1、1 研 究 生:黃 金 榮九十四學年度第二學期台灣科技大學營建系博士班研究生定期報告九十四學年度第二學期台灣科技大學營建系博士班研究生定期報告Value-Added Treatment Inference Model for Rule-based Uncertainty Knowledge2AgendAAgendA1.論文題目與摘要論文題目與摘要2.論文研究流程論文研究流程3.規則式知識相似度計算及知識關聯規則式知識相似度計算及知識關聯 4.規則式確定性知識加值處理推論模式規則式確定性知識加值處理推論模式5.規則式不確定性知識加值處理推論模式規則式不確定性知識加值處理推論模式6.研究成果研究成
2、果7.結論結論31.論文題目與摘要論文題目與摘要Topic of reseArch規則式不確定性知識加值處理推論模式規則式不確定性知識加值處理推論模式(Value-Added Treatment Inference Model for Rule-based Uncertainty Knowledge)keywords條件機率條件機率(Conditional Probability),確定係數確定係數(Certainty Factor),知識加值知識加值(Knowledge Value-Added),可信度指數可信度指數(Reliable Factor),決策指數決策指數(Decision In
3、dex)41.論文題目與摘要論文題目與摘要ISSUE:1規則式知識相似度計算及知識關聯規則式知識相似度計算及知識關聯 ISSUE:2規則式確定性知識加值處理推論模式規則式確定性知識加值處理推論模式 ISSUE:3規則式不確定性知識加值處理推論模式規則式不確定性知識加值處理推論模式5知識庫中知識來源不同,專家意見不一,導致知識庫中知識存在相同意義、衝突或數據大小不一致。隨時空環境改變,新技術、新法規、新方法、新證據產生等因素,可能產生知識不適用。知識整體內涵及關聯無法顯現,且錯誤的知識,會導致錯誤的決策。BAckground1.論文題目與摘要論文題目與摘要6依據知識關聯,建構規則式不確定知識加值
4、處理推論模式,進行合併、整合、刪除、創新、及新增等加值處理作業。reseArch oBjecTive提升知識應用之附加價值。使知識表現更具整體性,知識間關聯映射及知識輔助決策指標能具體顯現,有效避免做出錯誤決策。1.論文題目與摘要論文題目與摘要71.Objective Formation2.文獻回顧文獻回顧 3.規則式知識相似度計算規則式知識相似度計算Next pagea.相似度衡量方法 b.知識再利用及知識加值c.不確定係數計算2.論文研究流程論文研究流程(1/2)4.知識相似度計算系統知識相似度計算系統 快速準確求得規則式知識相似度矩陣,確定知識間之關聯 85.知識關聯與知識關聯與確定確定
5、知識加值處理推論知識加值處理推論模式模式6.規則式不確定知識規則式不確定知識加值處理推論模式加值處理推論模式 End7.結論結論依據知識6種不同關聯,經合併、整合、刪除、創新及新增等加值推論處理(可信度指數理論)不確定知識6種不同關聯,經合併、整合、刪除、創新及新增等加值推論處理(決策指數理論)Up page2.論文研究流程論文研究流程(2/2)9摘要摘要3.ISSUE:1 規則式知識相似度計算及知識關聯規則式知識相似度計算及知識關聯提出規則式知識使用O-A-RV結構表示。整合條件機率、向量矩陣、人工智慧,建構條件 機率知識相似度演算法及知識相似度計算系統。快速準確求得規則式知識相似度矩陣,確
6、定知識間之關聯。作為提升知識附加價值處理之知識來源。10規則式知識使用O-A-RV結構表示ISSUE:1 規則式知識相似度計算及知識關聯規則式知識相似度計算及知識關聯規則式知識表示語法為IF前提(antecedent)THEN 結論(consequent)。無論前提或結論,均可使用一個句子(sentence)表示。使用物件(Object)、屬性(Attribute)、關係運算子(Relationship Operator)、語意值(Linguistic Value)等四部份組成O-A-RV結構表示。11句子句子O-A-RV結構表示結構表示ISSUE:1 規則式知識相似度計算及知識關聯規則式知識
7、相似度計算及知識關聯SentenceOARVThe temperature of the engine is more than 100 c.enginetemperature100 cThe vehicles color is red.vehiclecolor=redThe span of bridge is less than 50m.bridgespan50m前提向量前提向量=O1 A1 R1V1 O2 A2 R2V2 O3 A3 R3V3On An RnVn 第1個物件屬性 第2個物件屬性 第3個物件屬性 第n個物件屬性12O-A-RV分量轉換對應表示ISSUE:1 規則式知識相似度計
