JP2020009078A - Data processing system and data processing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a data processing system capable of accurately searching data for a database while reducing a user's load even if the characteristics of data are diverse. A data processing system for retrieving similar data based on a feature amount of user data, which includes a controller and a memory, and the controller executes a data processing program recorded in the memory. By doing so, the data structure to which the user data belongs is recognized, and the data structure includes a first node having the user data and a second node in the vicinity of the first node, and the second node is provided. The feature amount is calculated based on the node information. [Selection diagram] Fig. 1

Description

  The present invention relates to data processing for searching a database for data similar to user data.

  In the technical field dealing with various materials such as semiconductors and pharmaceuticals, databases are used for material development and research. The user of the database stores data such as experimental data and simulation data related to the material in the database, and performs data search in the database, for example, in order to obtain information that is useful for the material being developed by the user. There are many things.

  As a system for searching a database, for example, a system described in Patent Literature 1 is known. Patent Literature 1 discloses a substance that is a candidate from past simulations and experiments when there is a substance that is unknown in use when a new substance is generated by combining and blending substances each having multi-element / multi-component properties. Material action / combination analysis data that can provide information to users, has a method to save past experiments and costs and availability for each substance, and reflects them in calculations, and to search for the compounded substance. A conservative extraction mechanism is disclosed.

  Further, in Patent Literature 2, when a search condition is input, a search target database extraction unit extracts a database including data to be narrowed down by the search condition. Then, the data that has been individually added in the database to the same record is practically integrated, and based on the integrated information, a record that matches the search condition is searched. Upon receiving a systematic search command that specifies a specific record, another record that is systematically close to the specified record is searched, and the search result display unit displays the search result by the integrated search unit and the search result by the systematic search unit. A database search device displayed on a screen of a display device is disclosed.

JP 2012-048615 A JP-A-1999-175552

  When searching for a material in a database, a characteristic of the material is used as a scale for determining the similarity between the material to be searched and the material registered in the database. Features include physical measurable length, mass, intensity, and other information about molecules and atoms calculated by a computer, such as electronic states (energy, etc.), as well as calculations such as statistical processing. Some can be obtained by processing. As described above, since the types of features are various, when performing a data search in a database or registering a material, forcing a user to select or specify a feature not only increases the load on the user, Data retrieval accuracy may be reduced.

  Accordingly, an object of the present invention is to provide a data processing system capable of accurately performing data search on a database while reducing the load on a user even if the characteristics of the data are diverse.

  In order to achieve the above object, the present invention is a data processing system for searching for similar data based on a feature amount of user data, comprising a controller and a memory, wherein the controller records on the memory By executing the data processing program, the data structure to which the user data belongs is recognized, and the data structure includes a first node having the user data and a second node near the first node. And calculating the characteristic amount based on the information of the second node. Further, the present invention is also a data processing method for searching for similar data based on a feature amount of user data in order to achieve the above object.

  Advantageous Effects of Invention According to the present invention, it is possible to provide a data processing system and a data processing method capable of accurately performing data search for a database while reducing the load on a user even if the characteristics of the data are diverse.

FIG. 1 is an example of a block diagram of a data processing system. FIG. 2 is a block diagram showing the data structure of each of the databases DB1 and DB2. FIG. 3 is a block diagram showing the relationship between the data structure of the database, categories of user data, and characteristics of user data. FIG. 4 is a block diagram of the business server. FIG. 5 is a block diagram of the terminal computer. FIG. 6 is an example of the above-described similar data search screen of the terminal computer. FIG. 7 is an example of a user data registration screen of the terminal computer. FIG. 8 is an example of a category determination condition input screen of the terminal computer. FIG. 9 is an example of a category correspondence table of the business server. FIG. 10 is an example of a feature correspondence table of the business server. FIG. 11 is another example of the feature correspondence table of the business server. FIG. 12 is an example of a calculation formula list table of the business server. FIG. 13 is a flowchart showing the flow of processing performed by the business server from when the user data is input from the terminal computer to when the feature amount is calculated. FIG. 14 is a block diagram of a data structure for explaining that there are a plurality of routes at neighboring nodes within the range of the relative path number. FIG. 15 is a block diagram of a data structure with another route to a neighboring node. FIG. 16 is an example of the access history table of the business server.

