JP5584917B2 - Data analysis system and data analysis method - Google Patents

Description

  The present invention relates to a data analysis system that analyzes data, and more particularly to a data analysis system that automatically performs analysis.

  With the development of IT (Information Technology) technology, a large amount of data such as business logs and sensor data has been collected. Companies and local governments have a need to analyze these large amounts of data to optimize their operations.

  When trying to optimize a business using data such as a business log and sensor data, first, the data is aggregated and analyzed in order to grasp the overall tendency of the data. Here, the tabulation analysis refers to, for example, calculating a frequency distribution for each item constituting the data and executing cross tabulation on a plurality of items constituting the data.

  Usually, such total analysis is performed in the initial stage of data analysis work, and 100 or more total analysis is performed. It is necessary for experts who understand the meaning of the data and statistics experts to proceed with tabulation and analysis work through trial and error.

  For example, when the monthly sales of a certain product category are aggregated and analyzed, it is assumed that the sales in January are twice that of other months. In this case, it is conceivable that the analysis is further advanced by analyzing the sales in January for each product belonging to this product category. It is also possible to cross-tabulate sales for the product category on two axes, fiscal and monthly, and analyze whether the same trend appears every year.

  Thus, the total analysis work devise a new analysis method based on the previous analysis result, and executes the devised analysis method. Therefore, in the total analysis, a trial and error process is repeated many times, and finally 100 or more total analysis results are obtained.

  Such a tabulation analysis is costly and time consuming for specialists who work. Further, when 100 or more total analyzes are executed, it becomes difficult for a person who has executed the analysis to fully understand and interpret a large number of analysis results.

  A technique is known in which items constituting data are divided into a plurality of groups, cross tabulation is performed, and a graph is generated (for example, see Patent Document 1).

  A questionnaire survey is conducted when collecting and aggregating information transmission logs. A technology that extracts the characteristics of large temporal changes from the analysis results of the questionnaire survey and determines the scope and items of the questionnaire based on the extracted characteristics. It is known (see, for example, Patent Document 2).

  A technique is known in which a visual plot suitable for items constituting data is automatically generated (see, for example, Patent Document 3).

JP 2007-11468 A JP 2006-18357 A Special table 2009-508210

  In the techniques described in Patent Documents 1 to 3, since the total analysis is not automatically performed, the human cost and the time cost of the total analysis cannot be solved.

  Thus, a technique is desired in which a computer automatically performs a total analysis that is advanced by trial and error. As a result, it is expected that the human cost in the total analysis is reduced and the time required for the total analysis work is also shortened. Also, by presenting to the user the criteria used when the computer determines whether to analyze the analysis results further, the user can interpret the analysis results using the criteria and rank the analysis results. The criteria presented to the user helps to understand a large number of analysis results.

  However, there are the following technical problems for the computer to automatically perform the aggregation analysis.

  First, if the computer analyzes the analysis results at random, there is a problem that the number of analyzes diverges and computer resources are wasted. In particular, if a large number of low-value analysis results are preferentially executed, it takes a long time and a valuable analysis result cannot be obtained.

  Second, in order for the computer to further analyze the analysis results, an algorithm is required to examine the analysis results in detail and determine which part of the analysis results to deepen.

  Third, since it is necessary for a computer to perform a large number of aggregate analyzes on a large amount of data, an algorithm for efficiently performing the analysis is necessary. In particular, when a computer performs analysis processing on a large amount of data, file input / output due to the output of many low-value analysis results becomes a bottleneck, and the execution speed of analysis processing decreases. It is necessary to improve efficiency. Therefore, the computer executes as much analysis processing as possible, evaluates the analysis result after executing the analysis processing, and preferentially presents the analysis result with high evaluation to the user. Thereby, only the analysis result with high value can be presented to the user, and the accuracy of the analysis result presented to the user can be improved. For this purpose, the computer needs to execute as many analysis processes as possible, and it is an important issue to improve the execution speed of the analysis processes.

  The present invention has been made in view of the above, and an object of the present invention is to provide a data analysis system that automatically performs tabulation analysis.

  A typical example of the present invention is a data analysis system that includes a processor that executes a program and a storage area that stores a program executed by the processor, and analyzes the data. An analysis execution unit that executes a process and stores an analysis result of the analysis process in the storage area; a plurality of inference rule units that evaluate the analysis result and generate a new analysis process from the analysis result; An inference rule control unit for controlling the inference rule unit, and an analysis processing management unit for managing an unprocessed analysis process of the analysis execution unit, wherein the inference rule unit evaluates the analysis result And an analysis processing generation unit that generates the new analysis processing from the analysis result, and the analysis result evaluation unit reads the analysis result stored in the storage area. The analysis result is evaluated by calculating a feature value indicating the characteristic of the analysis result, the analysis result and the feature value are associated with each other and stored in the storage area, and the inference rule control unit The analysis result stored in the storage area is selected as a candidate for the generation source of the new analysis process, and the analysis process generation unit is selected by the inference rule control unit. A new analysis process is generated from the analysis result, and the analysis process management unit causes the analysis execution unit to execute the analysis process generated by the analysis process generation unit.

  ADVANTAGE OF THE INVENTION According to this invention, the data analysis system which performs a total analysis automatically can be provided.

It is explanatory drawing of the data analysis system of embodiment of this invention. It is a functional block diagram of a data analysis system of an embodiment of the present invention. It is explanatory drawing of an example of the analysis object data of embodiment of this invention. It is explanatory drawing of an example of the analysis object data of embodiment of this invention. It is explanatory drawing of an example of the analysis object data of embodiment of this invention. It is explanatory drawing of an example of the analysis script of embodiment of this invention. It is explanatory drawing of an example of the analysis script of embodiment of this invention. It is explanatory drawing of an example of the analysis result of embodiment of this invention. It is explanatory drawing of an example of the graph of the analysis result of embodiment of this invention. It is explanatory drawing of the table schema of the analysis script table of embodiment of this invention. It is explanatory drawing of the table schema of the evaluation value table of embodiment of this invention. It is a sequence diagram which shows the data flow of the analysis script production | generation process of embodiment of this invention. It is a flowchart of the analysis script production | generation process of embodiment of this invention. It is a sequence diagram which shows the data flow of the analysis execution process of embodiment of this invention. It is a flowchart of the analysis execution process of embodiment of this invention. It is explanatory drawing of the merge of the analysis script of embodiment of this invention. It is explanatory drawing of the analysis result list display screen which the analysis result display part of embodiment of this invention displays. It is explanatory drawing of the analysis result display screen which the analysis result display part of embodiment of this invention displays. It is explanatory drawing of the sort parameter setting screen which the analysis result display part of embodiment of this invention displays. It is explanatory drawing of the analysis script input screen which the analysis user interface part of embodiment of this invention displays. 1 is a block diagram of a data analysis system 100 configured by a plurality of computers according to an embodiment of this invention.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. In this specification, functions realized by a program in a processing unit such as a computer may be expressed as “processing”, “unit”, and “means”.

  The present embodiment relates to a data analysis system 100 (see FIG. 1) that automatically analyzes data.

  FIG. 1 is an explanatory diagram of a data analysis system 100 according to the embodiment of this invention.

  In FIG. 1, the data analysis system 100 is described as a general-purpose computer (computer). However, the present invention is not limited to this, and the data analysis system 100 may be configured by a plurality of general-purpose computers.

  The data analysis system 100 includes an input device 101, a network device 102, a central processing unit (processor or central processing unit: CPU) 103, a main storage unit 104, an auxiliary storage unit 105, and a display device 107.

  The input device 101, network device 102, CPU 103, main storage unit 104, auxiliary storage unit 105, and display device 107 are connected via a bus 106.

  The input device 101 is an input unit that receives input from a user, and is, for example, a keyboard and a mouse. The network device 102 is a network interface unit for connecting to a network such as the Internet.

