WO2017046906A1 - データ分析装置および分析方法 - Google Patents
データ分析装置および分析方法 Download PDFInfo
- Publication number
- WO2017046906A1 WO2017046906A1 PCT/JP2015/076363 JP2015076363W WO2017046906A1 WO 2017046906 A1 WO2017046906 A1 WO 2017046906A1 JP 2015076363 W JP2015076363 W JP 2015076363W WO 2017046906 A1 WO2017046906 A1 WO 2017046906A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- node
- records
- data
- value
- variable
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
- G06N5/045—Explanation of inference; Explainable artificial intelligence [XAI]; Interpretable artificial intelligence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/02—Knowledge representation; Symbolic representation
- G06N5/022—Knowledge engineering; Knowledge acquisition
- G06N5/025—Extracting rules from data
Definitions
- the present invention relates to a data analysis apparatus and an analysis method.
- ⁇ Analysis of output including various system states or intermediate results is required.
- An example of a system that requires state analysis is a control system, and system state analysis is required in order to deal with control failures such as a failure or a difference from a predetermined control result.
- An example of a system that requires output analysis is a sales management system, which can be used to modify and formulate sales plans in relation to customer preferences and sales time (hours, days of the week, months, etc.) It is necessary to analyze the output including intermediate results such as sales. These requests include accurate analysis.
- Patent Document 1 describes that a decision tree corresponding to a classification rule inherent in data is generated, and new attribute data is added to a classification rule with low classification accuracy to improve classification accuracy. .
- Patent Document 1 limits the number of attribute types of data and increases the number of attribute types of data when a desired classification accuracy cannot be obtained, thereby obtaining one combination of data attributes constituting a classification rule.
- the state and output are not always specified by one rule.
- the attribute of data reflecting the cause of the system abnormality or a combination thereof is not necessarily limited to one. That is, the improvement in accuracy includes not overlooking a plurality of important rules.
- the data analysis device is required to visualize the rules for classifying data, eliminating the complexity of modeling.
- a data analysis apparatus for analyzing data having a record including an objective variable and a plurality of explanatory variables creates a node defined by the conditions of the explanatory variable based on the objective variable and the explanatory variable of the record,
- a node creation unit for associating with a node
- an evaluation value generation unit for generating, as an evaluation value, a ratio of the number of records whose target variable is a target value of a plurality of records associated with the node, and based on the evaluation value
- a parameter extracting unit that selects a node and extracts and outputs the condition of the explanatory variable related to the selected node.
- rules for classifying data can be easily visualized.
- FIG. 1 is a configuration diagram of the data analysis apparatus 1.
- the data analysis apparatus 1 is connected to an analysis target system 2 such as a control system or a sales management system, and is connected to an output apparatus 3 that outputs a data analysis result.
- an analysis target system 2 such as a control system or a sales management system
- the data analysis apparatus 1 includes a data collection unit 10, a decision tree creation unit 30, an evaluation value generation unit 50, and a parameter extraction unit 80, a collection data table 20, a decision tree table 40, an evaluation result table 60, And the parameter table 90.
- the data analysis device 1 is a computer having a processing device that executes each processing unit and a storage device that stores each table.
- each processing unit is configured by a program running on a CPU (Central Processing Unit), and each table is configured by a database stored in a storage device.
- CPU Central Processing Unit
- the data collection unit 10 associates the data record collected from the analysis target system with the decision tree node that defines the node based on the condition relating to the data collected by the decision tree creation unit 30.
- the evaluation value generation unit 50 classifies the records into records belonging to each node based on the conditions relating to the data, further classifies the records into records according to the mode of the objective variable of the decision tree included in the data, A ratio of the number of records further classified with respect to the number of records belonging is generated as an evaluation value.
- the parameter extraction unit extracts a condition relating to data corresponding to an aspect of the objective variable based on a predetermined criterion relating to the generated evaluation value, and outputs the extracted condition.
- the operation of each processing unit is as follows.
- the data collection unit 10 stores the data collected from the analysis target system 2 in the collected data table 20.
- the decision tree creation unit 30 creates a decision tree from the record of the collected data table 20 and stores it in the decision tree table 40.
- the evaluation value generation unit 50 calculates an evaluation value of each node of the decision tree, and generates an evaluation result table 60 in which the evaluation value is added to the decision tree table 40.