  Next, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is an example of a block diagram of a data processing system. The data processing system includes a business server 101 connected to a plurality of databases (database servers) 10-n via a network (Net1), and a business server 101 connected to the business server 101 via a network (Net2). A user-side terminal computer 102 for requesting data registration, data retrieval, and the like to a database.

  The business server 101 provides a plurality of terminal computers 102 with a cloud service for searching for data and registering data for a plurality of databases. The terminal computer 102 can use, for example, a cloud service to enhance a solution for material development.

  The database is not particularly limited, and may be, for example, MongoDB, PostgreSQL, MySQL, MariaDB (the above is a unique name of the database). The database stores open data and user internal data. The database may be managed and practiced by a user. The terminal computer 102 registers experimental data and analysis data (such as three-dimensional simulation) associated with material development in the database via the business server 101, Alternatively, similar data can be searched based on a database.

  2. Description of the Related Art Conventionally, a user has widely performed detection and extraction of data similar to user data from a database by paying attention to characteristics of the data, such as experimental data and simulation data, or characteristics accompanying the data. . The characteristic is, for example, individual information of material data, and this includes various types such as tensile strength, density, and strength.

  The number of combinations of features provided by the material and features not provided by the material is enormous depending on the type and classification of the material, such as the composition, shape, structure, and physical properties of the material. Nevertheless, conventionally, the user himself had to select, select, or determine the feature, calculate the feature amount, and search for similar data from the database based on the feature amount. Not only was the load heavy for the user, but if the selection of the feature was not appropriate, the accuracy of the search results could be reduced.

  Thus, the business server 101 executes a data processing program in response to a user request from the terminal computer 102, and determines the type, range, and the like of the characteristics of the user data based on the data structure of the user data stored in the database. Determine, and use the user data to calculate feature weights, e.g., feature values, and then search the database based on the calculated feature values for similar data to reduce user load. Release and improve search accuracy.

  It is not easy for the business server 101 to determine the characteristics of the user data from the user data itself. This is because the values and formats of the experimental data and the simulation data greatly differ depending on the user or the data.

  On the other hand, in a database in which user data is recorded in a text-based format (for example, json or star schema), in a data structure, the vicinity of a user data node (which may be paraphrased as “periphery”) There are nodes that record as text-based metadata such as indexes, flags, summary sentences, keywords, and general words for data. This node is hereinafter referred to as a “neighbor node”. In addition, it may be paraphrased into another term such as “peripheral node”.

  The metadata is useful for determining, specifying, determining, or determining the type, classification, range, and the like of user data (hereinafter, these are collectively referred to as “categories”). Therefore, the business server 101 determines the category of the user data by means of a correspondence table, machine learning, calculation, or the like based on the storage information of the neighboring nodes. Once the category of the user data is determined, the types of features related to the category are also limited. Therefore, the business server 101 can easily find the value of the limited feature based on the user data. The correspondence between the category and the feature is made possible by means such as a table, machine learning, or calculation.

  The range of nodes near the user data node may be specified by a relative path starting from the user data node. A path is a link between nodes. The neighbor node may be a node having a specific number of relative paths from the origin node, or a plurality of nodes within a predetermined number of paths from the origin node. Must be a neighboring node. For example, a node far from the user data node loses its relationship with the user data and may not correctly reflect the category of the user data.

  Limiting the range of neighboring nodes is not very meaningful. Any node that includes metadata that reflects the category of user data may be said to be a neighboring node. For example, a node having a relative path number of 10 or less may be a neighboring node. The relative path may be defined by a combination of the number of paths to the parent node and the number of paths to the child node. For example, the neighboring nodes whose relative paths are defined by (Parent x <= 2) / (child x <= 2) are within two paths on the parent node side and then on multiple paths within two paths on the child node side. One or more of the nodes becomes a “neighbor node”.

  FIG. 2 is a block diagram showing the data structure of each of the databases DB1 and DB2. The data structure is obtained by connecting a plurality of nodes in a tree structure. In each of the data structures of DB1 and DB2, a node marked as calculation target data is a node of user data. The business server 101 can recognize the data structure of the database and, since it has the link information to the node of the calculation target data, can recognize the node of the user data in the data structure.