  The CPU 103 executes a program stored in the main storage unit 104. The main storage unit 104 is a storage area accessible by the CPU 103 and stores a program executed by the CPU 103. The main storage unit 104 is, for example, a memory. The auxiliary storage unit 105 is a storage area that cannot be directly accessed by the CPU 103, and stores various data including programs executed by the CPU 103. The auxiliary storage unit 105 is, for example, an HDD (Hard Disk Drive).

  The input device 101 and the display device 107 constitute a user interface unit.

  FIG. 2 is a functional block diagram of the data analysis system 100 according to the embodiment of this invention.

  The data analysis system 100 includes an analysis result database 200, an inference rule control unit 201, inference rule units 202A and 202B (hereinafter collectively referred to as an inference rule unit 202), a job control unit 208, an analysis execution unit 209, and an analysis result display unit. 213, an analysis user interface unit 214, and a graph generation unit 215.

  The inference rule unit 202A includes an analysis result evaluation unit 203A and an analysis script generation unit 204A. Similarly, the inference rule unit 202B includes an analysis result evaluation unit 203B and an analysis script generation unit 204B. The analysis result evaluation units 203A and 203B are collectively referred to as an analysis result evaluation unit 203, and the analysis script generation units 204A and 204B are collectively referred to as an analysis script generation unit 204.

  The analysis execution unit 209 includes an analysis script interpretation unit 210, an analysis target data management unit 211, and analysis target data 212.

  Part or all of the inference rule control unit 201, the inference rule unit 202, the job control unit 208, the analysis execution unit 209, the analysis result display unit 213, the analysis user interface unit 214, and the graph generation unit 215 are shown in FIG. This is realized by the CPU 103 executing the program.

  The analysis result display unit 213 and the analysis user interface unit 214 include the hardware configurations of the input device 101 and the display device 107 shown in FIG.

  The analysis result database 200 and the analysis target data 212 are stored in a storage area such as the main storage unit 104 or the auxiliary storage unit 105 shown in FIG.

  Each unit, the analysis result database 200, and the analysis target data 212 illustrated in FIG. 2 may be mounted on the same computer or may be distributed and mounted on a plurality of computers.

  The analysis result display unit 213 and the analysis user interface unit 214 may include a configuration connected to the input device 101 and the display device 107 in addition to the input device 101 and the display device 107.

  Hereinafter, each component of the data analysis system 100 will be described.

  The analysis result database 200 includes an analysis script table 1000 (see FIG. 10) in which an analysis result that is an execution result of an analysis script describing an analysis process is registered, and an evaluation value table 1100 in which an evaluation value of the analysis result is registered. (See FIG. 11). Here, the analysis script refers to an execution instruction for analysis processing described in DSL (Domain Specific Language) or a programming language. The evaluation value is also referred to as a feature amount indicating the characteristic of the analysis result.

  Details of the analysis script table 1000 will be described with reference to FIG. Details of the evaluation value table 1100 will be described with reference to FIG.

  The inference rule control unit 201 controls the inference rule unit 202 and operates in cooperation with a plurality of inference rule units.

  Specifically, the inference rule control unit 201 acquires the analysis result stored in the analysis script table 1000 of the analysis result database 200, and analyzes the characteristic amount (evaluation value) indicating the characteristic of the acquired analysis result as the analysis result evaluation unit. The evaluation value calculated by the analysis result evaluation unit 203 is stored in the evaluation value table 1100. Further, the inference rule control unit 201 acquires an analysis result whose evaluation value satisfies a predetermined condition from the analysis result stored in the evaluation value table 1100, and sends a new analysis script to the analysis script generation unit 204 from the acquired analysis result. The generated new analysis script is stored in the analysis script table 1000.

  The inference rule unit 202 executes an analysis result evaluation process and a new analysis script generation process based on the inference rules. The inference rule is a standard for the evaluation process and the generation process, such as frequency tabulation and cross tabulation.

  The inference rule unit 202 whose inference rule is frequency aggregation generates an analysis script for executing frequency aggregation, and evaluates the analysis result so that the evaluation value of the analysis result having a prominent value is high.

  On the other hand, the inference rule unit 202 whose inference rule is cross tabulation generates an analysis script for executing cross tabulation and evaluates the analysis result so that the evaluation value of the analysis result having a uniform value becomes high.

  Note that different inference rules are set in the different inference rule sections 202. As a result, the data analysis system 100 includes a plurality of inference rule units 202, thereby enabling various analyzes.

  The job control unit 208 acquires an unexecuted analysis script from the analysis script table 1000 of the analysis result database 200 and places the acquired analysis script in the queue of the analysis execution unit 209 as an analysis job. The job control unit 208 monitors the progress of the analysis script of the analysis execution unit 209, and the analysis result, which is the result of the analysis execution unit 209 executing the analysis script, is stored in the analysis script table 1000 and the evaluation value table 1100 of the analysis result database 200. Store. From the above, it can be said that the job control unit 208 functions as an analysis processing management unit that manages the analysis processing of the analysis execution unit 209.

  The analysis execution unit 209 interprets the analysis script and executes analysis processing on the analysis target data 212. The analysis execution unit 209 includes an analysis script interpretation unit 210, an analysis target data management unit 211, and analysis target data 212.

  The analysis script interpretation unit 210 interprets the analysis script and converts the analysis script into a format that can be executed by the CPU 103.

  The analysis target data management unit 211 manages the analysis target data 212 and executes processing necessary for the analysis target data 212.

  The analysis execution unit 209 corresponds to, for example, a data processing device and an RDBMS (Relational Database Management System) described in Japanese Patent Application Laid-Open No. 2011-13758.

  The analysis execution unit 209 may be a distributed calculation system configured on a large number of computers when executing analysis processing on a large amount of data.

  The analysis result display unit 213 displays the analysis result stored in the analysis result database 200 on the display device 107. The analysis result display unit 213 switches display contents displayed on the display device 107 and rearranges the analysis results displayed on the display device 107 based on an input from the user.

  The analysis user interface unit 214 receives an input of an analysis script created by the user via the analysis script input screen 2000 shown in FIG. 20 and causes the analysis execution unit 209 to execute the received analysis script. The analysis user interface unit 214 displays the analysis result of the analysis script input by the user on the display device 107. Further, the analysis user interface unit 214 stores the analysis result obtained by performing the analysis process by the analysis execution unit 209 in the analysis script table 1000 of the analysis result database 200 based on an instruction from the user.

  From the above, it can be said that the analysis user interface unit 214 functions as an analysis script input receiving unit that receives an input of an analysis script created by a user.

  The inference rule control unit 201 can also cause the inference rule unit 202 to generate a new analysis script based on the analysis result of the analysis script created by the user received by the analysis user interface unit 214. As a result, the user can correct the analysis policy of the aggregate analysis automatically advanced by the data analysis system 100 via the analysis user interface unit 214.

  The graph generation unit 215 generates graph data such as a bar graph and a scatter diagram based on the analysis result, and displays the generated graph data on the display device 107 via the analysis result display unit 213, the analysis user interface unit 214, and the like.

  Next, an example of the analysis target data 212 will be described with reference to FIGS.

  FIG. 3 is an explanatory diagram illustrating an example of the analysis target data 212 according to the embodiment of this invention.

  In FIG. 3, the analysis target data 212 is described using data 300 with an XML tag, but it is obvious that the analysis target data 212 may be data other than the data 300 illustrated in FIG. 3.

  The analysis target data 212 shown in FIG. 3 represents a log of the acceleration sensor attached to the body by the user. In this log, “user ID”, “behavior”, “date”, “time”, “acceleration in x direction”, “acceleration in y direction”, and “acceleration in z direction” are defined by XML tags, These values are stored.

  In FIG. 3, only the analysis target data 212 for one user is shown, but there is actually analysis target data 212 for a plurality of users.

  FIG. 4 is an explanatory diagram illustrating an example of the analysis target data 212 according to the embodiment of this invention.

  The analysis target data 212 may be data other than the data 300 with the XML tag shown in FIG. 3, and FIG. 4 shows the data 400 excluding the XML tag from the data 300 with the XML tag shown in FIG. .

