JPH0695880A - Example base inference device - Google Patents

Example base inference device

Info

Publication number
JPH0695880A
JPH0695880A JP4269129A JP26912992A JPH0695880A JP H0695880 A JPH0695880 A JP H0695880A JP 4269129 A JP4269129 A JP 4269129A JP 26912992 A JP26912992 A JP 26912992A JP H0695880 A JPH0695880 A JP H0695880A
Authority
JP
Japan
Prior art keywords
case
inference
new
causal relationship
condition
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
JP4269129A
Other languages
Japanese (ja)
Other versions
JP2632117B2 (en
Inventor
Hiroaki Tsutsui
宏明 筒井
Atsushi Kurosaki
淳 黒崎
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Azbil Corp
Original Assignee
Azbil Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Azbil Corp filed Critical Azbil Corp
Priority to JP4269129A priority Critical patent/JP2632117B2/en
Priority to US08/109,179 priority patent/US5918200A/en
Priority to DE69328956T priority patent/DE69328956T2/en
Priority to EP93113564A priority patent/EP0590305B1/en
Priority to CN93118822A priority patent/CN1047011C/en
Publication of JPH0695880A publication Critical patent/JPH0695880A/en
Application granted granted Critical
Publication of JP2632117B2 publication Critical patent/JP2632117B2/en
Anticipated expiration legal-status Critical
Expired - Lifetime legal-status Critical Current

Links

Landscapes

  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

PURPOSE:To process the example of continuous data with a small scale control ler or the like by inferring a new example while considering importance calcu lated from distance between a similar example and the new example and calcu lating reliability from similarity. CONSTITUTION:A causal relationship model generation means 21 generates the causal relationship model of the input factor and the result of example data in the past and, when the new example data are inputted, a causal relationship model learning means 22 learns the new example data and updates the causal relationship model. The causal relationship models are stored in a storage part 23. In the meantime, an inference means 25 infers the new example data inputted from a causal relationship model condition input part 24 based on the causal relationship model of the storage part 13, sends out the result to a final predicted value deciding means 16, integrates it with a recognized result by an automatic state recognition generation device 10 and displays the predicted value at a load predicted value display part 30. Also, the inference means 25 sends out the inferred result to a reliability calculation means 26, calculates the reliability and displays it at a reliability display part 50.

Description

【発明の詳細な説明】Detailed Description of the Invention

【0001】[0001]

【産業上の利用分野】本発明は、過去の経験的事例を事
例データベースに蓄積しておき、新事例の条件に類似し
た既存事例を選択,修正することにより、新事例の結論
を推論する事例ベース推論装置に関する。
BACKGROUND OF THE INVENTION The present invention is a case in which past empirical cases are stored in a case database and existing cases similar to the conditions of the new case are selected and corrected to infer the conclusion of the new case. Regarding the base reasoning device.

【0002】[0002]

【従来の技術】事例ベース推論技術は、“知的活動は、
過去の経験を中心にして行われる”という仮説に基づく
推論モデルの生成技術であり、基本的には次のような推
論が行われる。即ち、まず過去に経験した事例を蓄積し
て事例ベースを生成し、次いで新事例の条件が入力され
た場合にはこれと最も類似した既存事例を事例ベースか
ら検索すると共に、検索された既存事例を適当に修正し
て新事例の結論を推論する。そしてさらに、新事例を学
習して事例ベースを更新するものとなっている。このよ
うな、推論技術は、事例として例えば文字や単語等の個
別データを取り扱ういわゆる自然言語処理等の分野にそ
の適用がなされており、この分野における推論装置実現
のため、各種の手法が提案されている。
2. Description of the Related Art Case-based reasoning technology
It is a technique of generating an inference model based on the hypothesis that "it is performed based on past experience." Basically, the following inference is performed. When the condition of the new case is generated and then the existing case that is most similar to the condition of the new case is input, the case base is searched, and the searched existing case is appropriately modified to infer the conclusion of the new case. Furthermore, the new case is learned and the case base is updated.The reasoning technology is applicable to the field of so-called natural language processing, which handles individual data such as characters and words as a case. Various methods have been proposed for realizing an inference device in this field.

【0003】[0003]

【発明が解決しようとする課題】しかし、このような事
例ベース推論技術を、地域冷暖房システムの負荷予測,
プラントのバルブ操作量の制御及び機器特性の同定とい
ったような技術に適用しようとした場合、上記システム
や機器等では事例が文字や単語等の個別データでは無
く、例えば時系列で観測され連続的に変化する気温等の
連続データとなるため、次のような問題を生じている。
即ち、第1点として連続データの事例をどのような形式
で表現して事例ベースを作成するかという問題点、第2
点として連続データの事例間の類似度をどのように決定
して新事例に対する類似事例を検索するかという問題
点、第3点として連続データの類似事例をどのように修
正して新事例の結論として推論するかという問題点、第
4点として連続データの新事例をどのように学習して事
例ベースを更新するかという問題点がある。また、従来
の事例ベースは、32k個のプロセッサを有するよう
な、膨大な記憶容量と計算量を前提とした技術であるた
め、例えば上記システムや機器等で要求される「小資源
環境におけるリアルタイムな処理の実行」を実現するこ
とが困難であるという問題もあった。したがって本発明
は、気温等の連続データの事例を小規模コントローラ等
で処理可能にすることを目的とする。
However, such a case-based reasoning technique is used to predict the load of a district heating and cooling system.
When trying to apply to technology such as control of valve operation amount of plant and identification of equipment characteristics, in the above systems and equipment, the cases are not individual data such as letters and words, but are continuously observed, for example, in time series. Since it is continuous data such as changing temperature, the following problems occur.
That is, as a first point, there is a problem of how to form a case base by expressing a case of continuous data.
As a point, the problem of how to determine the similarity between cases of continuous data and search for a similar case to a new case, and the third point is how to modify the similar case of continuous data and conclude the new case. As a fourth point, there is a problem of how to learn a new case of continuous data and update the case base. In addition, the conventional case base is a technology that assumes a huge storage capacity and calculation amount such as having 32k processors, so that, for example, "real time in a small resource environment" required by the above-mentioned system or device. There is also a problem that it is difficult to realize “execution of processing”. Therefore, it is an object of the present invention to allow a small-scale controller or the like to process an example of continuous data such as temperature.

