights are equal. b. Simple pairing factor. The ratio of the total to the total. Where the matching and non-conforming weighting values are equal. This is the 'preset value of this system. c. Jaccard similarity ratio. This is a coefficient that, when executed, has ruled out that the union does not occur. Among them, the matching and non-conforming weights are equal. This option is also called the similarity ratio. * ° a _ neath coefficient. This coefficient is based on the conditional probability. This probability refers to the value of the 'characteristics of a certain item'. If an item is treated as the other 2 = predicted value, then the user can calculate this value after calculating the average number of individual f of these items. The number is calculated by subtracting the number of matches from two and dividing by C and dividing by the total number of items. Its range value ^ is the ambda value of G〇〇dman and KruSkal. It uses a project to predict (predict) to correspond to the error (PRE). Its range is between 〇 and 1. G-^le’s γ coefficient. This index is between the intersections of the 0 table and it is independent of the marginal sum. The scope is ☆ -1 to, (4) Figure 4 is the self-fine action of the Chinese medicine knowledge base generation system 147 of the present invention =
St:料；Z及資料探勘處方知識庫“Ο。 、有·尋找各疾病別之中醫證類特性剛、建= 12 201040756 目標資料集S402、清理不齊全之資料S403、將中醫證 型轉換成中醫辨證編碼B-Code S404、階層式集群、K 平均數集群S405、配合中醫師經驗決定群體數S406、 中醫師詮釋各中醫證類名稱S407。 實施例： 系統管理者可以設定每日一固定時間，要求系統主 動將病患病歷資料庫430中之病患資料進行分析，其分 析步驟如下： 步驟4-1 針對不同疾病別，從病患病歷資料庫430中選取同 〇 一疾病之資料尋找各種疾病別之中醫證類特性 (S401)，建立目標資料集（步驟：S402)，然後清理資 料不齊全之資料（步驟：S403 )後，將中醫證型轉換成 中醫辨證編碼B_Code (步驟：S404)，然後再從至少超 過一門權比率之B_Code納入分析。例如：某一疾病超 過5°/◦門檻比率之B_Code有陰虛，熱，濕，痹，血虛，瘀，氣 虛，肝，腎，厥，水飲，風，鬱滯，肺，胃，陽虛，失制，脾，皮， 大腸，動風，那系統就會拿這些資料進行分析。 步驟4-2 〇 之後系統會依照管理者要求之資料探勘工具對中醫 辨證編碼B_Code進行集群分析（步驟：S405)，中醫師 可以將此B_Code集群視為中醫證類（簡稱：B_Code證 類）（步驟：S406 )。例如：要求系統將B_Code資料分 成10群，則這10群之結果即為B_Code證類卜B_Code 證類10。 步驟4-3 中醫師將各中醫證類（B_Code證類）進行詮釋（步驟： S407 )或由系統直接將B_Code證類之各B_Code中心值 是否大於某一比率，如果大於某一比率則視為該 B_Code為該證類主要特性。並可進一步設定該證類之 13 201040756 單”可由系統將此B-c〇de證類出現 方知識康At早，為預設用藥並放入資料探勘處 有^虎1 BJ：〇de證類1中出現超過50%iB-Code 浐ί 1 ’dTo!濕，痹故其為該證類主要特性。而B-c〇de 二;現超過6〇%病患常用之方劑為龍膽填肝湯、 知柏地黃丸、甘露飲、防己。（如圖4一丨） 两 统處方集系統之資料探勘處方知識庫產生系 現所運用方法說明如下： 1.階層集群分析 階層集群分析會根據使用者所選取的中醫辨镫 編碼B—Code，試圖找出具有相對同質性的B_c〇d2 組別。它所使用的演算法，會從個別集群中的每一 個B_Code開始，然後再與集群組合，直到只剩下一 個為止。系統可以分析原始資料（預設），或從各種 不同的標準化轉換中選擇。「相似性」程序會產生距 離或相似性測量。每個階段都會顯示相似性測量， 以協助使用者選出最適用的集群數目K，並藉此資 訊瞭解各B—Code集群為何種β—Code證類。中醫師 可，將此B_Code證類設定應該使用那些合適方劑 或單位藥’亦可由系統將此、(^心證類出現頻率之 方劑或單位藥做為預設用藥。 2. K平均數集群分析 使用者可以利用這個程序，讓演算法（該法可 以處理大量觀察值）根據所選用的中醫辨證編碼 B_Code ’找出性質相似的B_c〇de群組。但是，此分 析方法會要求使用者指定集群數目κ。如果使用者& 道這個資訊’便可以指定初始B_Code集群中心然後 反覆更新各B一Code集群中心。使用者可以儲存 B—Code集群組員、距離資訊、和最後集群的 201040756 二i。^訊瞭解各B-Code集群為何種B—Code 出現頻率之方劑或單位藥撤2統將”-code也類 集群分析是本為預設用藥。κ平均數 家上產)生H顯：自動處方集系統之中醫師處方專 510、病串症解次含有：病患病歷資料庫之處方資料 豕庫530,以及步驟851〇到858斗〇。2〇中醫師處方專 〇 ♦统斤醫師可以要求系統配合中醫師經驗及 分:集系統之中醫師處方專家庫產生系統，其 步驟5-1 方=5中病別’從病患病歷資料庫之處 樂所治療之中醫辨證編碼B-C〇de為； 〇 ίΐΐι ΐ比率之[⑽納人分析，將方劑或單i 醫證型轉成B-code(步驟：S530)。例如： d λ一疾病超過5%門檻比率之❿鈕有陰虛， 胃、’ 生生f’齋，氣虛，肝’腎，厥，水飲，風，鬱滞，肺 ί進陽行1分ί，脾，皮，大腸，動風，那系統就會拿這些資 步驟5-2 执^中〇醫師指定之疾病別’從病患病歷資料庫之辨 ί換ΞI Γ疾病別及醫師別查詢病患辨證資料並 、士 B—Code (步驟：s54〇：)，從資料庫中選取同一 料，然後清理資料不齊全之資料後，將至少超過 ]檻比率之方劑與單位藥納入分析，並製作各中醫師 15 201040756 指定某-疾病超過5%門檻比率之單位藥龍膽師 甘露请毒丹，甘露飲，知柏地黃丸 前、生龍膽篇肝湯: 豬荼湯，秦完鱉甲湯，銀魅散，那李统七^:璉子飲， 行分析。 趙取邵系統就會拿這些資料進 步驟5-3 驟運包含：相異性量數或相似性量數）（步 ，如=:狀=高則B』。de與方劑的關 t 果大於某一比率則視為該B Code之常用方櫥= 早位藥，也就是該方劑所能治療證 = d #師_該療證型 患常；之方劑九二·肝B:cod: ί滿熱出現超過⑽病 是否需修正（步驟：糊，若中醫心要斷 則進行或單味藥所能治療之中㈣>， 类(步驟：s 5 9 0〉，而中醫師覺得系統所分析之结 Ϊ =相：：：其建立在中醫師處方專家庫53〇 Ϊ 5-ΐ)1理論做為後續參考。（如 巧統運關聯程度之衡量方法有兩類， 性量數及相似性量數，其定義如下： 尺傾兴 1.相異性量數 相異性量數可以用來衡量中醫辨證編碼B 與方劑之相異狀況，如果相異狀況愈低則B—c〇de與 方劑的關係就愈大。系統可使用下列相異性量數 a.歐基里得直線距離。這是從四重表格中依 SQRT(b+c)所計算得來的值，其中b和c代表 對角線上，對應到觀察值出現在某一項目、但未 201040756 出現在另一項目的儲存格。 察值的數目，它的最小值為〇 I 致觀 C·形式差異係數。範圍介於〇到「^^限: 料相異性量數。