  FIG. 5 is an explanatory diagram illustrating an example of the analysis target data 212 according to the embodiment of this invention.

  FIG. 5 shows data 500 obtained by converting the data 300 shown in FIG. 3 or the data 400 shown in FIG. 4 into a table format.

  Next, an example of an analysis script will be described with reference to FIGS.

  FIG. 6 is an explanatory diagram illustrating an example of the analysis script according to the embodiment of this invention.

  An example of the analysis script shown in FIG. 6 is a frequency tabulation analysis script 600 that calculates a frequency distribution of absolute values of acceleration from the analysis target data 212 that is a log of the acceleration sensor shown in FIG. Here, the absolute value of acceleration is the square root of the sum of squared values obtained by squaring the acceleration in the x direction, the acceleration in the y direction, and the acceleration in the z direction.

  “Logs: histogram (0, 20, 1)” in FIG. 6 means that the frequency distribution of the “logs” tag hierarchy of the data 300 shown in FIG. 3 is calculated, and the argument is a value from 0 to 20 Is counted in increments of 1. Further, “target: sqrt (x * x + y * y + z * z)” in FIG. 6 means that a value obtained by rounding off the absolute value of acceleration is targeted.

  FIG. 7 is an explanatory diagram illustrating an example of an analysis script according to the embodiment of this invention.

  FIG. 7 shows an analysis script 700 in which the analysis script 600 shown in FIG. 6 is described in the SQL language.

  Note that the analysis script of the present embodiment is not limited to the analysis script shown in FIGS. 6 and 7 and may be another analysis script.

  FIG. 8 is an explanatory diagram illustrating an example of an analysis result according to the embodiment of this invention.

  An example of the analysis result shown in FIG. 8 is analysis result data 800 obtained by executing the analysis script 600 shown in FIG. 6 on the data 300 (analysis target data 212) shown in FIG.

  The first line shown in FIG. 8 means that there are 21240 logs with a value of 0 obtained by rounding off the absolute value of acceleration. The second line shown in FIG. 8 means that 1246 logs have a value of 1 obtained by rounding off the absolute value of acceleration. Since the third and subsequent lines shown in FIG. 8 are the same as the first and second lines, the description thereof is omitted.

  FIG. 9 is an explanatory diagram illustrating an example of a graph of an analysis result according to the embodiment of this invention.

  An example of the analysis result graph illustrated in FIG. 9 is a graph 900 of the analysis result data 800 illustrated in FIG.

  The horizontal axis is the value obtained by rounding off the absolute value of acceleration, and the vertical axis is the number of logs of analysis results.

  Next, the analysis result database 200 will be described with reference to FIGS.

  As described with reference to FIG. 2, the analysis result database 200 includes an analysis script table 1000 that stores analysis scripts, analysis results, and the like, and an evaluation value table 1100 that stores evaluation values of analysis results.

  The analysis result database 200 is managed by a known technique such as an RDBMS (Relational Database Management System).

  FIG. 10 is an explanatory diagram of a table schema of the analysis script table 1000 according to the embodiment of this invention.

  The analysis script table 1000 includes analysis_id 1001, data_id 1002, created_at 1003, updated_at 1004, parent_analysis_ids 1005, rule_id 1006, title 1007, script 1008, expected_fitness 1009, and result 1010.

  In the analysis_id 1001, a unique identifier (hereinafter referred to as an analysis ID) of a record in the analysis script table 1000 is registered. Registered in data_id 1002 is an identifier indicating the type of analysis target data 212 to be analyzed.

  In created_at 1003, the date and time when the record was added is registered. In updated_at 1004, the date and time when the record was updated is registered.

  In parent_analysis_ids 1005, an analysis ID of an analysis result that is a creation source of the analysis script registered in script 1008 is registered.

  In rule_id 1006, a unique identifier (rule ID) of the analysis script generation unit 204 that created the analysis script registered in script 1008 is registered. In FIG. 10, cross and scale are registered in rule_id 1006, where cross indicates the rule ID of the cross tabulation inference rule unit 202, and scale indicates the rule ID of the frequency distribution inference rule unit 202.

  In title 1007, the title of the aggregate analysis by the analysis script registered in script 1008 is registered. In the script 1008, the analysis script generated by the analysis script generation unit 204 is registered. In expected_fitness 1009, a predicted value of the evaluation value of the analysis script registered in the record is registered. Details of the calculation method of the predicted value will be described with reference to FIG. In result 1010, the analysis result of the analysis script registered in the script 1008 of the record is registered.

  FIG. 11 is an explanatory diagram of a table schema of the evaluation value table 1100 according to the embodiment of this invention.

  The evaluation value table 1100 includes fitness_id 1101, analysis_id 1102, created_at 1103, updated_at 1104, rule_id 1105, measure 1106, and value 1107.

  In fitness_id 1101, a unique identifier of a record in the evaluation value table 1100 is registered. In analysis_id 1102, an analysis ID of the evaluated analysis result is registered. The evaluation value table 1100 is associated with the analysis script table 1000 by the analysis ID. Since one analysis result is evaluated by the analysis result evaluation unit 203 of the plurality of inference rule units 202, one record in the analysis script table 1000 corresponds to a plurality of records in the evaluation value table 1100, and these records are Corresponding by analysis ID.

  In created_at 1103, the date and time when the record of the evaluation value table 1100 was added is registered.

  In updated_at 1104, the date and time when the record of the evaluation value table 1100 was updated is registered.

  In rule_id 1105, an inference rule set in the analysis result evaluation unit 203 that calculates the evaluation value is registered. In the measure 1106, a unique identifier of the evaluation criterion for which the analysis result evaluation unit 203 has calculated the evaluation value is registered. An evaluation value is registered in value 1107.

  Next, the analysis script creation process will be described with reference to FIGS.

  FIG. 12 is a sequence diagram illustrating a data flow of the analysis script generation process according to the embodiment of this invention.

  Data used in the analysis script generation process is exchanged among the inference rule control unit 201, the analysis result database 200, and the inference rule unit 202.

  Each inference rule unit 202 is set with a unique identifier (rule ID). To simplify the explanation, in FIG. 12, the inference rule unit 202A is set with a rule ID “0”, and the inference rule unit 202B. It is assumed that the rule ID “1” is set in.

  The inference rule part 202A to which the rule ID “0” is set first executes the analysis script generation process, and then the inference rule part 202B to which the rule ID “1” is set executes the analysis script generation process.

  In FIG. 12, two inference rule units 202 are illustrated. However, even if there are three or more inference rule units 202, after one inference rule unit 202 executes the analysis script processing, the next inference rule is displayed. The unit 202 executes analysis script generation processing.

  First, the inference rule control unit 201 registers an analysis result that has not yet been evaluated by the inference rule unit 202A (rule_id = 0) that executes the analysis script generation process, from the record of the analysis script table 1000 of the analysis result database 200. A record is acquired (1201).

  Next, the inference rule control unit 201 passes the record of the analysis script table 1000 acquired in the process of step 1201 to the inference rule unit 202A. The inference rule unit 202A calculates the evaluation value of the record in accordance with the evaluation criterion set in the inference rule unit 202A, and passes the calculated evaluation value and the evaluation criterion used for calculating the evaluation value to the inference rule control unit 201. (1202).

  Next, the inference rule control unit 201 registers the evaluation value passed from the inference rule unit 202A in the evaluation value table 1100 of the analysis result database 200 (1203).

  Next, the inference rule control unit 201 refers to the evaluation value table 1100 of the analysis result database 200, and acquires a predetermined number of records that satisfy a predetermined condition as analysis script generation source candidates (1204). Details of the predetermined condition will be described in the processing of step 1305 in FIG.

  Next, the inference rule control unit 201 causes the inference rule unit 202A that executes the analysis script generation process to generate an analysis script based on the generation source candidate acquired in the process of step 1204 (1205).

  Next, the inference rule control unit 201 registers the analysis script generated by the inference rule unit 202A in the analysis script table 1000 of the analysis result database 200 (1206).