【0004】[0004]

【課題を解決するための手段】このような課題を解決す
るために本発明は、各事例間に連続性が成立するような
事例の推論を行う推論装置であって、既存事例の条件部
の各入力変数の値を離散化しこの離散値により条件部を
象徴化すると共に同一象徴に対する既存事例の結論部を
統合化しこの結論部と条件部との関係と,条件部の変化
に対する結論部の変化情報とを付随して生成する事例ベ
ース生成部と、推論を行うための前処理として既存事例
から推論精度を考慮した条件部の類似度を判定するため
に用いるしきい値を決定する手段と,推論に用いる類似
事例の個数の制約を決定する手段とからなるパラメータ
決定部と、新事例の条件部に対する既存事例の類似度を
条件部の位相の連続性に基づいて決定する手段と,条件
部の類似度に基づき事例ベースから新事例に対する類似
事例を条件部が結論部に与える影響度を考慮して検索す
る手段と,新事例に対する類似事例の重要度を位相によ
る条件部の距離から決定する手段と,重要度に基づき複
数の類似事例を条件部の変化に対する結論部の変化を考
慮して修正を行い統合化し新事例の結論部を推論する手
段と,条件部及び結論部の連続性により類似事例の条件
部の位相が近ければ結論部の位相も近いことを用い推論
結果の信憑性を判定する手段と,推論結果及びその信憑
性の情報を出力する手段とからなる事例ベース推論部
と、新事例をそのままリアルタイムに学習し事例ベース
を更新する事例ベース学習部とを設けたものである。
In order to solve such a problem, the present invention is an inference apparatus for inferring cases in which continuity is established between each case, which is a condition part of an existing case. The value of each input variable is discretized, the condition part is symbolized by this discrete value, and the conclusion part of the existing case for the same symbol is integrated, and the relation between this conclusion part and the condition part, and the change of the conclusion part for the change of the condition part A case-based generation unit that generates information and information, a unit that determines a threshold value used to determine the similarity of a conditional unit considering the inference accuracy from an existing case as preprocessing for performing inference, A parameter deciding unit comprising means for deciding the number of similar cases used for inference, means for deciding the similarity of the existing case to the condition section of the new case based on the phase continuity of the condition section, and the condition section Based on the similarity of A method for searching a similar case for a new case from a case base in consideration of the degree of influence of a conditional part on a conclusion part, a means for determining the importance of a similar case for a new case from the distance of the condition part by phase, and an importance level A method for inferring the conclusion part of a new case by modifying and consolidating multiple similar cases considering the change of the conclusion part with respect to the change of the condition part, and the condition part of the similar case by the continuity of the condition part and the conclusion part. If the phase of the inference result is close to that of the conclusion part, the case-based reasoning part consisting of means for judging the authenticity of the inference result and means for outputting the information of the inference result and its authenticity, and the new case as they are A case-based learning unit that learns in real time and updates the case base is provided.

【0005】[0005]

【作用】新事例が入力された場合、新事例の結論部の推
論に用いる事例を既存事例の中から選択して類似度を演
算し、類似度に基づく類似事例と新事例との距離を演算
すると共に距離による重要度を演算し、この重要度を考
慮しながら新事例の推論を行う一方、推論に用いる類似
事例が検索された場合、類似度によりその信憑性を演算
する。
[Operation] When a new case is input, a case to be used for inference of the conclusion part of the new case is selected from existing cases, the similarity is calculated, and the distance between the similar case and the new case is calculated based on the similarity. At the same time, the importance according to the distance is calculated, and the new case is inferred while considering the importance. On the other hand, when a similar case used for the inference is retrieved, the credibility is calculated by the similarity.

【0006】[0006]

【実施例】以下、本発明について図面を参照して説明す
る。図1は、本発明に係る事例ベース推論装置を適用し
たシステムの一実施例を示すブロック図であり、このシ
ステムは、地域冷暖房の空調負荷予測システムである。
同図において、1は過去に例えば或る時間毎に観測され
た外気温やこのときの不快指数等の事例データを記憶す
る記憶装置、2はこの事例データを各月毎に類別するデ
ータ類別部、10は各月毎の事例データから認識モデル
を作成すると共に,実際に外気温等の新たなデータが入
力されたときにこの認識モデルに基づき各月への帰属性
を認識し負荷予測を行う状態認識自動生成装置、20は
各月毎の事例データから各月の因果関係モデルを作成す
ると共に,新しいデータが入力されたときにこの因果関
係モデルに基づいて推論を行いかつその信憑性を算出す
る推論装置、30は負荷予測値表示部、40は信憑性統
合手段、50は信憑性表示部である。
DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of a system to which the case-based reasoning apparatus according to the present invention is applied. This system is an air conditioning load prediction system for district cooling and heating.
In the figure, 1 is a storage device for storing case data such as an outside temperature observed in the past at a certain time interval and a discomfort index at that time, and 2 is a data classification unit for classifying the case data for each month. Numeral 10 creates a recognition model from case data for each month, and when new data such as the outside temperature is actually input, recognizes the attribution to each month based on this recognition model and predicts the load. The state recognition automatic generation device, 20 creates a causal relationship model for each month from case data for each month, and when new data is input, infers based on this causal relationship model and calculates its credibility. An inference device, 30 is a load prediction value display unit, 40 is a credibility integrating unit, and 50 is a credibility display unit.