這是從四重表格中依St: material; Z and data exploration prescription knowledge base "Ο., have · looking for the characteristics of various diseases, TCM syndromes, just built = 12 201040756 target data set S402, clean up incomplete information S403, convert TCM syndromes into TCM syndrome code B-Code S404, hierarchical cluster, K average cluster S405, with the Chinese physician experience to determine the number of groups S406, TCM interpretation of each TCM certificate class name S407. Example: System administrator can set a fixed time per day The system is required to actively analyze the patient data in the patient medical record database 430, and the analysis steps are as follows: Step 4-1 For different diseases, select the data of the same disease from the patient medical record database 430 to find various The TCM syndrome type of the disease (S401), establish the target data set (step: S402), and then clean up the incomplete data (step: S403), and convert the TCM syndrome into the TCM syndrome code B_Code (step: S404) And then from the B_Code of at least one door weight ratio into the analysis. For example: B_Code of a disease exceeding 5 ° / threshold rate has yin deficiency, heat, humidity, phlegm, blood Virtual, phlegm, qi deficiency, liver, kidney, phlegm, water drink, wind, stagnation, lung, stomach, yang deficiency, disorder, spleen, skin, large intestine, dynamic wind, then the system will take these data for analysis. After 4-2, the system will analyze the T_ syndrome code B_Code according to the data exploration tool requested by the administrator (step: S405). The Chinese medicine practitioner can regard this B_Code cluster as a TCM certificate (abbreviation: B_Code certificate) (step :S406). For example, if the system is required to divide the B_Code data into 10 groups, the result of the 10 groups is the B_Code certificate class B_Code certificate class 10. Step 4-3 The Chinese medicine practitioner interprets each TCM certificate class (B_Code certificate class) (Step: S407) Or whether the B_Code center value of the B_Code certificate class is directly greater than a certain ratio by the system. If it is greater than a certain ratio, the B_Code is regarded as the main characteristic of the certificate class. 201040756 Single "can be used by the system to make this Bc〇de syndrome appear in the knowledge of At At, as the default medication and put into the data exploration department. ^虎1 BJ: 〇de syndrome class 1 appears more than 50% iB-Code 浐ί 1 'dTo! Wet, why is it? The main characteristics of the class. And B-c〇de 2; now more than 6〇% of the commonly used prescriptions for the patients are Longdan liver filling soup, Zhibai Dihuang Wan, Ganlu drink, anti-self. (Figure 4) The data mining and prescription knowledge base of the two system of prescription system is described as follows: 1. Hierarchical cluster analysis The cluster analysis will be based on the Chinese doctor's identification code B-Code. Trying to find the B_c〇d2 group with relative homogeneity. The algorithm it uses starts with each B_Code in the individual cluster and then combines with the cluster until there is only one left. The system can analyze raw data (presets) or choose from a variety of standardized conversions. The Similarity program produces distance or similarity measures. Similarity measures are displayed at each stage to assist the user in selecting the most appropriate number of clusters K and to learn what type of β-Code certificate each B-Code cluster is. Chinese medicine practitioners can set the B_Code certificate class to use those appropriate prescriptions or unit drugs'. This can also be used by the system, (the prescription of the frequency of the heart syndrome or the unit drug as the default drug. 2. K average cluster analysis Users can use this program to let the algorithm (which can handle a large number of observations) to find B_c〇de groups of similar nature according to the selected TCM syndrome code B_Code '. However, this analysis method will require the user to specify the cluster. The number κ. If the user & channel information can then specify the initial B_Code cluster center and then update each B-code cluster center. Users can store B-Code cluster members, distance information, and the last cluster 201040756 II. ^ News to understand the B-Code cluster for which B-Code frequency of the prescription or unit drug withdrawal 2 will be "-code also cluster analysis is the default medication. κ average number of production" H display: automatic In the formulary system, the physician's prescription special 510, the disease syndrome contains: the patient's medical record database, the data 豕 530, and the steps 851 858 858 〇 〇. Specialized ♦ Physician Physician can request the system to cooperate with the Chinese Physician's experience and points: the system of the physician's prescription expert library generation system, the step 5-1 square = 5 disease 'from the patient's medical record database The TCM syndrome code BC〇de is the (ίΐΐι ΐ ratio [(10) nanoanalysis, the prescription or single i medical certificate is converted into B-code (step: S530). For example: d λ-disease exceeds 5% threshold ratio The button has yin deficiency, stomach, 'shengsheng f' zhai, qi deficiency, liver 'kidney, sputum, water drink, wind, stagnation, lung ί into the line 1 point ί, spleen, skin, large intestine, dynamic wind, that system You will take these steps 5-2 to execute the disease specified by the doctor. 