  Subsequently, since the inference rule control unit 201 has executed the analysis script generation process on the inference rule unit 202A, the inference rule control unit 201 executes the analysis script generation process on the inference rule unit 202B (rule_id = 1).

  Since the analysis script generation process of the inference rule unit 202B is the same as the analysis script generation process of the inference rule unit 202A, the same reference numerals are given and description thereof is omitted.

  FIG. 13 is a flowchart of analysis script generation processing according to the embodiment of this invention.

  The analysis script generation process is executed by the CPU 103 that executes a program that implements the inference rule control unit 201 and the inference rule unit 202.

  The inference rule control unit 201 sets a variable (rule_id) indicating a rule ID set in the inference rule unit 202 that executes the analysis script generation process to 0 (1301).

  Next, the inference rule control unit 201 has not yet evaluated the inference rule unit 202 that executes the analysis script generation process from the record of the analysis script table 1000 of the analysis result database, as described in the process of step 1201 shown in FIG. Records in which analysis results that have not been registered are registered are acquired as evaluation target analysis results (1302).

  The processing in step 1302 will be described in detail.

  The inference rule control unit 201 acquires a record (analyzed record) in which records other than NULL are registered in the result 1010 among the records of the analysis script table 1000.

  Then, the inference rule control unit 201 acquires a record that has not been evaluated by any inference rule unit 202 among the acquired analyzed records as an evaluation target analysis result. Specifically, the inference rule control unit 201 acquires, as an evaluation target analysis result, an analyzed record in which the analysis ID registered in the analysis_id 1001 of the analyzed record is not registered in the analysis value id 1102 in the evaluation value table 1100.

  Further, the inference rule control unit 201 uses an analyzed record that has not been evaluated yet by the inference rule unit 202 specified by the current variable (rule_id) as an evaluation target analysis result, even if the analyzed record has already been evaluated. get. Specifically, the inference rule control unit 201 is configured so that the analysis ID registered in the analysis_id 1001 of the analyzed record has a record registered in the analysis_id 1102 in the evaluation value table 1100. When the rule ID registered in the rule_id 1105 of the record is different from the rule ID set in the processing of Step 1301, the record of the analysis script table 1000 in which the analysis ID is registered in the analysis_id 1001 is acquired as the evaluation target analysis result.

  Next, the inference rule control unit 201 causes the analysis result evaluation unit 203 of the inference rule unit 202 specified by the variable (rule_id) to evaluate the evaluation object analysis result acquired in the process of step 1302 (1303).

  Specifically, the inference rule control unit 201 analyzes the analysis ID registered in the analysis_id 1001 included in the record of the analysis script table 1000 that is the evaluation target analysis result, the analysis script registered in the script 1008 included in the record, and the The analysis result registered in the result 1010 included in the record is passed to the analysis result evaluation unit 203.

  The analysis result evaluation unit 203 calculates an evaluation value based on the analysis result passed from the inference rule control unit 201 and the evaluation standard set in the analysis result evaluation unit 203, and the calculated evaluation value and the evaluation standard are controlled by the inference rule control. To the unit 201.

  Here, the analysis result evaluation method in the analysis result evaluation unit 203 will be briefly described. Details of the evaluation method will be described later.

  The analysis result evaluation unit 203 is paired with the analysis script generation unit 204, and the analysis processing of the analysis script generated by the analysis script corresponds to the analysis result evaluation method by the analysis result evaluation unit 203.

  For example, when the analysis script generation unit 204 generates an analysis script for performing frequency aggregation, the analysis result evaluation unit 203 corresponding to the analysis script generation unit 204 evaluates the degree of protrusion of the analysis result. Specifically, the analysis result evaluation unit 203 calculates, as an evaluation value, a value obtained by subtracting an average value of other analysis result values excluding the maximum value from the maximum value of the analysis result.

  When the analysis script generation unit 204 generates an analysis script for executing cross tabulation, the analysis result evaluation unit 203 corresponding to the analysis script generation unit 204 evaluates the degree of uniformity of the analysis result. Specifically, the analysis result evaluation unit 203 calculates the absolute value of the difference between each analysis result and the average value of the analysis results as an evaluation value.

  Next, the inference rule control unit 201 registers the evaluation value passed from the analysis result evaluation unit 203 in the evaluation value table 1100 of the analysis result database 200 (1304).

  Specifically, the inference rule control unit 201 registers the analysis ID corresponding to the evaluation value passed from the analysis result evaluation unit 203 in analysis_id 1102, registers the rule ID set in the variable (rule_id) in rule_id 1105, The evaluation criterion passed from the analysis result evaluation unit 203 is registered in the measure 1106. The inference rule control unit 201 generates a unique identifier in the evaluation value table 1100 and registers the generated identifier in fitness_id 1101. The inference rule control unit 201 registers the current date and time in created_at 1103 and updated_at 1104.

  Next, the inference rule control unit 201 acquires a predetermined number of records satisfying a predetermined condition from the evaluation value table 1100 of the analysis result database 200 as analysis script generation source candidates (1305).

  The predetermined condition will be described.

  The inference rule control unit 201 refers to the evaluation value table 1100, and determines a predetermined number in descending order of evaluation values evaluated by the analysis result evaluation unit 203 of the inference rule unit 202 specified by the rule ID set in the variable (rule_id). Get the record (evaluation value record). Specifically, the inference rule control unit 201 has a record in which the rule ID registered in the rule_id 1105 of the evaluation value table 1100 matches the rule ID set in the variable (rule_id), and the evaluation value registered in the value 1107 is large. A predetermined number (for example, 20) is acquired in order.

  Then, the inference rule control unit 201 refers to the analysis script table 1000, and the inference rule unit 202 in which the analysis result that is the evaluation target of the acquired evaluation value record is specified by the rule ID set in the variable (rule_id). A record that is not the analysis result of the analysis script generated by the analysis script generation unit 204 is acquired as a generation source candidate. Specifically, the inference rule control unit 201 refers to the record of the analysis script table 1000 in which the analysis ID registered in the analysis_id 1102 of the evaluation value record matches the analysis ID registered in the analysis_id 1001 of the analysis script table 1000, and rule_id 1006 It is determined whether or not the rule ID registered in (1) matches the rule ID set in the variable (rule_id). When the rule ID registered in the rule_id 1006 matches the rule ID set in the variable (rule_id), the inference rule control unit 201 deletes the record in which the analysis ID is registered in the analysis_id 1102 from the evaluation value record and remains An evaluation value record is acquired as a generation source candidate.

  Thus, the analysis script generation unit 204 of the inference rule unit 202 specified by the rule ID set in the variable (rule_id) generates a new analysis script. By the above processing, the analysis script generation unit 204 is generated. It is possible to reduce useless processing in which the same analysis script generation unit 204 generates the same analysis script based on the analysis result of the analysis script generated by the.

  In the process of step 1205, the inference rule control unit 201 may add a condition of an analysis result that has an evaluation value equal to or greater than a predetermined value.

  Next, the inference rule control unit 201 passes the generation source candidate to the analysis script generation unit 204 of the inference rule unit 202 specified by the rule ID set in the variable (rule_id). The analysis script generation unit 204 newly generates a predetermined number of analysis scripts based on the generation source candidates passed from the inference rule control unit 201, generates the title of the generated analysis script, and the predicted value of the generated analysis script Is calculated (1306).

  The generation candidate that the inference rule control unit 201 passes to the analysis script generation unit 204 includes the information registered in the analysis_id 1102, created_at 1103, updated_at 1104, rule_id 1105, measure 1106, and value 1107 of the evaluation value table 1100, and the analysis ID registered in the analysis_id 1102 The analysis script registered in the script 1008 of the record of the analysis script table 1000 corresponding to is included.

  An example of an analysis script generated by the analysis script generation unit 204 will be described. The analysis script generation unit 204 of a certain rule ID (scale in FIG. 10) generates an analysis script that aggregates the frequency distribution by reducing the step size of the frequency distribution. Further, the analysis script generation unit 204 of another certain rule ID (cross in FIG. 10) generates an analysis script that adds the time to a new analysis result to a new axis and performs cross tabulation on two axes.