  When recognizing the range of neighboring nodes based on the number of relative paths from the originating node, the business server 101, when there are a plurality of nodes, sequentially scans these and determines a category based on data in the nodes.

  In the DB1 of FIG. 2, the business server 101 stores information within a range of a relative path from the origin node (calculation target data), that is, ((Parent x <= 2) / (child x <= 2) / *) The category is determined to be “metal” based on the information belonging to the “chemical component node” (TN200) as a neighboring node within the range (the text “Cu, Ni” which is information related to the category).

In the DB2 as well, the business server 101 determines that a neighboring node is within a range of a relative path ((Parent x <= 2) / (child x <= 2) / *) from the origin node (data to be calculated). The category is determined as “metal” based on the text “Ni” of “Component Element Prop” (TN200).

  Further, the business server 101 sets the “material shape” as a neighboring node within a range of a relative path ((Parent x <= 2) / (child x <= 2) / *) from the origin node (calculation target data). (TN202), that is, the category is determined to be “casting” based on the text “Cast Block” which is information related to the above-described category.

In the DB2 as well, the business server 101 determines that a neighboring node is within a range of a relative path ((Parent x <= 2) / (child x <= 2) / *) from the origin node (data to be calculated). The category is determined to be “casting” based on the information belonging to “tag” (TN202), that is, the text “Alloy cast iron” which is the information related to the above-described category. The business server 101 determines the category as "casting" based on the information belonging to the "material shape node" which is a neighboring node and the text "CastBlock" which is information related to the above-described category.

  FIG. 3 is a block diagram showing a relationship between a data structure 301 of a database, a category 302 of user data, and a feature 303 of user data. The business server 101 specifies a category from a plurality of categories based on “Cu, Ni” which is information related to the category and belongs to a “chemical component” node which is a node close to the node of the calculation target data. I do.

  The business server 101 has a plurality of categories (302). Relevant features 303 are assigned to each category in advance according to the type of category. The business server 101 specifies the feature 303 from the category 302 according to the assignment. The feature is associated with an expression for calculating the feature amount, and the business server 101 calculates the feature amount by applying the calculation target data to the calculation expression. The business server 101 can select a feature suitable for the calculation target data via the category.

  Next, the business server 101 will be described in detail. FIG. 4 is a block diagram of the business server 101. The business server 101 includes computer hardware resources, software resources, and data resources. The hardware resources include a controller (CPU) 402, a memory 403, and a communication module 404a. The software resources are realized by the programs in the memory 403, and are each implemented by a neighboring node extraction module 405, a category identification amount module 406, a feature determination module 407, an access history collection module 412, a feature amount calculation module 413, and similar data. A search module 414 and a user data registration module 416 are provided. The above-described module is achieved by the controller 402 executing a memory program.

  The data resource includes a category correspondence table 408, a characteristic correspondence table 409, a calculation formula list table 410, and an access history information table 411.

  The business server 101 executes the neighborhood node extraction module 405 to extract the neighborhood nodes (the “chemical component” node, the “material shape” node, etc. in FIG. 2) for the user data node based on the number of relative paths. The business server 101 executes the category specifying module 406 to scan the neighboring nodes, extract information on the category (“Cu”, “Ni”, “Cast” in FIG. 2), and perform the category based on this information. The “category” is specified with reference to the correspondence table 408 (FIG. 3: link between 301 and 302).

  Further, the business server 101 executes the feature determination module 407, and determines the feature based on the category and the feature correspondence table 409 (FIG. 3: link between 302 and 303). The business server 101 executes the feature amount calculation module 413, refers to the determined calculation formula list table 410, acquires a calculation formula corresponding to the feature, and calculates the feature value by applying user data to the calculation formula. I do.

  The business server 101 executes the similar data search module 414 to search the database for similar data based on the feature amount. The similar data may be data having a value in a range similar to the feature amount for the same or similar feature. The business server 101 executes the user data registration module 416 to store data uploaded from the terminal computer 102 in a node of a data structure of a predetermined database.