'Recognition from the patient's medical record database ΞI Γ Diseases and physicians do not check the patient's syndrome data and, B-Code (step: S54〇:), select the same material from the database, and then clean up the data with incomplete information, and then include at least the prescription and unit drug of the ratio, and make each physician 15 201040756 Designate a certain disease - more than 5% The threshold medicine unit gentian teacher Ganlu, please poison Dan, Ganlu drink, Zhibai Dihuang Pills before, raw gentian articles liver soup: Pork broth soup, Qin Wan 鳖 汤 soup, silver charm scattered, that Li Tong seven ^: 琏子饮, analysis. Zhao take Shao system will take these data into the steps 5-3 The sudden movement includes: the amount of dissimilarity or the similarity quantity) (step, such as =: shape = high, then B). If the de-party is greater than a certain ratio, it is regarded as the common side of the B Code. Cabinet = early medicine, that is, the prescription can cure = d #师_Therapeutic syndrome is common; the prescription is 92. Liver B:cod: ί Full heat appears more than (10) whether the disease needs correction (step: paste If the heart of the Chinese medicine is to be broken, or the single-drug can be treated (4) >, class (step: s 5 9 0), and the Chinese doctor thinks that the system analyzes the knot = phase::: it is established in the Chinese medicine practitioner The prescription expert library 53〇Ϊ 5-ΐ)1 theory is used as a follow-up reference. (If there are two types of methods for measuring the degree of association of Qiaotong, the number of sexual quantities and the similarity are defined as follows: 1. The amount of dissimilarity can be used to measure TCM syndrome B and prescription. In the case of different conditions, if the difference is different, the relationship between B-c〇de and the prescription is greater. The system can use the following dissimilarity quantity a. Euclid's linear distance. This is from the quadruple table. The value calculated by SQRT(b+c), where b and c represent the diagonal, corresponding to the cell in which the observed value appears in one item but not in 201040756 in another item. The minimum value is 〇I to C. Formal difference coefficient. The range is from “〇^”: the amount of material anisotropy. This is from the quadruple table.
上，對應到觀察值出現在某—項目、I d另變一異項數目的儲4格番而矣n *觀察值的總個數。 d.變異數。迳疋從四重表格中依（b+c)/4n所In the above, the total number of observations corresponding to the observed value appearing in a certain item, I d is changed to a different number of items. d. The number of variations.依 From the quadruple table, rely on (b+c)/4n
出來的係數’其中b和c代表對角線上 到觀察值出現在某一項目、但未出現在另一項目 :儲1存而η是觀察值總個數。其範圍介於〇 e. Lance與Williams係數。這是從四重表格中依 (b+c)/(2a+b+c)所計算出來的係數，其中a代 表對應至觀察值、且同時出現在兩個項目中的儲 存格，而b和c代表對角線上，對應到觀察值 出現在某一項目、但未出現在另一項目的儲存 格。這個量數的範圍介於〇到1之間。 2 ·相似性量數 相似性量數可以用來衡量中醫辨證編碼BCode 與方劑之相似狀況，如果相似狀況愈高則BCode與 方劑的關係就愈大。系統可使用下列相似性量數有： a. Russel與Rao。這是二元的内部（點）運算結 果。其中，相符和不相符的加權值相等。 b. 簡單配對係數。為相符者佔總值的比率。其中， 相符和不相符的加權值相等。這是本系統的預設 值0 c· Jaccard相似性比。這是一個係數，執行時，該 係數就已經排除聯合不出現的情況。其中，相符 17 201040756 和不相符的加權值相等。這個選項也叫做相似性 d. Sokal與Sneath係數。此係數是以條件機率 基礎’這個機率指的是，某個項目的特性正好符、、 合其他項目中的數值。如果某個項目被當成其^ 項目的預測值，那麼在使用者算出這些項目個 值的平均數之後’就可以計算出這個數值。 e. 】amann。此係數的計算方式是，將相符者個數 去不相符者個數，再除以總項目個數。盆 於—1到1之間。 .amh^da值。這個值是Go〇dman和的 ^ambda值。它會使用一個項目來預測另一個 方向都預測），來對應到誤差縮減比例 (PRE)。其範圍介於〇到1之間。 索引為2x2表格的交又比函 之門 匕與邊際t和無關。其範圍介於-1到1 其將1編碼*處方的㈣模組。 mm家庫或資料探勘處方知識庫 中療證型轉換成B-C* (步驟 病歷資料庫61”，將中醫師之辨 :續ϊ r r方射與辨證; 計模型模组可供選擇工具，有關聯程度模組與統 組，=行(步如選擇關聯程度模 獅），接著Am、/平均集群演算（步驟 (步驟S670)，最後呈現社、t醫師辨證關聯度排序 現m果，系統呈現結果並由尹醫 201040756 Ϊ = 藥師配藥（步驟：测)。有關於圖6之 砰細例子將於下文與圖7、圖8 一併詳述之。 (七）圖7為以資料探勘處方知識庫為主 處方集系統運作實施例’包含病患病歷資料庫中7醫〇〇自動資 料探勘處方知識庫71〇，以及步驟S72〇〜S77〇 、 為以中醫師處方專家庫為主之中醫自動4方集系統之 包含：病患病歷資料庫_、中處方、專 810, S820^S870〇* t K 7 ^^^ S720：S730 Ο ❹ i=2?】8l〜S83°相同，圖8中僅有選用『中 醫師處方專豕庫810』之不同，故不予贅述。 俨划本^!重點在於說明中醫師可以選擇依照資料 ，勘處方知識庫710或中醫師處方專家庫81〇進行自動 處方集開方系統之主要依據，並在看診時能夠 助其選擇適當㈣及單位藥。