  As an example of the calculation method of the predicted value of the analysis script generation unit 204, the analysis script generation unit 204 calculates the average value of the evaluation values of the analysis results of the generation source of the newly generated analysis script, and calculates the calculated average value May be used as a predicted value. As another example, the analysis script generation unit 204 may use the maximum evaluation value of the analysis result of the generation source of the newly generated analysis script as the predicted value.

  Note that the number of analysis scripts generated by the processing in step 1306 is calculated by Equation 1.

Pi-Wi × Qi (Formula 1)
Pi, Wi, and Qi will be described.

  Pi is the number of analysis scripts generated by the analysis script generation unit 204 when there is no job execution waiting for the analysis script generated by the analysis script generation unit 204 of a certain rule ID in the analysis execution unit 209. .

  Wi is a weighting value that can be set by the user for each rule ID.

  Qi is the number of analysis scripts waiting for job execution among the analysis scripts generated by the analysis script generation unit 204 of a certain rule ID.

  As described above, when the number of analysis scripts waiting for job execution is large, the number of analysis scripts generated by the analysis script generation unit 204 can be reduced, so that the processing load can be reduced. On the other hand, when the number of analysis scripts waiting for job execution is small, the number of analysis scripts generated by the analysis script generation unit 204 can be increased, so that more detailed aggregate analysis can be performed.

  Furthermore, since Wi can be set for each rule ID by the user, the number of analysis scripts generated by each analysis script generation unit 204 is determined in consideration of the processing load of the analysis script generated by each analysis script generation unit 204. Can be set.

  In the processing of step 1306, the analysis script generation unit 204 passes the newly generated analysis script, the title of the analysis script, and the predicted value of the analysis script to the inference rule control unit 201.

  The inference rule control unit 201 registers the analysis script, the analysis script title, and the predicted value passed from the analysis script generation unit 204 in the analysis script table 1000 of the analysis result database 200 (1307).

  Specifically, the inference rule control unit 201 registers, in data_id 1002, an identifier indicating the type of the analysis target data 212 that is the target of analysis processing of the newly generated analysis script, and generates a newly generated analysis script. The analysis ID of the original analysis result is registered in parent_analysis_ids 1005, the title of the analysis script passed from the analysis script generation unit 204 is registered in title 1007, and the analysis script passed from the analysis script generation unit 204 is registered in script 1008. The predicted value passed from the analysis script generation unit 204 is registered in the expected_fitness 1009. The inference rule control unit 201 generates a unique analysis ID in the analysis script table 1000, registers the generated analysis ID in analysis_id 1001, and registers the current date and time in created_at 1003 and updated_at 1004. Further, since the inference rule control unit 201 has not yet calculated the analysis result of the newly generated analysis script, the inference rule control unit 201 registers a NULL value in the result 1010. When a newly generated analysis script is executed by the analysis execution unit 209, the job control unit 208 registers the analysis result of the analysis script in the result 1010.

  Next, the inference rule control unit 201 newly sets a rule ID obtained by incrementing the rule ID set in the variable (rule_id) in the variable (rule_id) (1308).

  Then, the inference rule control unit 201 determines whether or not the variable (rule_id) set in step 1308 is equal to or less than N, which is the number of all inference rule units 202 included in the data analysis system 100 (1309). ).

  If it is determined in step 1309 that the variable (rule_id) is N or less, the process returns to step 1302 and steps 1302 to 1308 are performed for the inference rule part 202 of the rule ID set in the variable (rule_id). Execute the process.

  On the other hand, if it is determined in step 1309 that the variable variable (rule_id) is larger than N, the processing is completed for all the inference rule units 202, so that the inference rule control unit 201 sleeps for a certain period of time. After (1310), the processing is executed again from the processing in step 1301. Note that the analysis script generation processing is repeatedly executed in an infinite loop unless there is a forced termination from the user. In the analysis script generation process, an analysis script for executing the deep analysis is generated based on the previous analysis result, so that a new analysis script continues to be generated each time the analysis script generation process is repeated.

  Hereinafter, an example of evaluation value calculation and analysis script generation by the inference rule unit 202 will be described in detail.

  As an example of evaluation value calculation and analysis script generation by the inference rule unit 202, frequency subdivision can be considered.

  For example, in the graph of the analysis result shown in FIG. 9, 200,000 or more records with an absolute value of acceleration “0” protrude from other records. For this reason, as the next total analysis, it is conceivable that the frequency of the data in which the absolute value of the acceleration is in the vicinity of “0” is counted in 0.1 increments and graphed. In such a rule, the analysis result evaluation unit 203 evaluates the analysis result using a scale for measuring the degree of protrusion of the graph. For example, the analysis result evaluation unit 203 can evaluate the analysis result using a value obtained by subtracting the average of other frequency values excluding the maximum value from the maximum value of the frequency as an evaluation value. The analysis script generation unit 204 can generate a new analysis script in which the argument of “histogram (0, 20, 1)” shown in FIG. 6 is changed to “histogram (0, 1, 0.1)”. The analysis script generation unit 204 holds a model character string “frequency distribution of Z increments from X to Y of A” and converts “A” of the model into “sqrt (x * x + y * y + z *”. The title is generated by replacing “X” of the template with “0”, “Y” of the template with “1”, and “Z” of the template with “0.1” in “z)”.

  As another example of evaluation value calculation and analysis script generation by the inference rule unit 202, cross tabulation can be considered.

  For example, in the graph shown in FIG. 9, the absolute values “1” to “8” of acceleration are distributions close to a uniform distribution. For this reason, characteristics can appear in the frequency distribution of the absolute value of acceleration by adding other axes and performing cross tabulation on the two axes of the absolute value axis of acceleration and another axis (for example, the time axis). There is sex. For this reason, in the next tabulation analysis, it is conceivable to perform cross tabulation with the time as the horizontal axis and the absolute value of acceleration as the vertical axis. It should be noted that there may be two or more analysis results from which the cross tabulation analysis script is generated.

  When the cross tabulation is newly generated by the analysis script generation unit 204, the analysis result evaluation unit 203 evaluates the analysis result using a scale that measures the degree of uniformity of the graph. The evaluation value indicating the scale for measuring the degree of uniformity of the graph can be calculated by, for example, summing the absolute values of the differences between the frequencies and the average value of the frequencies.

  Further, the analysis script generation unit 204 that generates a cross tabulation analysis script changes “logs: histogram (0, 20, 1)” shown in FIG. 6 to “logs: histogram (1, 9, 1), histogram (0, 24, 1) '' and `` target: sqrt (x * x + y * y + z * z) '' to `` target: sqrt (x * x + y * y + z * z), hour (time) ”Can be used to generate a new script. Here, hour (time) is a function for acquiring time information from the time tag. Further, in the DSL shown in FIG. 6, it is assumed that cross tabulation can be performed by separating with a comma. With respect to the title, the analysis script generation unit 204 stores a character string “A and time cross tabulation” and replaces A with “sqrt (x * x + y * y + z * z)”. To generate a title.

  The inference rule unit 202 of the present invention is not limited to the rules in the above example, and various rules can be considered. By incorporating various inference rules into the present invention, useful aggregation analysis can be performed automatically.

  FIG. 14 is a sequence diagram illustrating a data flow of the analysis execution process according to the embodiment of this invention.

  Data used in the analysis execution process is exchanged among the job control unit 208, the analysis result database 200, and the analysis execution unit 209.

  First, the job control unit 208 acquires a record of an analysis script waiting for job execution from the analysis script table 1000 of the analysis result database 200 (1401).

  Next, the job control unit 208 passes the analysis script specified by the record acquired in step 1401 to the analysis execution unit 209 and passes an analysis process execution command to the analysis execution unit 209 (1402). When the analysis execution command is passed from the job control unit 208, the analysis execution unit 209 executes the analysis script passed from the job control unit 208.