8、算及知識關聯規則式知識相似度計算及知識關聯資料型態運算性質說明Nominal無法比較大小,無法算術運算車子顏色,比較結果為真或假Ordinal有限個,但具有次序關係,可以比較大小,無法算術運算衣服尺寸型號:SMLXLInterval and ratio數值型資料,可以比較大小,也可以算數運算影響範圍:很小小中大x 與知識庫案例與知識庫案例Ky 等各種狀況之等各種狀況之RV分量之轉換對應分量之轉換對應其中其中:V1 為為x與與y的關係,的關係,V2 為為K與與y的關係,的關係,V3 為為T與與x的關係。的關係。max表示分量表示分量V之最大值再加之最大值再加R。min表示分量表示分量V之最小值再
9、減之最小值再減R。18RV分量(Interval and ratio)轉換對應表示ISSUE:1 規則式知識相似度計算及知識關聯規則式知識相似度計算及知識關聯Transforming Mapping for x y,Tx and Ky Transforming Mapping for x y,T=x and Ky Transforming Mapping for x y,Ty 19RV分量(Interval and ratio)轉換對應表示ISSUE:1 規則式知識相似度計算及知識關聯規則式知識相似度計算及知識關聯20RV分量(Interval and ratio)轉換對應表示ISSUE:1
10、規則式知識相似度計算及知識關聯規則式知識相似度計算及知識關聯當當測試案例測試案例 x1 T x2與知識庫案例與知識庫案例 y1 K”Rule 1:IF a1 THEN V42Rule 2:IF a1 THEN V30Rule 3:IF a1 THEN V52Rule 4:IF a1 THEN V60Then constant value added inferences of this knowledge set RF are as follows:Rule 1:IF a1 THEN V30(RF=0)Rule 2:IF a1 THEN 30V 42(RF=0.25)Rule 3:IF a1
11、THEN 42V 52(RF=0.5)Rule 4:IF a1 THEN 5260(RF=1)44可信度指數表示實例:ISSUE:2 規則式確定性知識加值處理推論模式規則式確定性知識加值處理推論模式Example 2:if in knowledge base there are five rules with same antecedent a1Rule 1:IF a1 THEN V30Rule 3:IF a1 THEN V52Rule 4:IF a1 THEN V60Rule 5:IF a1 THEN V=50Then the value added of this knowledge se
12、t is inferenced below:Rule 1:IF a1 THEN V 30(RF=0.4)Rule 2:IF a1 THEN 30V42(RF=0.6)Rule 3:IF a1 THEN 42V 52(RF=0.4,V 50)Rule 4:IF a1 THEN V=50(RF=0.6)Rule 5:IF a1 THEN 52V60(RF=0.6)Rule 6:IF a1 THEN V 60(RF=0.4)45不同關聯知識群之加值處理推論模式ISSUE:2 規則式確定性知識加值處理推論模式規則式確定性知識加值處理推論模式NOT c1a1Rule 3c1NOT a1Rule 2c1a
13、1Rule 1ConsequentAntecedent Knowledge Previous knowledgeInference RF=RF4c1NOT a1Rule 4RF=RF3c1a1Rule 3RF=RF2NOT c1a1Rule 2Denote by RFInnovateRF=RF1c1a1Rule 1Remark Value-added treatmentReliable factor ConsequentAntecedent KnowledgeKnowledge value-added treatmentNOT c1a1Rule 3c1NOT a1Rule 2c1a1Rule
14、1ConsequentAntecedent Knowledge Previous knowledgeInference RF=RF4c1NOT a1Rule 4RF=RF3c1a1Rule 3RF=RF2NOT c1a1Rule 2Denote by RFInnovateRF=RF1c1a1Rule 1Remark Value-added treatmentReliable factor ConsequentAntecedent KnowledgeKnowledge value-added treatmentValue-Added treatment inference of knowledg