此㈣運0協 步驟6-1 二依照此次看診病患儲存於病患病歷資料庫7〇〇中之 主訴，並將病患病歷資料庫之辨證資料轉成B C〇de(步 驟·· S720、S820)，若缺乏辨證資料則要求中醫師輸入 辨證資料（㈣S730、S830)，另可由中醫辅助辨證系 統或中醫師自行辨證將中醫證型轉成中醫辨證編碼 B_Code代入自動處方集開方系統中。而系統會以該病 患=疾病別，從資料探勘處方知識庫71〇或中醫師處方 專豕庫810中選取與該病患同一疾病之知識進行比對 (步驟：S750〜S770 ; S850〜S870 )。而比對之方法是 依照中醫辨證編碼與處方的比對模組（簡稱比對模組） 來進行，將患者之中醫辨證編碼B_c〇de與資料探勘處 方知識庫710或中醫師處方專家庫81〇中處方可治療之 B_Code進行比對。 ’ 例如：有一門診病患之症狀為項強，眼睛乾澀羞明， 眼睛酸澀，流涕，鼻窒，清涕，鼻燥，腦、耳鳴，齦腫痛，咽 201040756 脈；ΐ色，台，苔薄，舌色紫，無苔，溫 輔助辨说系關，澀脈-右關，弦脈-右關。而經中醫 也系]斷可能B-CGde為陰，血，虛，濕，再經中 修正後為肝，陰，血，虛，濕，熱，‘，再將此 i灸會以該病患之疾病別與資料探勘處方知 y 710或中醫師處方專家庫81〇中處方 之 B-Code進行比對。（如圖6、7、 步驟6-2 驟圖6,當比對的方法為統計模型模組（步 Π〇)時，系統會將患者之中醫辨證編碼B—Code作 ίη ^模^之自變數’帶人圖7之資料探勘處方知識庫 以讓系統自動判斷為何種B一Code證類，及該 B-Code證類所適合之方劑為何。 經Γ患者之B—C〇de為肝，陰，血，虛，濕’熱，痺， 笮肝、β後與B—c〇de證類1相似，故合適之方劑為龍膽 知柏地黃丸、甘露飲、防己，然後系統會自動 =其、、Ό果放入中醫自動處方及開方系統。（如圖81) 步驟6-3 當比對的方法為關聯程度模組（步驟S65〇)時（圖 ’系統會將患者之各個B—Code逐一與圖8之中醫師 專家庫810中各方劑或單位藥所適應之B c〇de進行比 對’讓系統自動判斷該B_Code所適合之方劑為何。 例如：該患者之B—Code為肝，陰，血，虛，濕：妖痺， =比對後與中醫師專家庫中甘露飲之適應B_c〇de為陰 ，濕，熱、知柏地黃丸之適應BCode為下焦，陰，虛， 濕’熱、龍膽;寫肝湯之適應B Code為濕熱相似。故合適 之方劑為龍膽腐肝湯、知柏地黃丸、甘露飲，然後系統 會自動將其結果放入中醫自動處方及開方系統，而後中 醫師可依照系統所給之建議進行修改再加入例如防己 20 201040756 : 藥材。（如圖8-1) 統計模組名詞定義與實際做法： 1.階層集群分析 階層集群分析會根據使用者所選取的中醫辨證 編碼B—Code，試圖找出具有相對同質性的Bc〇de 組別。它所使用的演算法，會從個別集群中的每一 個Β—Code開始’然後再與集群組合，直到只剩下一 個為止。系統可以分析原始資料（預設），或從各種 不同的標準化轉換中選擇。「相似性」程序會產生距 ❹ 離或相似性測量。每個階段都會顯示相似性測量， ^協助使用者選出最適用的集群數目κ，並藉此資 訊瞭解各B—Code集群為何種B—c〇de證類。中醫師 了 =將此B_Code證類設定應該使用那些合適方劑 或單位藥，亦可由系統將此Bc〇de證類出現頻率之 ^劑或單位藥做為預設用藥。當需要比對病患時， ίϊίίί者i中醫辨證編碼B-c〇de作為原分析 貝科之其中一筆，讓集群分析判斷為會與哪一個 證類同一群，即可知道該病患B—c〇de證類 ❹ 2· K平均數集群分析 以虛利用這個程序’讓演算法（該法可 值）根據所選用的中醫辨證編碼The resulting coefficient 'where b and c represent the diagonal line until the observed value appears in one item, but does not appear in another item: storage 1 and η is the total number of observations. Its range is between 〇 e. Lance and Williams coefficients. This is the coefficient calculated from (b+c)/(2a+b+c) from the quadruple table, where a represents the cell corresponding to the observed value and appears in both items simultaneously, and b and c represents the diagonal, corresponding to the cell where the observed value appears in one item but does not appear in another item. This quantity ranges from 〇 to 1. 2 · Similarity measure The similarity measure can be used to measure the similarity between TCode and the prescription of TCM syndrome. If the similarity is higher, the relationship between BCode and prescription is greater. The following similarity quantities can be used by the system: a. Russel and Rao. This is the result of a binary internal (dot) operation. Among them, the matching and non-conforming weights are equal. b. Simple pairing factor. The ratio of the total to the total. Where the matching and non-conforming weighting values are equal. This is the default value of the system 0 c · Jaccard similarity ratio. This is a coefficient that, when executed, has ruled out that the union does not occur. Among them, the match 17 201040756 and the non-conforming weights are equal. This option is also called similarity d. Sokal and Sneath coefficients. This coefficient is based on the conditional probability. This probability refers to the characteristic of an item, and the value in other items. If an item is considered as the predicted value of its ^ item, then the value can be calculated after the user has calculated the average of the values of these items. e. 】amann. This coefficient is calculated by dividing the number of matches to the number of non-conformities and dividing by the total number of items. The basin is between -1 and 1. .amh^da value. This value is the ^ambda value of Go〇dman and . It uses a project to predict the other direction to predict) to correspond to the error reduction ratio (PRE). Its range is between 〇 and 1. The intersection of the index of the 2x2 table is not related to the margin of the letter. It ranges from -1 to 1 and it will encode 1 (pre) (4) modules. In the mm home library or data exploration prescription knowledge base, the type of treatment is converted into BC* (step medical record database 61), which distinguishes the Chinese medicine practitioners: Continued rr rr square shot and syndrome differentiation; the model model module is available for selection and related The degree module and the group, = line (steps such as selecting the degree of association lion), followed by Am, / average cluster