  When the job control unit 208 passes the analysis execution instruction to the analysis execution unit 209 in the process of step 1402, the job control unit 208 shifts to a standby state and periodically inquires the analysis execution unit 209 whether or not the execution of the analysis script is completed. (1403, 1404).

  When the analysis execution unit 209 completes the execution of the analysis script, the job control unit 208 acquires the analysis result from the analysis execution unit 209 (1405). Then, the job control unit 208 stores the analysis result acquired in step 1405 in the analysis script table 1000 (1406).

  FIG. 15 is a flowchart of analysis execution processing according to the embodiment of this invention.

  The analysis execution process is executed by the CPU 103 that executes a program that implements the job control unit 208 and the analysis execution unit 209.

  First, the job control unit 208 refers to the analysis script table 1000 of the analysis result database 200, and acquires analysis scripts waiting for job execution in descending order of predicted value (expected_fitness 1009) (1501). Note that if the result 1010 of the analysis script table 1000 is NULL, the job control unit 208 determines the analysis script registered in the script 1008 of the record as an analysis script waiting for job execution.

  Next, the job control unit 208 merges the analysis scripts acquired in step 1501 and combines them into one analysis script (1502). Thereby, the CPU 103 does not need to read the next analysis script every time the execution of the analysis script is completed, the number of file input / output operations can be reduced, and the analysis process can be executed at high speed. Details of the processing in step 1502 will be described with reference to FIG. Further, the processing in step 1502 may be executed by the analysis execution unit 209 instead of the job control unit 208.

  Next, the analysis execution unit 209 executes the analysis script synthesized in step 1502 (1503). Specifically, the analysis script interpretation unit 210 of the analysis execution unit 209 executes the analysis script on the analysis target data 212.

  The job control unit 208 waits for processing until the analysis execution unit 209 completes execution of the analysis script (1504). When the analysis execution unit 209 completes execution of the analysis script, the job control unit 208 acquires the analysis result from the analysis execution unit 209. (1505).

  Next, the job control unit 208 registers the analysis result acquired in step 1505 in the result 1010 of the analysis script table 1000 of the analysis result database 200 (1506).

  Next, after sleeping for a certain time (1507), the job control unit 208 returns to the process of step 1501 and restarts the analysis execution process. The job control unit 208 repeatedly executes the analysis execution process in an infinite loop as long as there is no forced termination from the user with a certain period of sleep. Since the inference rule control unit 201 keeps registering new analysis scripts in the analysis script table 1000 of the analysis result database 200, the job control unit 208 also causes the analysis execution unit 209 to execute the generated analysis script, and the analysis results are stored in the analysis result database. It must continue to be registered in the 200 analysis script tables 1000.

  FIG. 16 is an explanatory diagram of merging analysis scripts according to the embodiment of this invention.

  In FIG. 16, reference numerals 1600 and 1601 denote analysis scripts to be merged, and reference numeral 1602 denotes an analysis script synthesized into one after merging.

  The underlined portion above 1602 is a portion corresponding to the analysis script 1600, and the underlined portion below 1602 is a portion corresponding to the analysis script 1601.

  As shown in FIG. 16, a program of a part excluding the underlined portion is prepared in advance as a template, and the job control unit 208 or the analysis execution unit 209 refers to the analysis script and inserts the underlined portion into the template. A single analysis script is generated by merging a plurality of analysis scripts.

  The analysis execution unit 209 calls the mapper function 1602 on the analysis target data 212 illustrated in FIG. 3 and acquires the analysis result. The analysis execution unit 209 calls a reducer function on the acquired analysis result. Accordingly, the analysis execution unit 209 can simultaneously execute the analysis process described in the analysis script 1600 and the analysis process described in the analysis script 1601.

  Further, when the capacity of the analysis target data 212 is enormous, the processing speed of the analysis execution process decreases due to file input / output of the analysis target data 212. Further, since the analysis described in the analysis script 1600 and the analysis described in the analysis script 1601 analyze the same analysis target data 212, the analysis scripts 1600 and 1601 are merged into one analysis script 1602 to perform analysis. Analysis can be performed by reading the target data 212 once, and analysis execution time can be shortened. As a result, the number of analysis scripts that can be executed can be increased. Therefore, by displaying preferentially what is highly evaluated to the user, the accuracy in terms of the usefulness of the analysis results can be increased.

  FIG. 17 is an explanatory diagram of an analysis result list display screen 1700 displayed by the analysis result display unit 213 according to the embodiment of this invention.

  The analysis result list display screen 1700 includes a search keyword input field 1701, a search button 1702, an analysis result list display field 1703, and a sort button 1709.

  A search keyword input field 1701 is a text field in which a search keyword is input. The search button 1702 is operated to cause the analysis result display unit 213 to search for an analysis result that matches the search keyword input to the search keyword input field 1701 from the analysis result registered in the analysis result database 200. The analysis result search method searches, for example, an analysis result whose title registered in the title 1007 of the analysis script table 1000 of the analysis result database 200 matches the search keyword.

  The analysis result list display field 1703 is a field for displaying a list of analysis results that match the search keyword input in the search keyword input field 1701.

  The analysis result list display field 1703 includes a graph button 1704, an analysis ID 1705, a data ID 1706, a title 1707, and an inference rule 1708.

  The graph button 1704 is operated to display an analysis result display screen 1800 (see FIG. 18) on which the analysis result of the record is displayed on the analysis result display unit 213. An analysis ID is displayed in the analysis ID 1705. In the data ID 1706, an identifier indicating the type of the analysis target data 212 to be analyzed is displayed. The title 1707 displays the title of the analysis script that acquired the analysis result. The inference rule 1708 displays the rule ID of the inference rule unit 202 that generated the analysis script that acquired the analysis result.

  The analysis result list display field 1703 displays the analysis results sorted in a predetermined order (sort parameter). In FIG. 17, it is assumed that the predetermined order is the order from the highest evaluation value of the analysis result. The evaluation value used for this sort is called a comprehensive evaluation value, and specifically, an average value of evaluation values calculated by different inference rules for a certain analysis result.

  The sort button 1709 is a button operated to display the sort parameter setting screen 1900 (see FIG. 19) for setting the sort parameter of the analysis result displayed in the analysis result list display field 1703 on the analysis result display unit 213. It is.

  FIG. 18 is an explanatory diagram of an analysis result display screen 1800 displayed by the analysis result display unit 213 according to the embodiment of this invention.

  The analysis result display screen 1800 includes a graph display field 1801, an evaluation value display field 1802, a user name input field 1803, an inference rule input field 1804, an evaluation scale input field 1805, an evaluation value input field 1806, and an evaluation button 1807.

  A graph display field 1801 is a field for displaying a graph of an analysis result obtained by operating the graph button 1704. When the analysis result display unit 213 detects that the graph button 1704 has been operated, the graph generation unit 215 generates graph data of the analysis result in which the graph button 1704 has been operated, and the graph generated by the graph generation unit 215 Display data in a graph display field.

  The evaluation value display field 1802 is a field for displaying the evaluation value calculated by the analysis result evaluation unit 203 of each inference rule unit 202 of the analysis result. The evaluation value display field 1802 displays the rule ID of the inference rule unit 202 that calculated the evaluation value, the evaluation scale that calculated the evaluation value, and the evaluation value.

  A user name input field 1803, an inference rule input field 1804, an evaluation scale input field 1805, and an evaluation value input field 1806 are fields for inputting values when the user changes the evaluation value of the analysis result.

  A user name input field 1803 is a text field for inputting a user name. The inference rule input field 1804 is a field for inputting the rule ID of the inference rule unit 202 that has calculated the evaluation value to be changed. An evaluation scale input field 1805 is a field for inputting an evaluation scale in which an evaluation value to be changed is calculated. The evaluation value input field 1806 is a field for inputting an evaluation value to be changed.