15、e sets with same antecedent 46不同關聯知識群之加值處理推論模式ISSUE:2 規則式確定性知識加值處理推論模式規則式確定性知識加值處理推論模式Value-Added treatment inference of knowledge sets with same consequentc1a3Rule 3c1a2Rule 2c1a1Rule 1Consequent AntecedentKnowledgePrevious knowledgeInference RF=RF3c3a1Rule 5RF=RF2c2a1Rule 4there are contradictions
16、among consequents c1,c2,c3,or variable data size,denoted by RFInnovateRF=RF1c1a1Rule 3MergeRF=1NOT c1 OR NOT c2 OR NOT c3NOT a1 OR NOT a2 OR NOT a3 Rule 2No contradiction among consequents c1,c2,c3,or no variable data sizeMergeRF=1c1a1 OR a2 OR a3Rule 1Remark Value-added treatmentReliable factor Con
17、sequentAntecedent KnowledgeKnowledge value-added treatmentc1a3Rule 3c1a2Rule 2c1a1Rule 1Consequent AntecedentKnowledgePrevious knowledgeInference RF=RF3c3a1Rule 5RF=RF2c2a1Rule 4there are contradictionsamong consequents c1,c2,c3,or variable data size,denoted by RFInnovateRF=RF1c1a1Rule 3MergeRF=1NOT
18、 c1 OR NOT c2 OR NOT c3NOT a1 OR NOT a2 OR NOT a3 Rule 2No contradiction among consequents c1,c2,c3,or no variable data sizeMergeRF=1c1a1 OR a2 OR a3Rule 1Remark Value-added treatmentReliable factor ConsequentAntecedent KnowledgeKnowledge value-added treatment47規則式知識加值處理推論演算法架構(Rule-based Knowledge
19、Value-Added Treatment Inference Algorithm,RKVATIA)ISSUE:2 規則式確定性知識加值處理推論模式規則式確定性知識加值處理推論模式Conditional Probability Knowledge Similarity AlgorithmConditional Probability Knowledge Similarity AlgorithmKnowledge Value-AddedTreatmentKnowledge Value-AddedTreatmentRule-basedKnowledgeRule-basedKnowledgeReli
20、able FactorTheoryReliable FactorTheoryKnowledge RelationshipKnowledge RelationshipKnowledge Value-AddedKnowledge Value-AddedConditional Probability Knowledge Similarity AlgorithmConditional Probability Knowledge Similarity AlgorithmKnowledge Value-AddedTreatmentKnowledge Value-AddedTreatmentRule-bas
21、edKnowledgeRule-basedKnowledgeReliable FactorTheoryReliable FactorTheoryKnowledge RelationshipKnowledge RelationshipKnowledge Value-AddedKnowledge Value-Added48規則式知識加值處理推論演算法(Rule-based Knowledge Value-Added Treatment Inference Algorithm,RKVATIA)ISSUE:2 規則式確定性知識加值處理推論模式規則式確定性知識加值處理推論模式Input:一般規則式知識:
22、一般規則式知識Output:規則式加值處理知識:規則式加值處理知識Step 1:一般規則式知識表示:一般規則式知識表示O-A-RV結構結構Step 2:O-A-RV各分量轉換對應成數值並表成前提矩陣、結論矩陣及各分量轉換對應成數值並表成前提矩陣、結論矩陣及知識矩陣知識矩陣Step 3:計算兩兩前提相似度、兩兩結論相似度、兩兩知識相似度、前:計算兩兩前提相似度、兩兩結論相似度、兩兩知識相似度、前提結論相似度、結論前提相似度提結論相似度、結論前提相似度Step 4:建立前提相似度矩陣、結論相似度矩陣、知識相似度矩陣、前:建立前提相似度矩陣、結論相似度矩陣、知識相似度矩陣、前提結論相似度矩陣、結論前
23、提相似度提結論相似度矩陣、結論前提相似度Step 5:找出具有各種特殊關聯可以加值處理知識群:找出具有各種特殊關聯可以加值處理知識群Step 6:依知識關聯及可信度指數理論進行適當的加值推論處理:依知識關聯及可信度指數理論進行適當的加值推論處理 Step 7:儲存規則式加值處理知識:儲存規則式加值處理知識Step 8:停止:停止 49知識實例加值處理知識實例加值處理ISSUE:2 規則式確定性知識加值處理推論模式規則式確定性知識加值處理推論模式50摘要摘要5.ISSUE:3 規則式不確定知識加值處理推論模式規則式不確定知識加值處理推論模式提出可信度係數(RF)表示存在衝突或重疊或對應數據大小不
24、一致之知識可以採信程度確定指數(cf)顯示知識是否存在程度決策指數(RI)顯示知識輔助決策指標。依據知識關聯,建構規則式不確定知識加值處理推論模式進行合併、整合、刪除、創新、新增等加值處理作業,使知識經加值處理後,知識表現更具整體性,知識間關聯映射及知識輔助決策指標能具體顯現,有效避免做出錯誤決策。51單一前提知識確定係數單一前提知識確定係數(certainty factor,cf)計算計算ISSUE:3 規則式不確定知識加值處理推論模式規則式不確定知識加值處理推論模式Rule:IF THEN cfa,cfrule簡化表示 a,c cfa,cfrule=a,c cf知識之確定指數計算 cf=c
25、fa*cfrule其中cfa:單一前提發生之確定指數,表示該前提存在的機率,其值介於0與1之間。cfrule:規則推論之確定指數,表示該規則推論發生的機率,其值介於0與1之間。cf:組合單一前提及規則推論之知識淨確定係數,表示該知識存在的機率。定義其值介於0與1之間,52前提具有邏輯運算子前提具有邏輯運算子ANDAND,OROR知識確定係數計算知識確定係數計算ISSUE:3 規則式不確定知識加值處理推論模式規則式不確定知識加值處理推論模式53具有收斂對應具有收斂對應(多對一對應多對一對應)關聯知識確定係數計算關聯知識確定係數計算ISSUE:3 規則式不確定知識加值處理推論模式規則式不確定知識加
26、值處理推論模式54不確定知識各種關聯之淨確定係數計算推論不確定知識各種關聯之淨確定係數計算推論ISSUE:3 規則式不確定知識加值處理推論模式規則式不確定知識加值處理推論模式55不確定知識各種關聯之淨確定係數計算推論不確定知識各種關聯之淨確定係數計算推論ISSUE:3 規則式不確定知識加值處理推論模式規則式不確定知識加值處理推論模式56不確定知識各種關聯之淨確定係數計算推論不確定知識各種關聯之淨確定係數計算推論ISSUE:3 規則式不確定知識加值處理推論模式規則式不確定知識加值處理推論模式57決策指數決策指數(Decision Index,DI)理論理論ISSUE:3 規則式不確定知識加值處理
27、推論模式規則式不確定知識加值處理推論模式綜合表示知識是否存在之確定程度與可採信程度之高低綜合表示知識是否存在之確定程度與可採信程度之高低,有利於輔助決策。,有利於輔助決策。決策指數決策指數=確定指數確定指數*可信度指數可信度指數 DI=cf*RF DI範圍介於範圍介於0 和和 1之間。之間。DF值愈大,值愈大,表示該知識確定存在且可採信之程度愈高。表示該知識確定存在且可採信之程度愈高。58不確定知識各種關聯之加值處理推論模式不確定知識各種關聯之加值處理推論模式ISSUE:3 規則式不確定知識加值處理推論模式規則式不確定知識加值處理推論模式類型類型1 1:具有相同前提知識群:具有相同前提知識群(
28、一對多關係一對多關係)之加值處理推論之加值處理推論 RF3cf3=cfa1*cfrule3cfrule3cfa1c3a1RF2cf2=cfa1*cfrule2cfrule2cfa1c2a1RF1cf1=cfa1*cfrule1cfrule1cfa1c1a1RFcfcfrulecfaconsequentantecedent Previous knowledgeReasoningcf3*RF3RF=RF3cf3c3cf2*RF2RF=RF2cf2c2cf1*RF1RF=RF1cf1c1a1maxcf1,cf2,cf3*maxRF1,RF2,RF3maxRF1,RF2,RF3maxcf1,cf2,c
29、f3c1OR c2OR c3a1mincf1,cf2,cf3*minRF1,RF2,RF3minRF1,RF2,RF3mincf1,cf2,cf3c1AND c2AND c3a1mincf1,cf3*minRF1,RF3minRF1,RF3mincf1,cf3c1AND c3a1mincf1,cf2*minRF1,RF2minRF1,RF2mincf1,cf2c1AND c2a1DI RFcfconsequentantecedentKnowledge value-added treatmentRF3cf3=cfa1*cfrule3cfrule3cfa1c3a1RF2cf2=cfa1*cfrul