【0007】ここで、状態認識自動生成装置10は、認
識モデル生成手段11、認識モデル学習手段12、認識
モデル記憶部13、認識モデル状況入力手段14、帰属
性認識手段15、最終予測値決定手段16から構成され
ている。そして認識モデル生成手段11は、過去の事例
データから認識のための要因抽出を行い各要因による認
識モデルを生成すると共に、認識モデル学習手段12は
外気温等の新事例データを入力した場合にこれを学習し
て上記認識モデルの更新等を行い、これらの認識モデル
は認識モデル記憶部13に記憶される。一方、帰属性認
識手段15は、認識モデル状況入力部14からの新事例
データを入力した場合、記憶部13の認識モデルに基づ
いて各要因への帰属性を認識してこの認識結果を最終予
測値決定手段16へ送出すると共に、最終予測値決定手
段16は、この認識結果と推論装置20の推論結果とを
統合して最終の負荷予測値として出力するものとなって
いる。
Here, the state recognition automatic generation apparatus 10 includes a recognition model generation means 11, a recognition model learning means 12, a recognition model storage section 13, a recognition model status input means 14, a membership recognition means 15, and a final predicted value determination means. It is composed of 16. Then, the recognition model generation means 11 extracts factors for recognition from past case data to generate a recognition model by each factor, and the recognition model learning means 12 detects this when new case data such as the outside temperature is input. Are learned to update the recognition model and the like, and these recognition models are stored in the recognition model storage unit 13. On the other hand, when the new model data from the recognition model situation input unit 14 is input, the attribution recognition unit 15 recognizes the attribution to each factor based on the recognition model of the storage unit 13 and finally predicts the recognition result. The final predictive value determining means 16 integrates the recognition result and the inference result of the inferring device 20 and outputs the final load predictive value, while sending the value to the value determining means 16.

【0008】また、推論装置20は、因果関係モデル生
成手段21、因果関係モデル学習手段22、因果関係モ
デル記憶部23、因果関係モデル状況入力部24、推論
手段25、信憑性計算手段26から構成されている。そ
して、因果関係モデル生成手段21は、過去の事例デー
タについてその入力要因とその数時間後の結果である例
えば室温との因果関係のモデルを生成すると共に、因果
関係モデル学習手段は、新事例データが入力された場合
にこれを学習して上記因果関係モデルの更新等を行い、
これらの因果関係モデルは因果関係モデル記憶部23に
記憶される。一方、推論手段25は、因果関係モデル状
況入力部24からの新事例データを入力すると、この新
事例データについて記憶部13の因果関係モデルに基づ
き推論を行い、この推論結果を上記最終予測値決定手段
16へ送出して状態認識自動生成装置10による認識結
果と統合させ、負荷予測値表示部30にこの予測値を表
示させる。また推論手段25はその推論結果を信憑性計
算手段26へ送出して信憑性の演算を行わせ、信憑性表
示部50に表示させる。なお、推論手段25は、新事例
データを推論する場合、位相(topology)に基
づいて推論するようにしている。この位相とは、集合に
連続の概念が定義できるように与えられる構造のことを
言い、例えば新事例データと過去の事例データとの距離
や類似性の近さ等を示している。
Further, the inference apparatus 20 comprises a causal relationship model generating means 21, a causal relationship model learning means 22, a causal relationship model storage section 23, a causal relationship model situation input section 24, an inference means 25 and a credibility calculation means 26. Has been done. Then, the causal relationship model generating means 21 generates a model of a causal relationship between the input factor of the past case data and room temperature which is the result several hours after that, and the causal relationship model learning means uses the new case data. When is input, it is learned to update the above causal relationship model,
These causal relationship models are stored in the causal model storage unit 23. On the other hand, when the new case data is input from the causal relationship model status input unit 24, the inference unit 25 makes an inference on the new case data based on the causal relationship model in the storage unit 13, and determines the inference result as the final predicted value determination. It is sent to the means 16 and integrated with the recognition result by the state recognition automatic generation device 10 to display the predicted value on the load predicted value display unit 30. Further, the inference means 25 sends the inference result to the credibility calculation means 26 to perform the credibility calculation and display it on the credibility display section 50. When inferring new case data, the inference unit 25 is configured to infer based on the topology. This phase means a structure given to the set so that the concept of continuity can be defined, and indicates, for example, the distance between the new case data and the past case data, the closeness of similarity, or the like.

【0009】ここで推論装置20は、ハード的には、図
3に示すように、既存事例データ(過去の事例データ)
を入力するオフライン装置61、新事例データを入力す
るオンライン入力装置62、アルゴリズム記憶部63、
アルゴリズム記憶部63に記憶された所定のアルゴリズ
ムにしたがって推論動作を行うCPU64、各種のデー
タを記憶するデータ記憶部65、及び出力装置66から
構成されており、また機能的には、図2に示す各手段か
ら構成されている。
[0012] Here, the inference device 20 is, in terms of hardware, as shown in FIG. 3, existing case data (past case data).
Offline device 61 for inputting, online input device 62 for inputting new case data, algorithm storage unit 63,
It is composed of a CPU 64 that performs an inference operation according to a predetermined algorithm stored in the algorithm storage unit 63, a data storage unit 65 that stores various data, and an output device 66, and functionally shown in FIG. It is composed of each means.

【0010】即ち、上記した因果関係モデル生成手段2
1としてのメモリベース作成手段21A,パラメータ決
定手段21B、上記した因果関係モデル学習手段22、
推論手段25としての類似度決定手段25A,類似度事
例検索手段25B,重要度決定手段25C,事例統合手
段25D、上記した信憑性計算手段26、及び出力手段
27から構成されている。
That is, the above-mentioned causal relationship model generating means 2
1, memory base creating means 21A, parameter determining means 21B, causal relationship model learning means 22 described above,
The inference means 25 includes a similarity determining means 25A, a similarity degree case searching means 25B, an importance determining means 25C, a case integrating means 25D, the credibility calculation means 26, and an output means 27.