- the parameter extraction unit 80 refers to the evaluation result table 60, extracts parameters (data attributes) representing features inherent in the collected data, stores them in the parameter table 90, and outputs them to the output device 3.
- the output device 3 is, for example, a communication device for network output, a display for display, a terminal equipped with the display, or a printer for printing.
- FIG. 2 is a process flowchart of the data collection unit 10.
- the data collection unit 10 collects data from the analysis target system 2 (S11), and stores the collected data in the collected data table 20 (S12).
- the data to be collected is sensor data in a broad sense.
- the broad sensor data includes data such as a control target value set by the control device and an abnormality detection output of the abnormality detection device in addition to data of sensors such as a thermometer and an ammeter. These sensor data are output from each sensor at a cycle of 100 ms, 1 second, etc. based on a timer (clock) of the analysis target system 2.
- sensor data that is output in response to the occurrence of an event called abnormality detection, such as the abnormality detection output of the abnormality detection device.
- These sensor data are associated with time data (time stamp) representing the time output from each sensor.
- the data collection unit 10 may collect data from each sensor according to the output cycle, or may collect data by inputting sensor data once accumulated in the analysis target system 2.
- the data to be collected is log data output by the information processing system periodically or in response to an event that occurs.
- Log data also includes a time stamp. Therefore, log data output according to the log type, such as data reception of the information processing system, data writing to the storage device, program execution status, etc., is treated as sensor data when the log type is regarded as a sensor. be able to.
- the data to be collected can be handled in the same manner as the sensor data regardless of the type of the analysis target system 2, and will be described simply as sensor data.
- FIG. 3 is a configuration example of the collected data table 20.
- the collected data table 20 has an ID 21 for identifying each record and sensor data 22 to 26 collected from each sensor.
- the state 26 is an abnormality detection output when the above-described abnormality detection device is regarded as a sensor.
- the sensor data 22 to 25 are treated as explanatory variables in the decision tree analysis, and the state 26 is treated as an objective variable. Since the decision tree analysis itself is well known, a description thereof will be omitted.
- the sensor data 22 to 26 included in each record of the collected data table 20 can be regarded as the same time or the same time based on each time data not shown. A thing within a predetermined time (for example, 1/10 of a period) from the reference time is associated.
- a timer for example, 1/10 of a period
- the time data time stamp of a specific sensor whose sensor data is collected in the collection data table 20 may be used.
- the stirring time and reaction time Since it is known in advance, it is preferable to make corrections at these times (advance or delay the time data of each sensor data) and align the sensor data at the same time.
- the unit of each sensor data in the collected data table 20 is omitted, and the number of records is 40.
- the state 26 which is an objective variable is represented by two states (an aspect of the objective variable) of a character string “normal” or “abnormal”.
- the state may be three or more states depending on the analysis target system 2.
- each state in the state transition diagram can be selected.
- the analysis target system 2 is a sales management system
- the sales of a specific item is increasing, the sales of a specific item is decreasing, and the sales of a specific item are above the target value.
- Various states can be selected, such as (or below).
- FIG. 4 is a process flowchart of the decision tree creation unit 30.
- FIG. 5 is a configuration example of the decision tree table 40 created by the decision tree creation unit 30.
- the decision tree table 40 includes a node ID 41 of each node of the decision tree, a conditional statement 42 that defines the node (a condition that determines whether or not the target record belongs to the node), and a node in the hierarchy of the decision tree (tree structure).
- Depth hierarchical depth: the number of branches from the root node to the corresponding node) 43, the number of records 44 of the collected data belonging to the node, and the record 26 of the collected data state 26 representing “abnormal”
- the number of records (abnormal number of records) 45, the number of records 46 in which the collected data state 26 indicates “normal” (number of normal records) 46, and the node ID 47 of the parent node into which the node is divided are shown. Including.
- the decision tree creation unit 30 repeats the association of the node ID 41 of each node of the decision tree, the conditional statement 42 that defines the node, and the hierarchical depth 43 from the root node to the leaf node (terminal node), and the decision tree table 40 (S31).
- the root node includes all the records in the collected data table 20, and the node ID 41 is “0” and the hierarchical depth 43 is “0”.
- the node with the node ID 41 of “1” includes a record whose value of the sensor D25 in the collected data table 20 is 104.9 or less as indicated by the conditional statement 42.