30、e2cfrule2cfa1c2a1RF1cf1=cfa1*cfrule1cfrule1cfa1c1a1RFcfcfrulecfaconsequentantecedent Previous knowledgeReasoningcf3*RF3RF=RF3cf3c3cf2*RF2RF=RF2cf2c2cf1*RF1RF=RF1cf1c1a1maxcf1,cf2,cf3*maxRF1,RF2,RF3maxRF1,RF2,RF3maxcf1,cf2,cf3c1OR c2OR c3a1mincf1,cf2,cf3*minRF1,RF2,RF3minRF1,RF2,RF3mincf1,cf2,cf3c1AN
31、D c2AND c3a1mincf1,cf3*minRF1,RF3minRF1,RF3mincf1,cf3c1AND c3a1mincf1,cf2*minRF1,RF2minRF1,RF2mincf1,cf2c1AND c2a1DI RFcfconsequentantecedentKnowledge value-added treatment59不確定知識各種關聯之加值處理推論模式不確定知識各種關聯之加值處理推論模式ISSUE:3 規則式不確定知識加值處理推論模式規則式不確定知識加值處理推論模式類型類型1 1:具有相同前提知識群:具有相同前提知識群(一對多關係一對多關係)之加值處理推論範例之加
32、值處理推論範例 0.670.720.90.8c3a10.670.480.60.8c2a10.330.640.80.8c1a1RFcfcfrulecfaconsequentantecedent Previous knowledge Reasoning0.720.480.670.72c30.480.320.670.48c20.640.210.330.64c1a10.720.480.670.72c1OR c2OR c3a10.480.160.330.48c1AND c2AND c3a10.640.210.330.64c1AND c3a10.480.160.330.48c1AND c2a1DI(RF=
33、1)DI(RF1)RFcfconsequentantecedentKnowledge value-added treatment0.670.720.90.8c3a10.670.480.60.8c2a10.330.640.80.8c1a1RFcfcfrulecfaconsequentantecedent Previous knowledge Reasoning0.720.480.670.72c30.480.320.670.48c20.640.210.330.64c1a10.720.480.670.72c1OR c2OR c3a10.480.160.330.48c1AND c2AND c3a10.
34、640.210.330.64c1AND c3a10.480.160.330.48c1AND c2a1DI(RF=1)DI(RF1)RFcfconsequentantecedentKnowledge value-added treatment60不確定知識各種關聯之加值處理推論模式不確定知識各種關聯之加值處理推論模式ISSUE:3 規則式不確定知識加值處理推論模式規則式不確定知識加值處理推論模式類型類型2 2:具有相同結論的知識群:具有相同結論的知識群(多對一關係多對一關係)之加值處理推論之加值處理推論RF3cf3=cfa3*cfrule3cfrule3cfa3c1a3RF2cf2=cfa2*c
35、frule2cfrule2cfa2c1a2RF1cf1=cfa1*cfrule1cfrule1cfa1c1a1RFcfcfrulecfaconsequentantecedent Previous knowledgeReasoningcf3*RF3RF3cf3c1a3cf2*RF2RF2cf2c1a2cf1*RF1RF1cf1c1a1cf123*maxRF1,RF2,RF3maxRF1,RF2,RF3cf123c1a1OR a2OR a3cf23*maxRF2,RF3maxRF2,RF3cf23c1a2OR a3cf12*maxRF1,RF2maxRF1,RF2cf12c1a1OR a2DI R
36、FcfconsequentantecedentKnowledge value-added treatmentRF3cf3=cfa3*cfrule3cfrule3cfa3c1a3RF2cf2=cfa2*cfrule2cfrule2cfa2c1a2RF1cf1=cfa1*cfrule1cfrule1cfa1c1a1RFcfcfrulecfaconsequentantecedent Previous knowledgeReasoningcf3*RF3RF3cf3c1a3cf2*RF2RF2cf2c1a2cf1*RF1RF1cf1c1a1cf123*maxRF1,RF2,RF3maxRF1,RF2,R