【0011】推論装置20は、上記したように、過去の
事例データについてその入力要因とその数時間後の結果
である例えば室温との因果関係について扱うものであ
り、入力空間としてX=<x1,x2,・・・,xn>
を、また出力空間としてY=<y>を仮定した場合に、
時刻tにおいて生じた事象X1(t),X2(t),・
・・,Xn(t)がα時間後に出力Y(t+α)を生じ
るような、つまり入出力データとして{X1(t),X
2(t),・・・,Xn(t),Y(t+α)}(t=
1,・・・,N)の関係を有し、各入出力変数が連続的
に変化するようなデータを推論する装置である。
As described above, the inference device 20 handles the causal relationship between the input factor of the past case data and the room temperature which is the result after several hours, for example, X = <x1, as the input space. x2, ..., xn>
And Y = <y> as the output space,
Events X1 (t), X2 (t), occurring at time t
.., Xn (t) produces output Y (t + α) after α time, that is, as input / output data {X1 (t), X
2 (t), ..., Xn (t), Y (t + α)} (t =
1, ..., N), and a device for inferring data in which each input / output variable continuously changes.

【0012】このような推論装置においては、推論を行
う前にまず、因果関係モデル手段21により過去の事例
データについて事例ベースを作成する。即ち、まず入力
空間を離散化して有限個の入力事象に分割し、同一入力
現象に属する入出力データを統合化することで1つの事
例を生成する。このとき事例の条件部、つまり外気温や
不快指数等のデータは、離散化された入力データ{X
1,X2,・・・,Xn}となり、また事例の結論部、
つまりα時間後の室温は、出力データの重心値Y,同一
入力事象が起こった回数n及びその偏微分値の重心値Δ
Y/ΔX1,・・・,ΔY/ΔXn、即ち{Y,n,Δ
Y/ΔX1,・・・,ΔY/ΔXn}となる。
In such an inference apparatus, the causal relationship model means 21 first creates a case base for past case data before inferring. That is, first, an input space is discretized and divided into a finite number of input events, and input / output data belonging to the same input phenomenon is integrated to generate one case. At this time, the condition part of the case, that is, the data such as the outside temperature and the discomfort index is the discretized input data {X
1, X2, ..., Xn}, and the conclusion part of the case,
That is, the room temperature after α hours is the center of gravity value Y of the output data, the number of times n the same input event occurs, and the center of gravity value Δ of its partial differential
Y / ΔX1, ..., ΔY / ΔXn, that is, {Y, n, Δ
Y / ΔX1, ..., ΔY / ΔXn}.

【0013】例えば、入力データとして外気温X1
(℃)及び不快指数X2(%)が観測され、これらのデ
ータの或期間内の各時刻t(t=1,・・・,N)を通
しての最大値(max),最小値(min)がそれぞ
れ、 X1(max)=30.0,X1(min)=20.0 X2(max)=80.0,X2(min)=70.0 となり、かつ時刻t=t1 におけるデータ X1(t1 )=25.6,X2(t1 )=78.7 が得られたとき、この入力空間を離散化するための離散
化数を「10」(最大値と最小値間を10分割する)と
すれば、離散化されたデータは例えばX1=6,X2=
9として表され、{6,9}という1つの事象に象徴化
される。ここで、時刻t=t1 +αにおける室温Y(t
1 +α)を25.0とすると、{6,9}→25.0と
いう因果関係が得られたことになる。
For example, as the input data, the outside temperature X1
(° C.) and discomfort index X2 (%) are observed, and the maximum value (max) and minimum value (min) of these data at each time t (t = 1, ..., N) within a certain period are X1 (max) = 30.0, X1 (min) = 20.0 X2 (max) = 80.0, X2 (min) = 70.0, respectively, and data at time t = t 1 X1 (t 1 ) = 25.6, X2 (t 1 ) = 78.7 is obtained, the discretization number for discretizing this input space is set to “10” (the maximum value and the minimum value are divided into 10). Then, the discretized data is, for example, X1 = 6, X2 =
It is represented as 9 and is symbolized as one event {6, 9}. Here, at the time t = t 1 + α, the room temperature Y (t
When 1 + α) is set to 25.0, it means that the causal relation of {6,9} → 25.0 is obtained.

【0014】また、時刻t=t1 +1における事象が X1(t1 )=25.8,X2(t1 )=78.79で
室温Y(t1 +α+1)=25.5 であるときは、入力事象は同一の{6,9}に属し、同
一入力事象に対して生じた事象として {6,9}→25.25[={25.0+25.5}/
2、ここでn=2] として平均化され、事例データの圧縮が行われる。この
結果、事例ベースに要するメモリの容量を従来例に比べ
格段に少なくできる。また、上記した各偏微分値とは、
各入力変数の変化量に対する出力の変化量であり、この
場合各入出力変数が連続データであることから、この偏
微分値ΔY/ΔXi(t)は(1)式により計算するこ
とができる。
When the event at time t = t 1 +1 is X1 (t 1 ) = 25.8, X2 (t 1 ) = 78.79 and room temperature Y (t 1 + α + 1) = 25.5, The input events belong to the same {6,9}, and as an event generated for the same input event, {6,9} → 25.25 [= {25.0 + 25.5} /
2, where n = 2] and the case data is compressed. As a result, the memory capacity required for the case base can be significantly reduced as compared with the conventional example. Also, with each of the above partial differential values,
This is the amount of change in output with respect to the amount of change in each input variable. In this case, since each input / output variable is continuous data, this partial differential value ΔY / ΔXi (t) can be calculated by the equation (1).