- the node whose node ID 41 is “2” includes a record in which the value of the sensor D25 is greater than 104.9. Since the nodes with node IDs 41 “1” and “2” are immediately below the node with node ID 41 “0”, the layer depth 43 is “1”. In the same manner, the association is repeated up to the leaf node.
- the order of determination of the conditional sentence 42 is the order of determination of the value of the sensor D25, determination of the value of the sensor A22 in this example, and so on. It is determined.
- the judgment criterion is a boundary value of the ratio of the number of normal / abnormal records represented by each record. For example, when the value of sensor D25 is larger than 104.9 with reference to 104.9 (node 2), it is only an abnormal state record, and when it is smaller than 104.9 (node 1), it is a normal state record and an abnormal state. Both records are included. Node 1 is further divided into nodes 3 to 5 based on the value of sensor A22. The value of sensor A22 is based on 19.9 and 20.1, and includes an abnormal record only if it is greater than 19.9 and less than 20.1 (node 4), and less than 19.9 (node 3) or greater than 20.1 (node 5) does not include an abnormal record. The same applies to other sensors based on the numerical values shown.
- the sensor data (explanatory variable) is used as a conditional statement to try multiple divisions, and the ratio of the target variable state (normal state, abnormal state) is well divided (the proportion of normal state is large)
- Sensor data and values (divided into nodes and nodes with a high proportion of abnormal states) are adopted as conditional statements for actual division, and node division is performed.
- the division of the divided nodes is repeated until a predetermined condition (node hierarchy depth, number of records in the node, ratio between normal state and abnormal state, etc.) is reached.
- Judgment criteria are not necessarily numerical standards.
- the sensor is not only a type that outputs the target level, but also a type that differentiates the level and outputs a change in the level as sensor data.
- the temperature (level) is differentiated, and the degree of ascent or descent is, for example, five stages ("rapid rise", “rise”, “no change", ...) Some output as data.
- a criterion for determination corresponding to a numerical value or character string representing a stage is used.
- the decision tree creation unit 30 stores the number of records 44 of the collected data table 20 corresponding to each node of the decision tree table 40 in the decision tree table 40 (S32).
- the decision tree creation unit 30 stores the number of abnormal records 45 and the number of normal records 46 of the collected data table 20 corresponding to each node of the decision tree table 40 in the decision tree table 40 (S33).
- the number of records shown in FIG. 5 is a numerical example when the number of records in the collected data table 20 is 40.
- FIG. 6 is a diagram representing the contents (decision tree) of the decision tree table 40 in a tree structure.
- a conditional statement 42 from the upper node to the lower node is described.
- the leaf nodes whose records included in FIG. 6 are all abnormal records are node 2 and node 8. This indicates that the state and output of the analysis target system 2 are not necessarily specified by one rule (type of sensor data or a combination thereof).
- FIG. 7 is a process flowchart of the evaluation value generation unit 50.
- FIG. 8 is a configuration example of the evaluation result table 60 generated by the evaluation value generation unit 50.
- the node ID 61, the conditional statement 62, the hierarchy depth 63, the record number 64, the abnormal record number 68, and the normal record number 71 in the evaluation result table 60 are the same as the corresponding items in the decision tree table 40.
- the evaluation value generation unit 50 copies the contents of the decision tree table 40 and generates the evaluation result table 60.
- the evaluation result table 60 includes a cover degree 65, an abnormal moderate degree 66, an abnormal fitness degree 67, a normal moderate degree 69, and a normal fitness degree 70.
- the normal moderate degree 69 and the normal fitness degree 70 are omitted because the explanation is that the abnormality of the abnormal intermediate degree 66 and the abnormal fitness degree 67 is read as normal.
- the abnormal moderate 66, abnormal fitness 67, normal moderate 69, and normal fitness 70 are evaluation values.
- Abnormal medium 66 and normal medium 69 are evaluation values representing an index of the accuracy of the rule based on the decision tree.
- the abnormal fitness level 67 and the normal fitness level 70 are evaluation values as a standard that represent the ratio of the number of abnormal records 68 or the number of normal records 71 included in the node to the total number of records. Therefore, the target level may be called an evaluation value.
- the evaluation value generation unit 50 obtains the cover degree 65 of each node as described above, and stores it in the evaluation result table 60 (S51). As described above, the evaluation value generation unit 50 obtains the abnormal moderate 66 (S52), the abnormal fitness 67 (S53), the normal moderate 69 (S54), and the normal fitness 70 (S55). Then, each is stored in the evaluation result table 60.