37、F3cf123c1a1OR a2OR a3cf23*maxRF2,RF3maxRF2,RF3cf23c1a2OR a3cf12*maxRF1,RF2maxRF1,RF2cf12c1a1OR a2DI RFcfconsequentantecedentKnowledge value-added treatment61不確定知識各種關聯之加值處理推論模式不確定知識各種關聯之加值處理推論模式ISSUE:3 規則式不確定知識加值處理推論模式規則式不確定知識加值處理推論模式類型類型2 2:具有相同結論的知識群:具有相同結論的知識群(多對一關係多對一關係)之加值處理推論範例之加值處理推論範例0.670.72
38、0.90.8c1a30.670.360.60.6c1a20.330.560.80.7c1a1RFcfcfrulecfaconsequentantecedent Previous knowledge Reasoning0.480.670.72c1a30.240.67036c1a20.180.330.56c1a10.620.670.92c1a1OR a2OR a30.550.670.82c1a2OR a30.480.670.72c1a1OR a2DI RFcfconsequentantecedentKnowledge value-added treatment0.670.720.90.8c1a30
39、.670.360.60.6c1a20.330.560.80.7c1a1RFcfcfrulecfaconsequentantecedent Previous knowledge Reasoning0.480.670.72c1a30.240.67036c1a20.180.330.56c1a10.620.670.92c1a1OR a2OR a30.550.670.82c1a2OR a30.480.670.72c1a1OR a2DI RFcfconsequentantecedentKnowledge value-added treatment62不確定知識各種關聯之加值處理推論模式不確定知識各種關聯之
40、加值處理推論模式ISSUE:3 規則式不確定知識加值處理推論模式規則式不確定知識加值處理推論模式類型類型3:具有相同前提與結論知識群具有相同前提與結論知識群(多對多關係多對多關係)之加值處理推論之加值處理推論RF5cf5=cfa2*cfrule5cfrule5cfa2c3a2RF4cf4=cfa2*cfrule4cfrule4cfa2c1a2RF3cf3=cfa1*cfrule3cfrule3cfa1c3a1RF2cf2=cfa1*cfrule2cfrule2cfa1c2a1RF1cf1=cfa1*cfrule1cfrule1cfa1c1a1RFcfcfrulecfaconsequentant
41、ecedent Previous knowledgeReasoningcf3*RF3RF3cf3c3a1mincf4,cf5*minRF1,RF3minRF1,RF3mincf4,cf5c1AND c3a2mincf1,cf2,cf3*minRF1,RF2,RF3minRF1,RF2,RF3mincf1,cf2,cf3c1AND c2AND c3a1cf35*maxRF3,RF5maxRF3,RF5cf35c3a1OR a2cf14*maxRF1,RF4maxRF1,RF4cf14c1a1OR a2DI RFcfconsequentantecedentKnowledge value-added
42、 treatment63不確定知識各種關聯之加值處理推論模式不確定知識各種關聯之加值處理推論模式ISSUE:3 規則式不確定知識加值處理推論模式規則式不確定知識加值處理推論模式類型類型4:具有某個知識結論為另一知識前提之知識群具有某個知識結論為另一知識前提之知識群(因果關係因果關係)之之加值處理推論加值處理推論 Reasoning1cf3=cfc2*cfrule3cfrule3cfc2c3c21cf2=cfc1*cfrule2cfrule2cfc1c2c11cf1=cfa1*cfrule1cfrule1cfa1c1a1RFcfcfrulecfaconsequentantecedent Prev
43、ious knowledgecf31cf3c3c2cf2*cf31cf2*cf3c2AND c3c1cf1*cf21cf1*cf2c1AND c2a1cf1*cf2*cf31cf1*cf2*cf3c1AND c2AND c3a1DIRFcfconsequentantecedentKnowledge value-added treatmentReasoning1cf3=cfc2*cfrule3cfrule3cfc2c3c21cf2=cfc1*cfrule2cfrule2cfc1c2c11cf1=cfa1*cfrule1cfrule1cfa1c1a1RFcfcfrulecfaconsequenta