  Further, the access history collection module 413 collects the history of the server 101 accessing the database, and records this in the access history information table 411. The access history is useful when the neighboring node extraction module 405 extracts a neighboring node from the originating node. The communication module 404a executes communication with the terminal computer 102 and the database (DBn).

  FIG. 5 is a block diagram of the terminal computer 102. The terminal computer 102 includes a controller 417, a memory 415, and a communication module 404b. The memory 415 implements a GUI for supporting a request to the business server 101 for a user. The GUI includes an input screen 415A for the user to search the database (DBn) for data similar to the user data, an input screen 415B for the user to register the user data in the database (DBn), and the business server 101. An input screen (category determination condition input screen) 415C for registering the category correspondence table 408 (FIG. 4A) is provided to the user. The communication module 404b executes communication with the business server 101.

  FIG. 6 is an example of the similar data search screen 415A described above. Reference numeral 502B denotes an operation unit for a user to upload input data 502A ([C: \ input.json]: path to input data) of the terminal computer 102 to a data structure of a predetermined database via the business server 101. is there. Reference numeral 503 denotes an area for displaying the categories specified by the category specifying module 406 of the business server 101 (two categories of “metal” and “rod”).

  Reference numeral 504A denotes an area for displaying the types of characteristics (two characteristic names of “tensile strength” and “elongation at break”) specified by the characteristic determination module 407 of the business server 101. If the user data is already registered in the predetermined database, instead of inputting to the 502A and 502B, the business server 101 displays, on the similar data search screen 415A, link information to the database in which the user data is recorded. May be displayed.

  Reference numeral 504B denotes an area for displaying a feature amount corresponding to a feature name calculated by the feature amount calculation module 413 of the business server 101. The feature value for “tensile strength” (feature name) is “250 MPa”, and the feature value for “elongation at break” (feature name) is “0.2%”. Reference numeral 505 denotes an area for displaying similar data obtained by the similar data search module 414 of the business server 101 searching the database. Reference numeral 505 denotes a metal A (“tensile strength”: feature amount “250 MPa”) and a metal B (“tensile strength”: feature amount “235 MPa”) for the feature amount “250 MPa” of “tensile strength”. And that it was discovered.

  FIG. 7 is an example of the user data registration screen 415B. Reference numeral 508A denotes an area in which a category is displayed, and the business server 101 executes the neighborhood node extracting module 405 and the category specifying module 406 to specify the category by inputting to the area of 508B, and specifies the category. (For example, “metal” and “rod”) are displayed in the area 508A.

  Further, the business server 101 executes the feature determination module 407 and the feature amount calculation module 412 by inputting the information to the area 509C, and stores the determined feature names (eg, “tensile strength” and “elongation at break”) in the area. 509A, and the calculated value of the feature amount is displayed on 509B. The user may input a predetermined value in the area 508A and the areas 509A and 509B. In this case, the similar data search module 414 of the business server 101 may search for similar data based on the input feature value name and feature value.

  FIG. 8 is an example of a category determination condition input screen 415C. The area 511 is an area where the user inputs the range of the neighboring node (relative path from the origin node). The business server 101 may display the data structure (FIG. 2) of the user data existing in the database on the input screen to support the user's input. At this time, the business server may notify the user of the node of the user data in the data structure. Upon receiving this notification, the user may specify the range of the neighboring node.

  "(Parent x <= 2) / (Child x <= 3) / *" indicates a relative path from the origin node (user data node). It means that the node is within two nodes, and further from that, the node within three is specified as the range of the neighboring nodes on the child node side. The neighborhood node extraction module 405 of the business server 101 sequentially scans the neighborhood nodes based on the input information and searches for category related information. The user may input a specific node such as “(Parent x = 2) / (Child x = 3) / *”. Further, the neighboring node extraction module 405 may determine the range of the neighboring nodes or the neighboring nodes depending on the characteristics and attributes of the user data, for example, whether the data format of the user data is json or a star schema. Good.

  Next, the user inputs category determination conditions to 513A to 513E. For example, as illustrated, when the user inputs (513C) that includes (or does not include) Cu (513B) in the value (513A) of (neighboring node) and operates the determination (513D), the determination is made. The conditions are fixed, while the judgment conditions can be edited (513E). Further, when the determination condition is satisfied, the user inputs that the category to be determined is “metal” (514). The user may select an intended category from a plurality of categories. Note that the business server 101 may verify the affinity between the input conditions (513B, 513C) and the category name (514) and output a warning display to the user. For example, a category (polymer) does not fit into Cu or Ni.