calculus (step (step S670), and finally presented the social and t-doctoral correlation degree sorting results, the system presents the results And by Yin doctor 201040756 Ϊ = pharmacist dispensing (step: test). The detailed examples of Figure 6 will be detailed below with Figure 7 and Figure 8. (7) Figure 7 is the data base for data exploration prescription The main prescription system operation example "including the patient medical record database 7 medical automatic data exploration prescription knowledge base 71", and steps S72 〇 ~ S77 〇, for the Chinese medicine practitioners expert library for the main Chinese medicine automatic 4 The inclusion system includes: patient medical record database _, medium prescription, special 810, S820^S870〇* t K 7 ^^^ S720: S730 Ο ❹ i=2?] 8l~S83° the same, only in Figure 8 There is a difference in the use of the "Chinese medicine practitioner's prescription library 810", so it will not be described. ^! The key point is to show that the Chinese medicine practitioners can choose to follow the data, the prescription knowledge base 710 or the Chinese medicine practitioners expert library 81〇 to carry out the main basis of the automatic prescription set prescribing system, and can help them choose the appropriate (four) and unit medicine during the visit. According to the procedure of the patient's medical record database, the patient's medical record is stored in the patient's medical record database, and the syndrome data of the patient's medical record database is converted into BC〇de (step· · S720, S820), if there is a lack of dialectical information, the Chinese medicine practitioner is required to input the dialectical information ((4) S730, S830), and the Chinese medicine aided syndrome differentiation system or the Chinese medicine practitioner can dialect the TCM syndrome type into the Chinese medicine syndrome code B_Code into the automatic prescription set. In the system, the system will compare the knowledge of the same disease with the patient from the data exploration prescription knowledge base 71〇 or the Chinese medicine prescription special library 810 (step: S750~S770; S850~S870). The method of comparison is based on the comparison module of TCM syndrome code and prescription (referred to as comparison module), and the patient TCM syndrome code B_c〇de and data exploration prescription are known. The library 710 or the Chinese medicine practitioner's prescription expert library 81〇 can be compared with the prescription B_Code. ' For example: the symptoms of a patient are strong, the eyes are dry and shy, the eyes are sore, drooling, nasal discharge, clearing, Nasal dryness, brain, tinnitus, phlegm and pain, pharynx 201040756 pulse; twilight, Taiwan, thin fur, purple tongue, no moss, temperature-assisted identification, Guanmai-Right Guan, Xuanmai-Right Guan. After the Chinese medicine is also broken] B-CGde may be yin, blood, imaginary, wet, and then corrected for liver, yin, blood, imaginary, wet, hot, ', then this moxibustion will be the disease The disease is compared with the B-Code of the prescription in the data exploration prescription y 710 or the prescription of the Chinese medicine practitioner prescription 81. (Figure 6, 7 and Step 6-2 Figure 6. When the comparison method is a statistical model module (step), the system will use the patient TCM syndrome code B-Code as the ίη ^模^ The variable 'takes the data exploration prescription knowledge base of Figure 7 to let the system automatically judge what kind of B-code certificate class, and what kind of prescription is suitable for the B-Code certificate class. The B-C〇de of the patient is the liver. Yin, blood, imaginary, wet 'hot, sputum, sputum liver, β and B-c〇de syndrome 1 similar, so the appropriate prescription for gentian Zhibai Dihuang Wan, Ganlu drink, anti-self, then the system will automatically = Its, and the results are placed in the Chinese medicine automatic prescription and prescription system. (Figure 81) Step 6-3 When the comparison method is the association level module (step S65〇) (Figure 'System will be the patient's individual B-Code is compared with the B c〇de of the prescription agent or the unit medicine in the physician expert library 810 in FIG. 8 to let the system automatically determine the prescription of the B_Code. For example: the patient's B-Code For the