【0015】[0015]

【数1】 [Equation 1]

【0016】以上のように、過去の事例(既存事例)に
ついて事例ベースの作成を行った後、次に新事例につい
ての推論を推論手段25により行う。まず、新事例の条
件部を{Xi* }(i=1,2,・・・,n)とし、既
存事例を{Xi,Y,n,ΔY/ΔXi}(i=1,
2,・・・,n)とする。ここで、新事例の条件部は、
入力と同時に上記のように離散化され、かつ象徴化され
ている。
As described above, after the case base is created for the past case (existing case), the inference means 25 infers the new case. First, the conditional part of the new case is set to {Xi * } (i = 1, 2, ..., N), and the existing case is set to {Xi, Y, n, ΔY / ΔXi} (i = 1,
2, ..., N). Here, the condition part of the new case is
Simultaneously with the input, it is discretized and symbolized as described above.

【0017】ここで新事例の推論を行う場合は、まず新
事例に対する既存事例の類似度を類似度決定手段25A
により決定する(この類似度とは、位相における近傍系
という概念に相応する)。新事例に対する既存事例の類
似度は次のような定義により決定する。即ち、 類似度0は、|Xi* −Xi|=0 (i=1,
2,・・・,n) 類似度1は、|Xi* −Xi|≦qXi (i=1,
2,・・・,n) 類似度2は、|Xi* −Xi|≦qXi+1(i=1,
2,・・・,n) 類似度3は、|Xi* −Xi|≦qXi+2(i=1,
2,・・・,n) のように定義する。ここで、qXiはしきい値と呼ばれ、
既存事例データから、Y(既存事例の結論部)の許容精
度に対するXi(既存事例の条件部)の分散により決定
したデジット値である。
When inferring a new case, first, the similarity of the existing case to the new case is determined by the similarity determining means 25A.
(This similarity corresponds to the concept of a neighborhood system in phase). The similarity of the existing case to the new case is determined by the following definition. That is, the similarity 0 is | Xi * −Xi | = 0 (i = 1,
2, ..., n) The similarity 1 is | Xi * −Xi | ≦ q Xi (i = 1,
2, ..., n) Similarity 2 is | Xi * −Xi | ≦ q Xi +1 (i = 1,
2, ..., n) The similarity 3 is | Xi * −Xi | ≦ q Xi +2 (i = 1,
2, ..., N). Where q Xi is called the threshold,
It is a digit value determined from the existing case data by the variance of Xi (condition part of the existing case) with respect to the allowable accuracy of Y (the conclusion part of the existing case).

【0018】一例として、X1を外気温,X2を不快指
数とし,Y=α時間後の室温を考えた場合、室温Yと予
測値との誤差が2度以内の予測をしようとしたとき、入
力空間と同様に1デジットが2度となるように離散化を
行う。そして離散化数が例えば「20」であったとする
と、 デジット1〜20それぞれに対し同一デジットに属す
るY* (新事例の結論部)を既存事例データから収集し
クラスタリングを行い、同一のクラスに属した新事例の
条件部X1,X2の分散を求める。 Y(既存事例の結論部)のデジットiに属する新事例
の条件部X1,X2の各クラスターの分散を離散値とし
て計算し、クラスター数で平均化することでデジット値
qiX1,qiX2を求める。 デジット値qiX1,qiX2を下式にしたがって平均化
する。即ち、 qX1=ΣqiX1/20,qX2=ΣqiX2/20
As an example, when X1 is the outside temperature and X2 is the discomfort index, and the room temperature after Y = α hours is considered, when the error between the room temperature Y and the predicted value is to be predicted within 2 degrees, the input is made. Similar to space, discretization is performed so that one digit becomes twice. If the discretization number is, for example, “20”, Y * (conclusion part of new case) belonging to the same digit for each of digits 1 to 20 is collected from the existing case data, clustered, and belongs to the same class. The variance of the conditional parts X1 and X2 of the new case is calculated. Calculate the variance of each cluster of conditional parts X1 and X2 of the new case belonging to the digit i of Y (conclusion part of the existing case) as a discrete value, and obtain the digit values qi X1 and qi X2 by averaging by the number of clusters. . The digit values qi X1 and qi X2 are averaged according to the following equation. That is, q X1 = Σqi X1 / 20, q X2 = Σqi X2 / 20

【0019】ここで、最も近いデジット値として、qX1
=2,qX2=3(実際には、qX1=2.123・・・,
X2=3.456・・・)とする。しかし、Yが1度以
内を必要とすれば、qX1=1,qX2=2となり、要求さ
れる精度によってデジット値qXiは異なる。即ち、各変
数Xiに対し既存事例と新事例との距離がqXiより小さ
ければ条件部の位相が近いとされ、その時には新事例の
結論部は、既存事例の結論部に対して要求される精度内
に入っていると見なされる。
Here, q X1 is the closest digit value.
= 2, q X2 = 3 (actually, q X1 = 2.123 ...,
q X2 = 3.456 ...). However, if Y needs to be within 1 degree, q X1 = 1 and q X2 = 2, and the digit value q Xi differs depending on the required accuracy. That is, the distance between the existing case and new case for each variable Xi is the phase of the condition part is smaller than q Xi is closer, the conclusion part of the new case at that time is required for the conclusion of the existing case Considered to be within precision.

【0020】次に、類似事例検索手段25Bにより新事
例に対する類似事例を検索する。新事例に対する類似度
が高い順に、最適事例の既存事例を類似事例として抽出
する。この最適事例数は、例えば既存事例によるシミュ
レーションから最も推論が良くなる事例数を選択する。
同一類似度の既存事例が最適事例数より多く存在する場
合は、各変数XiがYに与える影響度、即ち相関係数R
Xiの大小によって各変数に優先度を設定して抽出する。
Next, the similar case retrieval means 25B retrieves a similar case for the new case. The existing cases of the optimum cases are extracted as similar cases in descending order of similarity to the new case. For this optimum number of cases, for example, the number of cases that gives the best inference from a simulation based on existing cases is selected.
When there are more existing cases with the same degree of similarity than the optimum number of cases, the degree of influence of each variable Xi on Y, that is, the correlation coefficient R
Priority is set for each variable depending on the size of Xi and extraction is performed.