  FIG. 9 is an example of the category correspondence table 408. The category specifying module 406 updates and records the input information (415C in FIG. 5, FIG. 8) in the category correspondence table 408. In the category correspondence table 408, a record 602 for recording information of a relative path from the origin node, a record 603 of the condition of the category related information, and a record 604 of the category when the category related condition is satisfied correspond to each other. It is recorded as follows.

  FIG. 10 is a table as an example of the feature correspondence table 409, in which categories are associated with feature names. This correspondence table 409 associates category names 702 with feature names 703. This table may be created by the user and registered in the business server 101, or may be based on the analysis of the big data by the business server 101 or the like. The feature determining module 407 determines a feature name based on the category specified by the category specifying module 406.

  A plurality of feature names may be assigned to one category, or a plurality of categories may be assigned to the same feature name. The feature correspondence table 409 (FIG. 10) further has a record 704 called “feature weight”. The feature weight refers to the degree of influence that the feature name corresponding to the category recorded in the category name has in this category. In FIG. 10, the “tensile strength” exerted on “metal” is 0.3, whereas the “tensile strength” exerted on “metal” is 0.3. It is low 0.2. That is, “tensile strength” is a factor that should be emphasized in the “bar-shaped metal” category over the simple “metal” category. When calculating the feature amount for the category specified by the related information of the neighboring node, the feature amount calculation module 413 can improve the detection accuracy of similar data by reflecting the weight on the feature amount.

  FIG. 11 shows a feature list table 801 as another example of the feature correspondence table 409. The feature determination module 407 reads the categories of the category correspondence table 408 in order from the top, and registers the feature name and the feature weight corresponding to the category in the feature list table 801 with reference to the correspondence table 409 (FIG. 10). Reference numeral 804 denotes a feature weight total table serving as a feature correspondence table 409. The feature determination module 407 refers to the feature list table 801 to sum the feature weights of the same feature name, and stores the sum in the feature total list table 804. register. The feature weight sum table 804 determines the types of features included in the user data and the degree of influence of the features in the user data.

  FIG. 12 shows an example of the calculation formula list table 410. The calculation formula list table 410 includes a record 902 of a feature name, a record 903 of a computable condition, and a record 904 of a calculation formula for calculating a feature amount. This table may be defined in the business server 101 by the user. The feature value calculation module 413 refers to the feature names in the feature weight total table 804 in order, and based on the calculation formula list table 410 (FIG. 9), the computable condition 903 and the feature value calculation formula 904 corresponding to the feature name. To determine.

  Then, the feature amount calculation module 413 determines whether or not each of the feature names recorded in the feature weight total table 804 matches the calculatable condition, and if so, converts the user data into a feature amount calculation formula. By applying, the value of the feature amount is calculated for each feature name. The feature value calculation module 413 calculates a correction value of the feature value by multiplying the feature value for each feature name by the total weight according to the feature weight sum table 804, and uses the calculated value as a search screen in FIG. 6 (504B in FIG. 6). Output to The similar data search module 414 searches the database based on the correction value for each feature name, and displays records having similar feature amounts as a list on the search screen (FIG. 6: 505).

  Next, the flow of processing from when the business server 101 receives user data input from the terminal computer 102 to calculate a feature amount for searching for similar data will be described again with reference to the flowchart in FIG. . First, when the business server 101 receives a similar data search target request from the terminal computer 102 (S1000), a list of user data (a list of which database contains which user data folder exists) is sent to the terminal computer 102. Output. Upon receiving the designation of the user data from the terminal computer 102, the business server 101 executes the neighboring node extracting module 405 to extract the neighboring nodes of the user data node according to the relative path number (511 in FIG. 8) (S1002). ).