liver, yin, blood, imaginary, wet: enchanting, = after the comparison with the Chinese herbalist expert library in the nectar drink B_c〇de for yin, wet, hot , Zhibai Dihuang Pills adapt to BCode for the lower focus, yin, virtual, wet 'hot, gentian; written liver soup adaptation B Code is similar to damp heat. So the appropriate prescription for Longdan rot liver soup, Zhibai Dihuang Wan甘露饮, then the system will automatically put the results into the Chinese medicine automatic prescription and prescription system, and then the Chinese Physician can modify according to the recommendations given by the system and then add, for example, Fang 20 201040756 : medicinal materials (Figure 8-1) Module noun definition and practical practice: 1. Hierarchical cluster analysis Hierarchical cluster analysis will try to find the Bc〇de group with relative homogeneity according to the TCM syndrome code B-Code selected by the user. The algorithm used by it It will start from each of the individual clusters - Code and then combine with the cluster until there is only one left. The system can analyze the original data (preset) or choose from a variety of different standardized conversions. The program will measure the distance or similarity. Each stage will display the similarity measure, ^ assist the user to select the most suitable cluster number κ, and use this information to understand each B-Code set. What kind of B-c〇de syndrome is the group. Chinese medicine doctors = set the B_Code certificate class to use those appropriate prescriptions or unit medicines, or the system can use the frequency or unit medicine of the Bc〇de syndrome type as Presupposition medication. When it is necessary to compare the patients, ίϊίίί i Chinese medicine syndrome code Bc〇de as one of the original analysis of Becco, let the cluster analysis judge which one of the syndromes will be the same group, you can know the patient B-c〇de syndrome class · 2· K average cluster analysis to virtual use of this program 'Let the algorithm (the method is valueable) according to the selected TCM syndrome code
Code，找出性質相似的B—c〇de群組。但是，此八 析方法會要求使用者指定集群數目κ。 知道這個資訊，便可以指定初始B—c〇de集果群使中用二 後反覆更新各B Code隼群中心。蚀田土、 …、 … 果砰T 使用者可以儲存 集群組員、距離資訊、和最後B 集 的中心，並藉此資訊瞭解各β—c〇de集群為何種、 兮證類。中醫師可以將此β-Code證類設定應 該使用那些合適方刺或單位藥，亦可由系1將此， 21 201040756 之相鱗㈣做為預設用 關聯I Λ H純之方法與階層集群分析相同 關聯程度模組名同定義及實際做法說明： 1.相異性量數 R rl異可以用來衡量患者之中醫辨證編碼 m、中醫師專家庫之方劑所適應B-c〇de進行 i tin，上果相異狀況愈低則患者與方劑所適 的關係就愈大。比對模組可使用下列相異 性重数· a.歐基里得直線距離。這是從四重表格中依 SQRT(b+c)所計算得來的值，其中b和〇 對角線上，對應到觀察值出現在某一項目但 出現在另一項目的儲存格。 b·歐基里得直線距離平方。計算得到的是不一致觀 察值的數目，它的最小值為〇,但無最大值上限。 c. 形式差異係數。範圍介於〇到1之間的二元 料相異性量數。這是從四重表格中依bc/(n^2) 所計算出來的係數，其中b和c代表對角線 上’對應到觀察值出現在某一項目、但未出現在 另一項目的儲存格，而n是觀察值的總個數。 d. 變異數。這是從四重表格中依（b+c)/4n所計算 出來的係數’其中b和c代表對角線上，對應 到銳察值出現在某一項目、但未出現在另一項目 的儲存格，而η是觀察值總個數。其範圍介於〇 到1之間。 e. Lance與Williams係數。這是從四重表格中依 (b+c)/(2a+b+c)所計算出來的係數，其中&代 表對應至觀察值、且同時出現在兩個項目中的儲 存格，而b和c代表對角線上，對應到觀察值 出現在某一項目、但未出現在另一項目的儲存 22 201040756 格。這個量數的範圍介於〇到1之間。 2.相似性量數 相似性量數可以用來衡量患者之中醫辨證編碼 B-Code與中醫師專家庫之方劑所適應b—code進行 之相似狀況’如果相似狀況愈高則患者與方劑所適 應B—Code的關係就愈大。比對模組可使用下列相異 性量數： 、 a. Russel與Rao。這是二元的内部（點）運算妹 果。其中’相符和不相符的加權值相等。 〇 b·簡單配對係數。為相符者佔總值的比率。其中， 相符和不相符的加權值相等。這是本系統的'預設 值。 c. Jaccard相似性比。這是一個係數，執行時，該 係數就已經排除聯合不出現的情況。其中，相符 和不相符的加權值相等。這個選項也叫做相似性 比。Code, find the B-c〇de group of similar nature. However, this method of analysis requires the user to specify the number of clusters κ. Knowing this information, you can specify the initial B-c〇de fruit set to use the second and then update each B Code group center. Eclipse fields, ..., ... 砰 T users can store cluster members, distance information, and the center of the last B set, and use this information to understand what kind of β-c〇de cluster, what kind of scorpion. The Chinese medicine practitioner can set the β-Code syndrome class to use those suitable square thorns or unit medicines, or the system 1 can use this, 21 201040756 phase scale (4) as the default association I Λ H pure method and hierarchical cluster analysis The same associative degree module name definition and actual practice description: 1. The dissimilarity quantity R rl can be used to measure the patient's TCM syndrome code m, the prescription of the Chinese medicine expert library is adapted to Bc〇de for i tin, the result The lower the difference, the greater the relationship between the patient and the prescription. The comparison module can use the following reciprocal multiplicity. a. Euclidean