【0021】次に、重要度決定手段25Cにより新事例
に対する類似事例の重要度を決定する。入力空間に距離
を定義して事例間の位相を考慮する。ここでは例として
(2)式に示すような距離Lを導入する。
Next, the importance determining means 25C determines the importance of the similar case with respect to the new case. Consider the phase between cases by defining the distance in the input space. Here, as an example, the distance L as shown in the equation (2) is introduced.

【0022】[0022]

【数2】 [Equation 2]

【0023】ここで、Φiは変数Xiにおける距離の重
みである。そして抽出されたm個の類似事例の推論時の
重要度Wjを(3)式を用いて定義する。即ち、
Here, Φi is a distance weight in the variable Xi. Then, the importance Wj at the time of inference of the extracted m similar cases is defined by using the expression (3). That is,

【0024】[0024]

【数3】 [Equation 3]

【0025】なお図5は、このような類似事例の重要
度、即ち類似事例の重み付け前後の状況を説明する説明
図である。こうして抽出された最適事例数m個の類似事
例を用いて、新事例Xi* (i=1,2,・・・,n)
に対する推論値Y* を(4)式を用いて計算し統合化す
る(事例統合手段25D)。
FIG. 5 is an explanatory diagram for explaining the importance of such similar cases, that is, the situation before and after weighting of similar cases. A new case Xi * (i = 1, 2, ..., N) is created by using the similar cases with the optimum number of cases m extracted in this way.
The inference value Y * for is calculated and integrated using the equation (4) (case integration means 25D).

【0026】[0026]

【数4】 [Equation 4]

【0027】ここで、Lijはi番目の事例のj入力変数
軸上での入力データからの距離、yiはi番目の類似既
存事例の結論値、ΔY/ΔXjはi番目の類似既存事例
のj番目の変動が結論値に与える変動の割合をそれぞれ
示している。
Here, Lij is the distance from the input data on the j input variable axis of the i-th case, yi is the conclusion value of the i-th similar existing case, and ΔY / ΔXj is j of the i-th similar existing case. The respective changes of the second change give the conclusion value.

【0028】次に推論結果の信憑性判定を信憑性計算手
段26の計算結果に基づいて行う。即ち、推論に使用さ
れた類似事例の新事例に対する類似度を用い、推論結果
に対する信憑性を判定する。例えば、推論に用いられた
類似事例の新事例に対する類似度の中で、最も高い類似
度がこの推論結果に対する信憑性であると定義すると、
最も高い類似度が「1」である推論結果は、信憑性が
「1」であると判定される。この場合、信憑性「0」が
最も信憑度が高く、数字が大きくなるにつれ信憑度が低
くなる。出力手段27は、こうして得られた推論結果及
びその信憑性を出力すると共に推論に使用した類似事例
を出力することもできる。
Next, the credibility of the inference result is judged based on the calculation result of the credibility calculation means 26. That is, the credibility of the inference result is determined by using the similarity of the similar case used for the inference to the new case. For example, if we define the highest similarity among the similarities used for inference for new cases as the credibility for this inference result,
The inference result having the highest similarity of “1” is determined to have the credibility of “1”. In this case, the credibility "0" has the highest credibility, and the credibility decreases as the number increases. The output unit 27 can output the inference result thus obtained and its credibility, and can also output the similar case used for the inference.

【0029】次に事例ベース学習を行う場合には、因果
関係モデル学習手段22により新事例を学習して事例ベ
ースを更新する。このような事例ベースの更新は次の手
順により行われる。ただし、*が付いているものは新事
例を示している。即ち、前回までの同一条件部の事象回
数をnとすると、この事象回数を1つ増加させてn+1
にすると共に、出力値Yを(Y×n+Y* )/(n+
1)とし、さらに偏微分値ΔY/ΔX1を(ΔY/ΔX
1×n+ΔY/ΔX1* )/(n+1)とする。
When case-based learning is performed next, the causal relationship model learning means 22 learns a new case and updates the case base. Such case-based updating is performed by the following procedure. However, those marked with * indicate new cases. That is, assuming that the number of events of the same condition part up to the previous time is n, the number of events is increased by 1 to n + 1.
And the output value Y is (Y × n + Y * ) / (n +
1), and the partial differential value ΔY / ΔX1 is (ΔY / ΔX
1 × n + ΔY / ΔX1 * ) / (n + 1).

【0030】次に、図4は以上のような推論を行う推論
手段25の動作を要約して示したフローチャートであ
る。即ち、新事例の条件部としての外気温等の新事例デ
ータXが入力された場合、新事例の結論部である所定時
間後の室温の推論に用いる事例を既存事例の中から選択
し、新事例と既存事例との類似度を計算すると共に、類
似事例を検索する(ステップST1)。そして、推論に
用いる検索された類似事例と新事例との距離を計算する
と共に、距離による重要度を計算し(ステップST
2)、この重要度を考慮しながら新事例の結論部の推論
を行い(ステップST3)、推論結果yを得る(ステッ
プST4)。一方、推論に用いる類似事例が検索された
場合、類似度によりその信憑性を計算し(ステップST
5)、その結果を得る(ステップST6)。そして、こ
の信憑性結果と推論結果yとは出力される(ステップS
T8)。
Next, FIG. 4 is a flow chart summarizing the operation of the inference means 25 for performing the above inference. That is, when the new case data X such as the outside temperature as the condition part of the new case is input, the case used for the inference of the room temperature after the predetermined time which is the conclusion part of the new case is selected from the existing cases and the new case is selected. The similarity between the case and the existing case is calculated, and the similar case is searched (step ST1). Then, the distance between the retrieved similar case used for inference and the new case is calculated, and the importance according to the distance is calculated (step ST
2) The inference of the conclusion part of the new case is performed in consideration of this importance (step ST3), and the inference result y is obtained (step ST4). On the other hand, when a similar case used for inference is retrieved, its credibility is calculated based on the similarity (step ST
5), the result is obtained (step ST6). Then, the credibility result and the inference result y are output (step S
T8).