  The evaluation button 1807 is an evaluation specified by the rule ID input in the inference rule input field 1804 and the evaluation scale input in the evaluation scale input field 1805 among the evaluation values of the analysis results displayed in the graph display field 1801. The button is operated to change the value to the evaluation value input in the evaluation value input field 1806.

  Specifically, the analysis result display unit 213 includes the analysis ID registered in the analysis_id 1102 among the records in the evaluation value table 1100 of the analysis result database 200 and the analysis ID of the analysis result displayed in the graph display field 1801 as a graph. Get matching records. Then, the analysis result display unit 213 matches the rule ID registered in the rule_id 1105 with the rule ID input in the inference rule input field 1804 among the acquired records, and the evaluation scale registered in the measure 1106 is an evaluation scale. The value 1107 of the record that matches the evaluation scale input in the input field 1805 is changed to the evaluation value input in the evaluation value input field 1806. In this case, the analysis result display unit 213 registers the current date and time in updated_at 1104 of the record in which value 1107 is changed. In addition, the analysis result display unit 213 displays the analysis ID registered in the analysis_id 1001 of the analysis script table 1000 of the analysis result database 200 in the title 1007 of the record that matches the analysis ID of the analysis result displayed in the graph display field 1801. The user name input in the name input field 1803 is added, and the current date and time is registered in updated_at 1004.

  Thereby, the user can set the evaluation value, the evaluation value of the analysis result that the user wants to further analyze can be set high, and the analysis can be prevented from being scattered.

  FIG. 19 is an explanatory diagram of a sort parameter setting screen 1900 displayed by the analysis result display unit 213 according to the embodiment of this invention.

  The sort parameter setting screen 1900 includes a sort button 1901 and a weight designation field 1902.

  The sort button 1901 recalculates the evaluation values with the weights specified in the weight specification field 1902, and sorts the analysis results displayed in the analysis result list display field 1703 shown in FIG. 17 in the order of the recalculated evaluation values. The button to be operated.

  The weight designation field 1902 is a field for accepting designation of weight for each evaluation scale of each inference rule unit 202, and includes an inference rule 1903, an evaluation scale 1904, and a weight 1905.

  In the inference rules 1903, rule IDs of all inference rule sections 202 that have evaluated the analysis results displayed in the analysis result list display field 1703 are displayed. The evaluation scale 1904 displays the evaluation scales of all inference rule units 202 that have evaluated the analysis results displayed in the analysis result list display field 1703.

  A weight 1905 is a slide bar for the user to adjust (change) the weight of the evaluation scale of each inference rule unit 202. The user can specify a weight desired by the user by adjusting the slide bar having the weight 1905.

  A case where the sort button 1901 is operated will be described in detail.

  When the analysis result display unit 213 detects that the sort button 1901 has been operated, the analysis result display unit 213 applies the weight of each evaluation scale of each inference rule unit 202 specified in the weighting specification field 1902 to obtain a comprehensive evaluation value of each analysis result. Is calculated. The total evaluation value is a value obtained by calculating a value obtained by multiplying the evaluation value for which the analysis result is calculated by the weighting of the inference rule unit 202 for calculating the evaluation value and the evaluation criterion, and adding all the weighted evaluation values. It is.

  For example, for an analysis result, the evaluation value E1 is calculated with the evaluation scale “uniform” of the inference rule “cross”, the evaluation value E2 is calculated with the evaluation scale “extreme value” of the inference rule “scale”, and the inference rule “ The evaluation value E3 is calculated with the evaluation scale “maximum value” of “scale”, designated as the weight W1 of the evaluation scale “uniform” of the inference rule “cross”, and the evaluation scale “extreme value” of the inference rule “scale” If the weighting W2 of the evaluation rule “scale” and the weighting W3 of the “maximum value” are designated, the total evaluation value is calculated by (W1 × E1 + W2 × E2 + W3 × E3) / 3.

  The analysis result display unit 213 sorts the analysis results in descending order of the comprehensive evaluation values (in descending order of the comprehensive evaluation values), and displays the analysis results in the analysis result list display field 1703 in the sorted order.

  As described above, the list of analysis results can be sorted according to the evaluation criteria that matches the user's viewpoint.

  FIG. 20 is an explanatory diagram of an analysis script input screen 2000 displayed by the analysis user interface unit 214 according to the embodiment of this invention.

  The analysis script input screen 2000 is a screen displayed by the analysis user interface unit 214 when the user makes a predetermined input, and is a screen for accepting input of an analysis script created by the user.

  The analysis script input screen 2000 includes an analysis script input field 2001, a test execution button 2002, an analysis execution button 2003, a log display field 2004, a graph display field 2005, a user name input field 2405, a title input field 2406, and a save button 2407. .

  An analysis script input field 2001 is a text field for the user to input an analysis script.

  The test execution button 2002 is a button operated to cause the analysis execution unit 209 to execute a test on the analysis script input in the analysis script input field 2001.

  The analysis execution button 2003 is a button operated to cause the analysis execution unit 209 to execute the analysis script input in the analysis script input field 2001.

  In the log display field 2004, an analysis result of the analysis script when the test execution button 2002 is operated or an analysis execution button 2003 is operated, a log regarding the progress of the analysis processing of the analysis script, and the like are displayed.

  In the graph display field 2005, the analysis result of the analysis script input in the analysis script input field 2001 is displayed in a graph.

  The user name input field 2405 is a text field for inputting the user name of the user who created the analysis script. A title input field 2406 is a text field into which a title of an analysis script created by the user is input.

  The save button 2407 is a button operated by the analysis user interface unit 214 to register the analysis result of the analysis script created by the user in the analysis result database 200.

  A case where the save button 2407 is operated will be described in detail.

  When the save button 2407 is operated, the analysis user interface unit 214 registers the analysis result created by the user in the analysis script table 1000 of the analysis result database 200. Specifically, the analysis user interface unit 214 adds a new record to the analysis script table 1000, registers a unique analysis ID in the analysis_id 1001, and sets the analysis target data 212 that is the analysis target of the analysis script in the data_id 1002. An identifier indicating the type of the current data is registered, the current date and time are registered in created_at 1003 and updated_at 1004, and NULL is registered in parent_analysis_ids 1005. Also, the analysis user interface unit 214 registers the title input in the title input field 2406 and the user name input in the user name input field 2405 in the title 1007 of the new record, and inputs the analysis script input field 2001 in the script 1008. The registered analysis script is registered, and NULL is registered in expected_fitness 1009 and result 1010.

  Since the analysis result of the analysis script created by the user is registered in the analysis script table 1000, a new analysis script is generated from the analysis result. As a result, it is possible to correct that the analysis executed in the data analysis system 100 proceeds in a direction not intended by the user by creating a desired analysis script by the user.

  FIG. 21 is a block diagram of the data analysis system 100 including a plurality of computers according to the embodiment of this invention.

  In FIG. 21, a case where the data analysis system 100 includes a management computer 2100, inference rule computers 2110A, 2110B, and an analysis execution computer 2120 will be described as an example in which each part of the data analysis system 100 is distributed to a plurality of computers. 21 that are the same as those in FIG. 2 are given the same reference numerals, and descriptions thereof are omitted.

  The management computer 2100 includes an analysis result database 200, an inference rule control unit 201, a job control unit 208, an analysis result display unit 213, an analysis user interface unit 214, and a graph generation unit 215.

  The inference rule computer 2110A includes an inference rule unit 202A, and the inference rule computer 2110B includes an inference rule unit 202B.

  The analysis execution computer 2120 includes an analysis execution unit 209.

    As described above, each unit of the data analysis system 100 according to the present embodiment may be distributed among a plurality of computers.

  The present invention relates to a data analysis apparatus, and in particular, can be applied to a technique for automatically executing a total analysis.