- FIG. 9 is a process flowchart of the parameter extraction unit 80.
- the parameter is each sensor data reflected in each state (hereinafter, normal) 26 that is an objective variable, and a condition represented by a conditional statement 62 relating to each sensor data is a rule. Therefore, each rule includes parameters (data) and conditions based thereon.
- selection criteria are set in advance, such as a rule having a target level of 100% as an evaluation value, a rule having a target level greater than or equal to a predetermined value, and a rule having a target level of k from the top.
- the selection criteria may be set by the user of the data analysis apparatus.
- a node having a large number of records may be preferentially selected in consideration of values related to the number of records such as the number of records, the degree of coverage, and the fitness. This is because a node with a large number of records is important and considered to be highly reliable, and a node with a small number of records is considered not to have relatively high reliability and importance. Since the user of the data analysis apparatus can confirm a highly important rule according to the selection criteria as illustrated, a data analysis apparatus capable of visualizing features inherent in sensor data without being involved in unnecessary sensor data is obtained. become.
- the parameter extraction unit 80 refers to the evaluation result table 60, selects a rule with a high target level for each state, and stores it in the parameter table 90 (S81).
- FIG. 10 is a configuration example of the parameter table 90.
- the parameter table 90 associates the state 91 and the rule 92 having a high target level.
- the state 91 is “abnormal” and “normal” like the state 26.
- the state 91 may be various states including three or more states.
- FIG. 10 illustrates a case where “rule with 100% target level” is used as a selection criterion according to the state.
- the rule in which the abnormal degree 66 of the evaluation result table 60 indicates 100% when the state 91 is “abnormal” corresponds to the node 2 and the node 8, and the conditional statement 62 is the target. It stores in the rule 92 with high degree.
- the rule in which the normal medium 69 of the evaluation result table 60 indicates 100% corresponds to the node 3, the node 5, the node 6, and the node 9. It is stored in the rule 92 having a high target level.
- the selection criterion may be set so as to extract a conditional statement of a node having a large number of records (coverage and fitness) and a high target level. .
- the parameter extraction unit 80 outputs (displays) the contents of the parameter table 90 to the output device 3 (S82). Since the content of the parameter table 90 output to the output device is a rule with a high target level selected according to a predetermined selection criterion, the user of the data analysis unit is particularly focused on when the system status is abnormal. Priority can be given to actions that correspond to rules with high degrees.
- the parameter extracting unit 80 may extract important parameters or calculate the importance of parameters in addition to outputting the extracted conditional statements (parameters and their values) of each node as they are.
- the conditional statement of the node having a high normal state ratio has parameters D and A, which are the most important, and then C is important.
- the importance of each parameter may be converted into a numerical value based on the target level of the node, the number of records, and the like.