【0031】ここで事実Y(新事例の結論部)が得られ
た場合、この事実Yと推論結果yとを比較する(ステッ
プST9)。そして、事実Yと推論結果yとが異なり、
かつ新事例Xが過去の事例に存在しない場合は、新事例
の条件部であるXとその結論部であるYとを新事例とし
て登録する(ステップST10)。
If the fact Y (the conclusion of the new case) is obtained here, the fact Y is compared with the inference result y (step ST9). And the fact Y and the inference result y are different,
If the new case X does not exist in the past cases, the condition part X of the new case and its conclusion part Y are registered as new cases (step ST10).

【0032】[0032]

【発明の効果】以上説明したように、本発明によれば、
新事例が入力された場合、新事例の結論部の推論に用い
る事例を既存事例の中から選択して類似度を演算し、類
似度に基づく類似事例と新事例との距離を演算すると共
に距離による重要度を演算し、この重要度を考慮しなが
ら新事例の推論を行う一方、推論に用いる類似事例が検
索された場合、類似度によりその信憑性を演算するよう
にしたので、事例データが連続的に変化する複雑なシス
テムの予測,制御,同定を行う際に不可欠の、高精度の
推論,推論結果の信憑性評価,及びリアルタイムの適応
学習が可能になると共に、事例データが象徴化されるこ
とによってメモリの容量や計算量が削減され、この結果
小規模コントローラによるリアルタイムな処理実行が可
能になる。
As described above, according to the present invention,
When a new case is entered, the case to be used for inference in the conclusion part of the new case is selected from the existing cases, the similarity is calculated, and the distance between the similar case and the new case based on the similarity is calculated and the distance is calculated. While inferring a new case while taking into account this importance, the credibility is calculated based on the similarity when a similar case used for inference is searched. High-accuracy inference, credibility evaluation of inference results, and real-time adaptive learning, which are indispensable for predicting, controlling, and identifying continuously changing complex systems, are possible, and case data are symbolized. As a result, the memory capacity and calculation amount are reduced, and as a result, real-time processing can be executed by a small-scale controller.

【図面の簡単な説明】[Brief description of drawings]

【図1】本発明に係る事例ベース推論装置を適用したシ
ステムの一実施例を示す機能ブロック図である。
FIG. 1 is a functional block diagram showing an embodiment of a system to which a case-based reasoning apparatus according to the present invention is applied.

【図2】上記推論装置の機能ブロック図である。FIG. 2 is a functional block diagram of the inference apparatus.

【図3】上記推論装置のブロック図である。FIG. 3 is a block diagram of the inference apparatus.

【図4】上記推論装置の動作を示すフローチャートであ
る。
FIG. 4 is a flowchart showing an operation of the inference apparatus.

【図5】上記推論装置において類似事例の重要度の計算
による重み付け状況を示す図である。
FIG. 5 is a diagram showing a weighting situation by calculation of importance of similar cases in the inference apparatus.

【符号の説明】[Explanation of symbols]

21 因果関係モデル生成手段 21A メモリベース作成手段 21B パラメータ決定手段 22 因果関係モデル学習手段 25 推論手段 25A 類似度決定手段 25B 類似事例検索手段 25C 重要度決定手段 25D 事例統合手段 26 信憑性計算手段 27 出力手段 21 causal relationship model generating means 21A memory base creating means 21B parameter determining means 22 causal relationship model learning means 25 inference means 25A similarity determining means 25B similar case searching means 25C importance determining means 25D case integrating means 26 credibility calculating means 27 output means

Claims (1)

【特許請求の範囲】[Claims] 【請求項1】 各事例間に連続性が成立するような事例
の推論を行う推論装置であって、 既存事例の条件部の各入力変数の値を離散化しこの離散
値により条件部を象徴化すると共に同一象徴に対する既
存事例の結論部を統合化しこの結論部と条件部との関係
と,条件部の変化に対する結論部の変化情報とを付随し
て生成する事例ベース生成部と、推論を行うための前処
理として既存事例から推論精度を考慮した条件部の類似
度を判定するために用いるしきい値を決定する手段と,
推論に用いる類似事例の個数の制約を決定する手段とか
らなるパラメータ決定部と、新事例の条件部に対する既
存事例の類似度を条件部の位相の連続性に基づいて決定
する手段と,条件部の類似度に基づき事例ベースから新
事例に対する類似事例を条件部が結論部に与える影響度
を考慮して検索する手段と,新事例に対する類似事例の
重要度を位相による条件部の距離から決定する手段と,
重要度に基づき複数の類似事例を条件部の変化に対する
結論部の変化を考慮して修正を行い統合化し新事例の結
論部を推論する手段と,条件部及び結論部の連続性によ
り類似事例の条件部の位相が近ければ結論部の位相も近
いことを用い推論結果の信憑性を判定する手段と,推論
結果及びその信憑性の情報を出力する手段とからなる事
例ベース推論部と、新事例をそのままリアルタイムに学
習し事例ベースを更新する事例ベース学習部とを備えた
ことを特徴とする事例ベース推論装置。
1. An inference device for inferring cases in which continuity is established between each case, wherein the values of each input variable of the condition part of an existing case are discretized and the condition part is symbolized by the discrete value. In addition, the case base generation unit that integrates the conclusion part of the existing case for the same symbol and generates the relationship between the conclusion part and the condition part and the change information of the conclusion part for the change of the condition part is inferred. As a pre-processing for the above, means for determining the threshold value used to judge the similarity of the conditional part considering the inference accuracy from the existing case,
A parameter deciding unit comprising means for deciding the number of similar cases used for inference, means for deciding the similarity of the existing case to the condition section of the new case based on the phase continuity of the condition section, and the condition section A method for searching similar cases for a new case from the case base in consideration of the degree of influence of the conditional part on the conclusion part based on the similarity of the method, and determining the importance of the similar case for the new case from the distance of the conditional part by phase Means,
A method of inferring the conclusion part of a new case by modifying and consolidating multiple similar cases considering the change of the conclusion part with respect to the change of the condition part based on the importance, and the continuity of the condition part and the conclusion part A case-based reasoning unit consisting of a means for judging the credibility of the inference result by using the fact that the phase of the condition part is close to that of the conclusion part, and a means for outputting the information of the reasoning result and the credibility, and a new case A case-based reasoning apparatus comprising: a case-based learning unit that learns the real time as it is and updates the case base.
JP4269129A 1992-08-31 1992-09-14 Case-based reasoning device Expired - Lifetime JP2632117B2 (en)