101 Input device 102 Network device 103 CPU
DESCRIPTION OF SYMBOLS 104 Main memory part 105 Auxiliary memory part 106 Bus 200 Analysis result database 201 Inference rule control part 202 Inference rule part 203 Analysis result evaluation part 204 Analysis script generation part 208 Job control part 209 Analysis execution part 210 Analysis script interpretation part 211 Analysis object data Management unit 212 Analysis target data 213 Analysis result display unit 214 Analysis user interface unit 215 Graph generation unit

Claims (16)

A data analysis system comprising a processor for executing a program, and a storage area for storing a program executed by the processor, and analyzing data,
Performing an analysis process on the data, and storing an analysis result of the analysis process in the storage area;
A plurality of inference rule parts for evaluating the analysis result and generating a new analysis process from the analysis result;
An inference rule control unit for controlling the plurality of inference rule units;
An analysis processing management unit that manages unprocessed analysis processing of the analysis execution unit,
The inference rule unit includes an analysis result evaluation unit that evaluates the analysis result, and an analysis process generation unit that generates the new analysis process from the analysis result,
The analysis result evaluation unit reads the analysis result stored in the storage area, evaluates the analysis result by calculating a feature amount indicating a characteristic of the analysis result, and associates the analysis result with the feature amount. And store it in the storage area,
The inference rule control unit selects an analysis result in which the feature quantity satisfies a predetermined condition from analysis results stored in the storage area as a candidate for a generation source of the new analysis process,
The analysis processing generation unit generates a new analysis process from the analysis result selected by the inference rule control unit,
The analysis processing management unit causes the analysis execution unit to execute the analysis processing generated by the analysis processing generation unit. An input unit for receiving input from the user;
An analysis result display unit for outputting an analysis result display screen showing one analysis result,
The analysis result display unit
Including a feature value designation area that can be designated from the input unit as a value of the feature value of the output analysis result, and outputting the analysis result display screen;
2. The data analysis system according to claim 1, wherein a value designated in the feature quantity designation area is stored in the storage area as the feature quantity of the output analysis result. An input unit for receiving input from the user;
An analysis result that generates a list of analysis results obtained by extracting analysis results that match a predetermined condition from the analysis results stored in the storage area, and that outputs an analysis result list display screen showing the list of the generated analysis results A display unit,
The analysis result display unit
For each of the analysis result evaluation units that have calculated the feature amount, a weight designation area that can be designated from the input unit is output.
The analysis result display unit weights the feature amount based on the weighting designated in the weighting designation region,
The data analysis system according to claim 1, wherein the analysis results displayed on the analysis result list display screen are rearranged in the order of the weighted feature amounts. An analysis script input receiving unit for receiving an input from a user of an analysis script describing the analysis process;
The analysis script input receiving unit
Causing the analysis execution unit to execute the analysis processing by the analysis script that has received the input;
The data analysis system according to claim 1, wherein an analysis result of an analysis script that has received the input is stored in the storage area. When the analysis process generation unit generates the new analysis process, the analysis process generation unit generates the feature quantity of the analysis result that is a generation source of the new analysis process, and calculates the feature quantity of the analysis result by the new analysis process. Set it to the predicted value,
The data analysis system according to claim 1, wherein the analysis processing management unit determines a processing order of the analysis processing generated by the analysis processing generation unit based on the predicted value. The analysis processing management unit or the analysis execution unit merges an analysis script describing the plurality of analysis processes into one script,
The data analysis system according to claim 1, wherein the analysis execution unit executes a plurality of analysis processes by executing the merged script.   The analysis processing generation unit determines the number of the analysis processes to be generated based on the weight set for each inference rule unit and the number of execution processes not yet executed by the analysis execution unit. The data analysis system according to claim 1. The analysis result evaluation unit calculates the feature amount based on a rule corresponding to an analysis process generated by the analysis process generation unit belonging to the same inference rule unit,
The inference rule control unit has a predetermined feature amount calculated by the analysis result evaluation unit belonging to the same inference rule unit as the analysis processing generation unit that generates the new analysis process among the analysis results stored in the storage area. The data analysis system according to claim 1, wherein an analysis result that satisfies a condition is selected as a candidate for a generation source of the new analysis process. A data analysis method for analyzing data in a computer system comprising a processor for executing a program and a storage area for storing a program executed by the processor,
The computer system is
Performing an analysis process on the data, and storing an analysis result of the analysis process in the storage area;
A plurality of inference rule parts for evaluating the analysis result and generating a new analysis process from the analysis result;
An inference rule control unit for controlling the plurality of inference rule units;
An analysis processing management unit that manages unprocessed analysis processing of the analysis execution unit,
The inference rule unit includes an analysis result evaluation unit that evaluates the analysis result, and an analysis process generation unit that generates the new analysis process from the analysis result,
The method
The analysis result evaluation unit reads the analysis result stored in the storage area, and the rule corresponding to the analysis process generated by the analysis process generation unit belonging to the same inference rule unit, the feature amount indicating the characteristic of the analysis result Evaluating the analysis result by calculating based on, and storing the analysis result and the feature amount in association with each other in the storage area;
Among the analysis results stored in the storage area, the inference rule control unit is an analysis result evaluated by an analysis result evaluation unit belonging to the same inference rule unit as an analysis processing generation unit that generates a new analysis process, Selecting an analysis result for which the feature quantity satisfies a predetermined condition as a candidate for a generation source of the new analysis process;
The analysis processing generation unit generating a new analysis processing from the analysis result selected by the inference rule control unit;
And a step of causing the analysis execution management section to cause the analysis execution section to execute the analysis process generated by the analysis process generation section. An input unit for receiving input from the user;
An analysis result display unit for outputting an analysis result display screen showing one analysis result,
The method
The analysis result display unit including a feature amount designation area in which a user can designate a value of the feature amount of the output analysis result, and outputting the analysis result display screen;
The analysis result display unit includes a step of storing the value designated in the feature quantity designation area in the storage area as the feature quantity of the output analysis result. Data analysis method. An input unit for receiving input from the user;
An analysis result that generates a list of analysis results obtained by extracting analysis results that match a predetermined condition from the analysis results stored in the storage area, and that outputs an analysis result list display screen showing the list of the generated analysis results A display unit,
The method
The analysis result display unit outputs a weighting designation area in which weighting can be designated from the input unit for each analysis result evaluation unit that has calculated the feature amount;
The analysis result display unit weights the feature amount based on the weighting designated in the weighting designation region,
The data analysis according to claim 9, wherein the analysis result display unit includes a step of rearranging the analysis results displayed on the analysis result list display screen in the order of the weighted feature quantities. Method. An analysis script input receiving unit for receiving an input from a user of an analysis script describing the analysis process;
The method
The analysis script input receiving unit causing the analysis execution unit to execute the analysis processing by the analysis script that has received the input;
The data analysis method according to claim 9, further comprising a step in which the analysis script input reception unit stores an analysis result of the analysis script that has received the input in the storage area. When the analysis process generation unit generates the new analysis process, the feature value of the analysis result that is a generation source of the new analysis process is set as the feature value of the analysis result by the new analysis process. A step to set the predicted value;
The data analysis method according to claim 9, further comprising: a step in which the analysis processing management unit determines a processing order of the analysis processing generated by the analysis processing generation unit based on the predicted value. The analysis processing management unit or the analysis execution unit, merging the analysis script describing the plurality of analysis processing into one script;
The data analysis method according to claim 9, further comprising: executing a plurality of analysis processes by executing the merged script by the analysis execution unit.   The analysis processing generation unit includes the step of determining the number of analysis processes to be generated based on the weight set for each inference rule unit and the number of execution processes not yet executed by the analysis execution unit. The data analysis method according to claim 9. In the step of calculating the feature value by the analysis result evaluation unit, the feature value is calculated based on a rule corresponding to an analysis process generated by the analysis process generation unit belonging to the same inference rule unit,
In the step in which the inference rule control unit selects a candidate for the generation source of the new analysis process, the same inference rule as the analysis process generation unit that generates the new analysis process among the analysis results stored in the storage area The data analysis method according to claim 9, wherein an analysis result in which a feature amount calculated by the analysis result evaluation unit belonging to a condition satisfies a predetermined condition is selected as a candidate for a generation source of the new analysis process.

Images