  Subsequently, the category specifying module 406 scans the neighboring nodes, refers to the category correspondence table 408 (FIG. 8, FIG. 9: 603), and stores the category related information (“Cu,“ Ni ”) of the category correspondence table in the neighboring nodes. (Fig. 8: 513B, Fig. 9: 603), "Cast", and "Bar" (Fig. 9: 603)) include the category (Fig. 8: 514, Fig. 9: 604) corresponding to the related information. Is specified, and the specified category is displayed on the terminal computer 102 (503 in FIG. 6), and S1004 ends.

  Subsequently, the feature determination module 407 selects a feature name and a feature weight for the specified category with reference to the feature correspondence table 409 (FIG. 10), and registers these in the feature list table 801 (S1006). Subsequently, the feature determination module 407 calculates the sum of the feature weights of the rows including the same feature name in the feature correspondence table 409, and registers this in the feature weight total table 804 (S1008). In this step, in order to prevent a feature name having a small influence, that is, a feature name having a small total weight value from being registered, and to reduce a calculation load of the feature amount, the total weight value exceeds a predetermined threshold value. Only in this case, the feature amount name may be registered in the feature total list table 804.

  Subsequently, the feature value calculation module 413 refers to the calculation formula list table 410 (FIG. 12), selects a feature value calculation formula 904 corresponding to the feature name 902, and adds the user data and the sum of the feature name The feature amount is calculated by applying the weight (804 in FIG. 11) (S1100). Note that the feature amount calculation module 413 may determine the condition (903 in FIG. 12) of the calculation formula list table 410, and calculate the feature amount, for example, when the user data satisfies a predetermined condition. For example, the condition under which the elongation at break (feature name) can be calculated is “user data: format (3D)”, which indicates that the user data must be three-dimensional data in calculating the elongation at break. ing.

  Thus, the business server 101 ends the calculation of the feature amount for similar data search. The similar data search module 414 of the business server 101 searches the database based on the feature amount to search for similar data, so that even if the characteristics of the data are diverse, the data search for the database can be performed with reduced user load. While being executed with high accuracy.

  As described above, the range of a neighboring node is determined by the number of relative paths. Then, the neighboring node extracting module 405 extracts nodes within the range of the relative path number as neighboring nodes. By the way, there is a mode in which a plurality of nodes exist in a node within the range of the relative path number. The neighbor node extraction module 405 extracts all nodes. The neighboring node extracting module includes means for classifying extracted nodes, and can also extract the nodes by limiting them to some of the classified nodes. The access history collection module 411 described above is an example of this means. The access history is a history in which a user has accessed for registering and updating user data, and the history includes a record of a date and a route to an access destination.

  First, a description will be given of the fact that there are a plurality of routes at neighboring nodes in the range of the relative path number. When a plurality of pieces of neighboring data (data of neighboring nodes) exist in user data (original data for calculating a feature amount), neighboring data indicated by a relative path cannot be uniquely identified. For example, as shown in FIG. 14, user data (experiment X data combining material A1 and material A2) includes a neighbor node 1101 including a parent node “summary of material A1” and a parent node “summary of material A2”. It has two routes consisting of neighboring nodes 1102 via nodes.

  Further, as shown in FIG. 15, since the input data also includes the experiment data, the user data includes a parent node “Summary of Experiment X”, which is completely different from the system in FIG. The node 1103 will exist.

  Next, the access history will be described. When the terminal computer 102 accesses the GUI from the GUI, one layer of data is displayed on one display screen to the user. Therefore, after accessing the web page “Overview of material A1” in the layer of 1101 (FIG. 14), When accessing the web page "Experiment X data combining material A1 and material A2", a screen transition occurs. In such a screen transition, a history linking the previous access page and the current access page is generated. By using this history, the neighboring node extraction module 405 can limit the neighboring nodes within the range of the relative path number to “chemical combination A1”.

  When accessing user data from a query language such as SQL, the parent node and its child nodes can be accessed with a single query, and the target data name of a JOIN used to display multiple nodes simultaneously Is the access route history. For example, the query statement “SELECT * FROM experiment X data JOIN outline of material A1 USING material ID” is a query statement that simultaneously refers to “experiment X data combining material A1 and material A2” and its parent “summary of material AI”. It is. The data name on the right side of "JOIN", that is, "Summary of material A1" is used as the first access, the data name on the left side, "Experiment X data" is used as the next access, and the order of JOIN may be used as the access path history. .