- the importance of the parameter may be extracted for each state (abnormal or normal) of the objective variable, or may be extracted for all states.
- the data analysis apparatus described it is possible to visualize the characteristics inherent in the data as rules, eliminating the complexity of the decision tree.
- 1 data analysis device
- 2 analysis target system
- 3 output device
- 10 data collection unit
- 20 collection data table
- 30 decision tree creation unit
- 40 decision tree table
- 50 evaluation value generation unit
- 60 Evaluation result table
- 80 parameter extraction unit
- 90 parameter table.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Databases & Information Systems (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Medical Informatics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
Description
Claims (10)
- 目的変数と複数の説明変数を含むレコードを有するデータを分析するデータ分析装置において、
前記レコードの目的変数と説明変数に基づいて前記説明変数の条件で規定されたノードを作成し、前記レコードを前記ノードに対応付けるノード作成部、
前記ノードに対応付けられた複数のレコードの、前記目的変数が対象値であるレコードの数の割合を評価値として生成する評価値生成部、および、
前記評価値に基づいてノードを選択し、当該選択したノードにかかる前記説明変数の条件を抽出して出力するパラメータ抽出部、を有することを特徴とするデータ分析装置。 - 請求項1において、
前記評価値生成部は、前記目的変数が取りうる値を前記対象値とし、前記ノードごとにそれぞれの前記対象値について前記評価値を生成することを特徴とするデータ分析装置。 - 請求項2において、
前記ノード抽出部は、前記対象値ごとに前記ノードを抽出することを特徴とするデータ分析装置。 - 請求項2または3において、
前記ノード抽出部は、前記ノードの評価値と、前記ノードのレコード数にかかる値に基づいて、前記ノードの選択を行い、
前記ノードのレコード数にかかる値は、前記ノードに含まれるレコード数、前記ノードに含まれる前記目的変数が対象値のレコード数、前記ノードに含まれる前記目的変数が対象値のレコード数の全レコード数に対する割合、のいずれかであることを特徴とするデータ分析装置。 - 請求項1乃至4のいずれかにおいて、
前記ノード作成部は、決定木分析により、前記ノードを作成することを特徴とするデータ分析装置。 - 目的変数と複数の説明変数を含むレコードを有するデータを分析するデータ分析装置におけるデータ分析方法であって、前記データ分析装置は、
前記レコードの目的変数と説明変数に基づいて前記説明変数の条件で規定されたノードを作成し、
前記レコードを前記ノードに対応付け、
前記ノードに対応付けられた複数のレコードの、前記目的変数が対象値であるレコードの数の割合を評価値として生成し、
前記評価値に基づいてノードを選択し、
当該選択したノードにかかる前記説明変数の条件を抽出して出力することを特徴とするデータ分析方法。 - 請求項6において、前記データ分析装置は、
前記目的変数が取りうる値を前記対象値とし、前記ノードごとにそれぞれの前記対象値について前記評価値を生成することを特徴とするデータ分析方法。 - 請求項7において、
前記データ分析装置は、前記対象値ごとに前記ノードを抽出することを特徴とするデータ分析方法。 - 請求項7または8において、前記データ分析装置は、
前記ノードの評価値と、前記ノードのレコード数にかかる値に基づいて、前記ノードの選択を行い、
前記ノードのレコード数にかかる値は、前記ノードに含まれるレコード数、前記ノードに含まれる前記目的変数が対象値のレコード数、前記ノードに含まれる前記目的変数が対象値のレコード数の全レコード数に対する割合、のいずれかであることを特徴とするデータ分析方法。 - 請求項6乃至9のいずれかにおいて、前記データ分析装置は、
決定木分析により、前記ノードを作成することを特徴とするデータ分析方法。
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/JP2015/076363 WO2017046906A1 (ja) | 2015-09-16 | 2015-09-16 | データ分析装置および分析方法 |
US15/557,542 US11244235B2 (en) | 2015-09-16 | 2015-09-16 | Data analysis device and analysis method |
JP2017540404A JP6458157B2 (ja) | 2015-09-16 | 2015-09-16 | データ分析装置および分析方法 |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/JP2015/076363 WO2017046906A1 (ja) | 2015-09-16 | 2015-09-16 | データ分析装置および分析方法 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2017046906A1 true WO2017046906A1 (ja) | 2017-03-23 |
Family
ID=58288361
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2015/076363 WO2017046906A1 (ja) | 2015-09-16 | 2015-09-16 | データ分析装置および分析方法 |
Country Status (3)
Country | Link |
---|---|
US (1) | US11244235B2 (ja) |
JP (1) | JP6458157B2 (ja) |
WO (1) | WO2017046906A1 (ja) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2018195133A (ja) * | 2017-05-18 | 2018-12-06 | 富士ゼロックス株式会社 | 情報処理装置及び情報処理プログラム |
JP2020126331A (ja) * | 2019-02-01 | 2020-08-20 | 株式会社オービック | データ分析装置、データ分析方法およびデータ分析プログラム |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP6917942B2 (ja) * | 2018-04-11 | 2021-08-11 | 株式会社日立製作所 | データ分析サーバ、データ分析システム、及びデータ分析方法 |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2000222486A (ja) * | 1999-02-02 | 2000-08-11 | Toshiba Corp | 訪問販売における営業職員活動分析システム |
JP2004220236A (ja) * | 2003-01-14 | 2004-08-05 | Nippon Telegr & Teleph Corp <Ntt> | データ分析方法、装置、プログラム及び該プログラムを記録した記録媒体 |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2006048129A (ja) | 2004-07-30 | 2006-02-16 | Toshiba Corp | データ処理装置、データ処理方法及びデータ処理プログラム |