Priority Applications (5)

Application Number Priority Date Filing Date Title
JP4269129A JP2632117B2 (en) 1992-09-14 1992-09-14 Case-based reasoning device
US08/109,179 US5918200A (en) 1992-08-31 1993-08-19 State estimating apparatus
DE69328956T DE69328956T2 (en) 1992-08-31 1993-08-25 System for estimating the state of a system based on recorded input-output data for the system
EP93113564A EP0590305B1 (en) 1992-08-31 1993-08-25 State estimating apparatus of a system on the basis of recorded input/output data for the system
CN93118822A CN1047011C (en) 1992-08-31 1993-08-31 State estimating apparatus

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
JP4269129A JP2632117B2 (en) 1992-09-14 1992-09-14 Case-based reasoning device

Publications (2)

Publication Number Publication Date
JPH0695880A true JPH0695880A (en) 1994-04-08
JP2632117B2 JP2632117B2 (en) 1997-07-23

Family

ID=17468098

Family Applications (1)

Application Number Title Priority Date Filing Date
JP4269129A Expired - Lifetime JP2632117B2 (en) 1992-08-31 1992-09-14 Case-based reasoning device

Country Status (1)

Country Link
JP (1) JP2632117B2 (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2000155681A (en) * 1998-11-24 2000-06-06 Fujitsu Ltd Predicting device performing prediction based on analogous example and method
JP2000172671A (en) * 1998-12-11 2000-06-23 Fujitsu Ltd Display device for, result of similar prediction by k- neighborhood method
WO2007080688A1 (en) 2006-01-13 2007-07-19 Jfe Steel Corporation Prediction formula making device and prediction formula making method
WO2014002645A1 (en) * 2012-06-29 2014-01-03 アズビル株式会社 Load value prediction device and load value prediction method

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP4634242B2 (en) * 2005-07-08 2011-02-16 株式会社山武 Energy saving amount estimation apparatus, method, and program
JP4486004B2 (en) * 2005-07-27 2010-06-23 株式会社山武 River pollution load estimation system, method, and program

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2000155681A (en) * 1998-11-24 2000-06-06 Fujitsu Ltd Predicting device performing prediction based on analogous example and method
JP2000172671A (en) * 1998-12-11 2000-06-23 Fujitsu Ltd Display device for, result of similar prediction by k- neighborhood method
WO2007080688A1 (en) 2006-01-13 2007-07-19 Jfe Steel Corporation Prediction formula making device and prediction formula making method
WO2014002645A1 (en) * 2012-06-29 2014-01-03 アズビル株式会社 Load value prediction device and load value prediction method
JP2014010683A (en) * 2012-06-29 2014-01-20 Azbil Corp Load amount prediction device and load amount prediction method

Also Published As

Publication number Publication date
JP2632117B2 (en) 1997-07-23

Similar Documents

Publication Publication Date Title
Chen et al. Fuzzy forecasting based on two-factors second-order fuzzy-trend logical relationship groups and particle swarm optimization techniques
Filev et al. An Extended Version of the Gustafson‐Kessel Algorithm for Evolving Data Stream Clustering
Hassan et al. A hybrid of multiobjective Evolutionary Algorithm and HMM-Fuzzy model for time series prediction
EP0645025A1 (en) System and method for controlling a plant
JP3762840B2 (en) Prediction apparatus and method for performing prediction based on similar cases
Petelin et al. Control system with evolving Gaussian process models
JP2021103083A (en) Machine learning device, demand control system, and air conditioning control system
Haouari et al. Naïve possibilistic network classifiers
JP7481902B2 (en) Management computer, management program, and management method
US11624624B2 (en) Methods and apparatuses for predicting user destinations
Wolkenhauer Possibility theory with applications to data analysis
Lemos et al. Evolving intelligent systems: Methods, algorithms and applications
JP2632117B2 (en) Case-based reasoning device
Kargin et al. Method of using data from intelligent machine short-term memory in Fuzzy Logic System
JP2802469B2 (en) State prediction device
Trofimov Prototype approach to design of the automated expert systems with multi-structural recognition of complex images
El Afia et al. Fuzzy logic controller for an adaptive Huang cooling of simulated annealing
JP3284280B2 (en) Existing case effective use type inference device
Sahin et al. Review study on fuzzy time series and their applications in the last fifteen years
Malaviya et al. Automatic generation of fuzzy rule base for online handwriting recognition
WO2024128090A1 (en) Information processing device, control system, search method, and search program
JPH06187318A (en) Adaptive learning device
CN114511023B (en) Classification model training method and classification method
Yan et al. Matching decision method for knowledgeable manufacturing system and its production environment
Modirrousta et al. DQLAP: Deep Q-Learning Recommender Algorithm with Update Policy for a Real Steam Turbine System

Legal Events

Date Code Title Description
FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20080425

Year of fee payment: 11

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20080425

Year of fee payment: 11

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20090425

Year of fee payment: 12

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20090425

Year of fee payment: 12

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20100425

Year of fee payment: 13

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20100425

Year of fee payment: 13

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20110425

Year of fee payment: 14

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20120425

Year of fee payment: 15

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20130425

Year of fee payment: 16

EXPY Cancellation because of completion of term
FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20130425

Year of fee payment: 16