  The access history collection module 413 collects the access history of the user and records the collected history in the access history information table 411. FIG. 16 is an example of the access history information table 411. The access route information includes a column 1205 of a single session ID given for each access from the user and a column 1206 of the access route. In 1206, an access route from a Web browser by a user or an access route determined from a query is recorded.

  The neighboring node extraction module 405 may extract a neighboring node according to an access history of a predetermined session ID among a plurality of session IDs. For example, in FIG. 16, when the session ID is the latest session, a neighboring node may be extracted based on this. By using the latest session, the feature amount of the user data can be obtained in accordance with the latest data.

  When the neighboring node extraction module 405 extracts a neighboring node based on the access history of the session ID 1, it acquires “material A1 + material A2_experiment X” in the access route “material list / material A1 / material A1 + material A2_experiment X”. If the relative path is “(Parent)” starting there, “material A1” is extracted as a neighboring node. As described above, the business server extracts the neighboring nodes using the access history, so that the range of the neighboring nodes can be limited, and the calculation of the feature amount becomes efficient.

  In the above-described embodiment, the user data is described as experimental data related to material development, and / or simulation data. However, the present invention is not limited to this, and particularly limited, such as big data for product development. It should not be. Furthermore, of the total weights (804 in FIG. 11), only the total weights equal to or greater than a predetermined threshold may be registered in the table 804. By doing so, it is possible to reduce the computational load on the server except for a type of feature amount that is less involved in searching for data similar to user data.

  The controller of the business server 101 executes the data processing program to realize the artificial intelligence, makes the artificial intelligence learn the combination of the category, the feature name, and the user data as learning data, and based on the learning result, the user as the data processing target. The characteristics of the user data may be determined based on the score calculated for the data. Further, although the table of FIG. 12 describes one calculation formula associated with one feature, a predetermined calculation formula from a plurality of calculation formulas may be associated with one feature.

  The present invention further relates to the business server 101, the terminal computer 102 for causing the business server 101 to execute data processing, the data processing program of the business server 101, and the data processing program recorded (non-temporary). Target), such as a hard disk and a flash memory. The embodiments described above are not intended to limit the technical features of the present invention, but are merely examples, and can be appropriately modified.

101: business server 102: terminal computer DB1, DB2,... DBn: database 402: controller

Claims (10)

A data processing system for searching for similar data based on a feature amount of user data,
A controller,
And a memory,
The controller executes a data processing program recorded in the memory,
Recognizing a data structure to which the user data belongs, the data structure comprising: a first node having the user data; and a second node near the first node.
Calculating the feature quantity based on the information of the second node;
Data processing system. The controller is
Acquiring related information related to the type of the feature of the user data from the information of the second node;
Identifying the characteristics of the user data based on the related information,
Calculating the feature amount based on the specified feature,
The data processing system according to claim 1. The memory includes a table that associates the related information with the feature,
The controller specifies the feature from the related information based on the table,
The data processing system according to claim 2. The memory includes a table that associates the specified feature with a calculation formula for the feature amount,
The controller is
Specify a formula based on the table,
Calculating the feature amount by applying the user data to the specified calculation formula,
The data processing system according to claim 3. The controller extracts the second node based on a relative path number with the first node as a starting node.
The data processing system according to claim 1. The first node is configured based on a text-based format;
The controller obtains the related information based on text information of the second node,
The data processing system according to claim 2. The controller is
Identifying a plurality of characteristics of the user data;
Setting a weight according to the degree of influence on the user data for each of the plurality of features,
Applying the weight to the calculation of the feature amount;
The data processing system according to claim 2. The controller is
Of the plurality of features, sum the weights set for the same feature,
Applying the total value of the weights to the calculation of the feature amount;
The data processing system according to claim 7. The controller is
Obtaining an access history to the first node;
Extracting the second node based on the relative path number and the access history;
The data processing system according to claim 5. A data processing method in which a computer searches for similar data based on a feature amount of user data,
The computer executes a data processing program,
Recognizing a data structure to which the user data belongs, the data structure comprising: a first node having the user data; and a second node near the first node.
Calculating the feature quantity based on the information of the second node;
Data processing method.

Images