US8250008B1 (en) * | 2009-09-22 | 2012-08-21 | Google Inc. | Decision tree refinement |
US9299035B2 (en) * | 2013-11-01 | 2016-03-29 | International Business Machines Corporation | Iterative refinement of pathways correlated with outcomes |
JP6451736B2 (ja) * | 2014-03-28 | 2019-01-16 | 日本電気株式会社 | 価格推定装置、価格推定方法、及び、価格推定プログラム |
US9973520B2 (en) * | 2014-07-15 | 2018-05-15 | Cisco Technology, Inc. | Explaining causes of network anomalies |
-
2015
- 2015-09-16 JP JP2017540404A patent/JP6458157B2/ja active Active
- 2015-09-16 WO PCT/JP2015/076363 patent/WO2017046906A1/ja active Application Filing
- 2015-09-16 US US15/557,542 patent/US11244235B2/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2000222486A (ja) * | 1999-02-02 | 2000-08-11 | Toshiba Corp | 訪問販売における営業職員活動分析システム |
JP2004220236A (ja) * | 2003-01-14 | 2004-08-05 | Nippon Telegr & Teleph Corp <Ntt> | データ分析方法、装置、プログラム及び該プログラムを記録した記録媒体 |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2018195133A (ja) * | 2017-05-18 | 2018-12-06 | 富士ゼロックス株式会社 | 情報処理装置及び情報処理プログラム |
JP2020126331A (ja) * | 2019-02-01 | 2020-08-20 | 株式会社オービック | データ分析装置、データ分析方法およびデータ分析プログラム |
JP7197391B2 (ja) | 2019-02-01 | 2022-12-27 | 株式会社オービック | データ分析装置、データ分析方法およびデータ分析プログラム |
Also Published As
Publication number | Publication date |
---|---|
JP6458157B2 (ja) | 2019-01-23 |
US20180046927A1 (en) | 2018-02-15 |
US11244235B2 (en) | 2022-02-08 |
JPWO2017046906A1 (ja) | 2017-12-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP6243080B1 (ja) | プリプロセッサおよび異常予兆診断システム | |
JP6601433B2 (ja) | イベント解析装置、イベント解析システム、イベント解析方法、イベント解析プログラム及び記録媒体 | |
US20170109668A1 (en) | Model for Linking Between Nonconsecutively Performed Steps in a Business Process | |
CN111459700A (zh) | 设备故障的诊断方法、诊断装置、诊断设备及存储介质 | |
WO2021143268A1 (zh) | 基于模糊推理理论的电力信息系统健康评估方法及系统 | |
US20170109636A1 (en) | Crowd-Based Model for Identifying Executions of a Business Process | |
US20170109667A1 (en) | Automaton-Based Identification of Executions of a Business Process | |
JP6835098B2 (ja) | 要因分析方法、要因分析装置および要因分析プログラム | |
US20170109639A1 (en) | General Model for Linking Between Nonconsecutively Performed Steps in Business Processes | |
JP2015011027A (ja) | 時系列データにおける異常を検出する方法 | |
JP2006318263A (ja) | 情報分析システム、情報分析方法及びプログラム | |
CN106164795B (zh) | 用于分类报警的优化方法 | |
CN107944005B (zh) | 一种数据展示方法及装置 | |
CN116450399A (zh) | 微服务系统故障诊断及根因定位方法 | |
US20170109638A1 (en) | Ensemble-Based Identification of Executions of a Business Process | |
Gitzel | Data Quality in Time Series Data: An Experience Report. | |
JP6458157B2 (ja) | データ分析装置および分析方法 | |
US20170109640A1 (en) | Generation of Candidate Sequences Using Crowd-Based Seeds of Commonly-Performed Steps of a Business Process | |
JP4314123B2 (ja) | アラーム解析装置、アラーム解析方法及びアラーム解析プログラム | |
US8543552B2 (en) | Detecting statistical variation from unclassified process log | |
JP2019197245A (ja) | 情報処理装置、品質関連式生成方法、および品質関連式生成プログラム | |
JP2010250864A (ja) | 情報処理装置およびプログラム | |
WO2020066124A1 (ja) | プロセス状態解析装置及びプロセス状態表示方法 | |
US20170109637A1 (en) | Crowd-Based Model for Identifying Nonconsecutive Executions of a Business Process | |
US20170109670A1 (en) | Crowd-Based Patterns for Identifying Executions of Business Processes |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 15904095 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 2017540404 Country of ref document: JP Kind code of ref document: A |
|
WWE | Wipo information: entry into national phase |
Ref document number: 15557542 Country of ref document: US |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 15904095 Country of ref document: EP Kind code of ref document: A1 |