44、ntecedent Previous knowledgecf31cf3c3c2cf2*cf31cf2*cf3c2AND c3c1cf1*cf21cf1*cf2c1AND c2a1cf1*cf2*cf31cf1*cf2*cf3c1AND c2AND c3a1DIRFcfconsequentantecedentKnowledge value-added treatment64不確定知識各種關聯之加值處理推論模式不確定知識各種關聯之加值處理推論模式ISSUE:3 規則式不確定知識加值處理推論模式規則式不確定知識加值處理推論模式類型類型5:具有知識前提或結論出現相反結論之知識群具有知識前提或結論出現相
45、反結論之知識群(矛盾關係矛盾關係)之之加值處理推論加值處理推論 RF3cf3=cfa1*cfrule3cfrule3cfa1NOT c1a1RF2cf2=cfNOT a1*cfrule2cfrule2cfNOT a1c1NOT a1RF1cf1=cfa1*cfrule1cfrule1cfa1c1a1RFcfcfrulecfaconsequentantecedent Previous knowledgecf3*RF3RF3cf3NOT c1a1cf2*RF2RF2cf2c1NOT a1cf1*RF1RF1cf1c1a1maxcf1,cf3*maxRF1,RF2maxRF1,RF2maxcf1,c
46、f3c1OR(NOT c1)a1mincf1,cf3*minRF1,RF3minRF1,RF3mincf1,cf3c1AND(NOT c1)a1cf12*maxRF1,RF2maxRF1,RF2cf12c1a1OR(NOT a1)DIRFcfconsequentantecedentKnowledge value-added treatmentReasoningRF3cf3=cfa1*cfrule3cfrule3cfa1NOT c1a1RF2cf2=cfNOT a1*cfrule2cfrule2cfNOT a1c1NOT a1RF1cf1=cfa1*cfrule1cfrule1cfa1c1a1R
47、Fcfcfrulecfaconsequentantecedent Previous knowledgecf3*RF3RF3cf3NOT c1a1cf2*RF2RF2cf2c1NOT a1cf1*RF1RF1cf1c1a1maxcf1,cf3*maxRF1,RF2maxRF1,RF2maxcf1,cf3c1OR(NOT c1)a1mincf1,cf3*minRF1,RF3minRF1,RF3mincf1,cf3c1AND(NOT c1)a1cf12*maxRF1,RF2maxRF1,RF2cf12c1a1OR(NOT a1)DIRFcfconsequentantecedentKnowledge
48、value-added treatmentReasoning65不確定知識各種關聯之加值處理推論模式不確定知識各種關聯之加值處理推論模式ISSUE:3 規則式不確定知識加值處理推論模式規則式不確定知識加值處理推論模式類型類型6:具有各自獨立知識群具有各自獨立知識群(平行關係平行關係)之加值處理推論之加值處理推論 delete1cf2c2a2When new evidence,or a new regulation,process,procedure or knowledge item is found and establishedinnovate1cf3c3a1RemarkDFRFcfcon
49、sequentantecedentKnowledge value-added treatment1cf2=cfc1*cfrule2cfrule2cfa2c2a21cf1=cfa1*cfrule1cfrule1cfa1c1a1RFcfcfrulecfaconsequentantecedent Previous knowledgeReasoningdelete1cf2c2a2When new evidence,or a new regulation,process,procedure or knowledge item is found and establishedinnovate1cf3c3a
50、1RemarkDFRFcfconsequentantecedentKnowledge value-added treatment1cf2=cfc1*cfrule2cfrule2cfa2c2a21cf1=cfa1*cfrule1cfrule1cfa1c1a1RFcfcfrulecfaconsequentantecedent Previous knowledgeReasoning66規則式不確定知識加值處理推論演算法架構規則式不確定知識加值處理推論演算法架構(Rule-based Uncertainty Knowledge Value-Adding Treatment Inference Algo