linear distance. This is the value calculated from the quadruple table according to SQRT(b+c), where b and 〇 diagonally, corresponding to the observations appearing in one item but appearing in the cell of another item. b·Euclide has a linear distance squared. The calculated number of inconsistent observations is 最小值, but there is no upper limit. c. Formal difference coefficient. The amount of binary dissimilarity ranging from 〇 to 1. This is the coefficient calculated from the quadruple table according to bc/(n^2), where b and c represent the cells on the diagonal that correspond to the observed value appearing in one item but not appearing in another item. And n is the total number of observations. d. The number of variations. This is the coefficient calculated from (b+c)/4n from the quadruple table, where b and c represent the diagonal, corresponding to the observation that the sharp value appears in one item but does not appear in another item. Grid, and η is the total number of observations. Its range is from 〇 to 1. e. Lance and Williams coefficients. This is the coefficient calculated from (b+c)/(2a+b+c) from the quadruple table, where & represents the cell corresponding to the observed value and appears in both items simultaneously, and b And c represents the diagonal, corresponding to the observations appearing in one item, but not appearing in another item of storage 22 201040756. This quantity ranges from 〇 to 1. 2. The similarity measure similarity quantity can be used to measure the similarity of the patient's TCM syndrome code B-Code and the Chinese medicine expert's prescription to the b-code. If the similarity is higher, the patient and the prescription are adapted. The relationship between B-Code is greater. The comparison module can use the following dissimilarity quantities: , a. Russel and Rao. This is the binary internal (point) operation sister. Where the 'consistent and non-conforming weights are equal. 〇 b·Simple pairing factor. The ratio of the total to the total. Where the matching and non-conforming weighting values are equal. This is the 'preset value of this system. c. Jaccard similarity ratio. This is a coefficient that, when executed, has ruled out that the union does not occur. Among them, the matching and non-conforming weights are equal. This option is also called the similarity ratio.
d. fokal與Sneath係數。此係數是以條件機率為 基礎，這個機率指的是，某個項目的特性正好符 合其他項目中的數值。如果某個項目被當成其他 項目的預測值，那麼在使用者算出這些項目個別 值的平均數之後，就可以計算出這個數值。 e. famann。此係數的計算方式是，將相符者個數減 去不相符者個數，再除以總項目個數。其範圍介 於-1到1之間。 f. Lambda 值。這個值是 Go〇dfflan 和 Kruskal 的 值。它會使用一個項目來預測另一個項目 方向都預測），來對應到誤差縮減比例 (PRE)。其範圍介於0到1之間。d. fokal and Sneath coefficients. This coefficient is based on a conditional probability that the characteristics of an item exactly match the values in other items. If an item is considered as a predictor of other items, then the user can calculate this value after calculating the average of the individual values of those items. e. famann. This coefficient is calculated by subtracting the number of matches from the number of non-conformities and dividing by the total number of items. Its range is between -1 and 1. f. Lambda value. This value is the value of Go〇dfflan and Kruskal. It uses a project to predict the direction of another project to predict, and corresponds to the error reduction ratio (PRE). Its range is between 0 and 1.
Rule’s γ係數。此索引為2χ2表格的交叉比函 數，而且它與邊際總和無關。其範圍介於-丨到1 23 201040756 之間。 利之申請要件，爰依挺生優/、，為符合發明專 准本創作，以保障發明申請，盼審委早曰賜 審委有任何稽疑，請 2苦研發’倘若鈞 力配合，實感德便。 來函指示，發明人定者^ 201040756 【圖式簡單說明】 圖1為本發明中醫智能處方集系統。 圖2為本發明中醫四診資料輸入系統。 :心)為本發明病患自覺症狀勾選輸入介面（口語 ίΓίίίί;中醫標準化症狀同義詞庫。 圖3為本發明申醫輔助辨證系統。 di本發明中醫輔助辨證系統辨證結果及病患 ❹ :3~2為本發明中醫辨證輪人系統。 以庫為產本d醫自域方集系統之資料探勘處方 以=發明資料探勘處方知識庫之Bc〇de證類 :系5統為中醫自動處方集系統之中醫師處方專家庫產 圖5-1為本發明方劑或單味 碼B—Code建議表。 樂所此治療中醫辨證編 圖6為本發明中醫辨證編 ❹ 圖7為本發明以資料探勘處對模組。 動處方集系統。 慝方知識庫為主之中醫自 圖8為本發明以中醫師處 處方集系統。 万專豕庫為主之中醫自動 圖Μ為本發明中醫自動處方集開方系統。 【主要元件符號說明】 中醫四診資料輸入系統 100 110 120 中醫辅助辨證系統 以 資料探勘處方知識庫為主之中醫自動處方集系統 25 201040756 以中醫師處方專家庫為主之中醫自動處方集系統 125 中醫辨證編碼與處方的比對模組 130 資料探勘辨證知識庫 140 資料探勘處方知識庫 145 中醫自動處方集系統之資料探勘處方知識庫產生系統14 7 中醫師辨證專家庫 150 中醫自動處方集系統之中醫師處方專家庫產生系統 157 資料探勘處方知識庫 145 中醫師處方專家庫 155 病患自覺症狀勾選輸入介面（口語化介面）2 0 0 病患自覺症狀文字輸入介面 210 舌診儀輔助介面 220 脈診儀輔助介面 230 中醫標準化症狀同義詞庫 240 病患病歷資料庫 260 病患病歷資料庫 300 相異性量數與相似性量數 310 線性迴歸、LOGISTIC迴歸、分類樹 3 2 0 資料探勘辨證知識庫 330 中醫師辨證專家庫 340 中醫辨證編碼B_Code輸出並供給中醫自 動處方集系統使用 3 5 0 病患病歷資料庫 430 資料探勘處方知識庫 440 病患病歷資料庫之處方資料 510 26 201040756 病患病歷資料庫之辨證資料 5 2 0 中醫師處方專家庫 5 3 0 資料探勘處方知識庫或中醫師處方專家庫 6 0 0 病患病歷資料庫 6 10 病患病歷資料庫 7 0 0 資料探勘處方知識庫 7 10 病患病歷資料庫 8 0 0 中醫師處方專家庫 8 10 步驟 S 10 1 步驟 S 3 0 1 步驟 S 40 1 〜S 408 步驟 S 5 1 0 〜S 5 9 0 步驟 S 6 1 0 〜S 680 步驟 S 7 2 0 〜S 770 步驟 S 8 2 0 〜S 870Rule’s gamma coefficient. This index is the cross-ratio function of the 2χ2 table, and it is independent of the marginal sum. It ranges from -丨 to 1 23 201040756. Lizhi's application requirements, 爰依挺生优/,, in order to meet the invention of the invention, in order to protect the invention application, I hope that the trial committee will give the audit committee any doubts, please 2 research and development 'if the cooperation, real sense Will. Letter to the instructor, the inventor ^ 201040756 [Simplified illustration of the drawings] Figure 1 is a Chinese medicine intelligent formulary system. 2 is a Chinese medicine four diagnostic data input system of the present invention. :心) is the input interface for the patient's self-conscious symptoms (spoken ίΓίίίί; Chinese medicine standardized symptom synonym database. Figure 3 is the invention of the medical aid-assisted syndrome system. di differentiation of the Chinese medicine-assisted syndrome differentiation system and patient ❹: 3 ~2 is the TCM syndrome-victer system of the invention. The library is the data-exploration prescription of the self-domain prescription system of the invention. The Bc〇de syndrome of the invention data exploration prescription knowledge base: the system is the automatic prescription of Chinese medicine In the system, the physician prescription expert library production diagram 5-1 is the prescription or single-code B-Code recommendation form of the present invention. The treatment of Chinese medicine syndrome differentiation diagram 6 is the invention of the Chinese medicine dialectical compilation Figure 7 is the invention for data exploration The module is set. The prescription system is based on the prescription. The knowledge base of the Chinese side is based on the prescription system of the Chinese medicine practitioner. The special prescription of the traditional Chinese medicine is the automatic prescription of the Chinese medicine practitioner. The opening system. [Main component symbol description] Chinese medicine four diagnostic data input system 100 110 120 Chinese medicine auxiliary syndrome differentiation system based on data exploration prescription knowledge base is the main Chinese medicine automatic prescription system 25 201040756 Teacher Prescription Expert Library Main TCM Automatic Formulary System 125 TCM Syndrome Coding and Prescription Comparison Module 130 Data Exploration Syndrome Knowledge Base 140 Data Exploration Prescription Knowledge Base 145 TCM Automated Formulary System Data Exploration Prescription Knowledge Base Production System 14 7 TCM Syndrome Experts Library 150 TCM Automated Formulary System Physician Prescription Experts Library Production System 157 Data Exploration Prescription Knowledge Base 145 TCM Physician Prescription Expert Library 155 Patient conscious symptoms check input interface (spoken interface) 2 0 0 disease Susceptible symptoms text input interface 210 Tongue diagnosis device auxiliary interface 220 Pulse diagnosis device auxiliary interface 230 TCM standardized symptom synonym library 240 Patient medical record database 260 Patient medical record database 300 Dissimilarity quantity and similarity quantity 310 Linear regression, LOGISTIC regression, classification tree 3 2 0 data exploration and syndrome knowledge base 330 TCM syndrome diagnosis expert library 340 TCM syndrome code B_Code output and supply TCM automatic formulary system use 3 5 0 patient medical record database 430 data exploration prescription knowledge base 440 patients Medical records database location information 510 26 201040756 disease Dialectical data of the database 5 2 0 Physician prescription expert library 5 3 0 data exploration prescription knowledge base or Chinese medicine prescription expert database 6 0 0 patient medical record database 6 10 patient medical record database 7 0 0 data exploration prescription knowledge base 7 10 Patient Medical Record Database 8 0 0 Physician Prescription Experts Library 8 10 Step S 10 1 Step S 3 0 1 Step S 40 1 to S 408 Step S 5 1 0 to S 5 9 0 Step S 6 1 0 〜S 680 Step S 7 2 0 to S 770 Step S 8 2 0 to S 870