Data analysis system and classification processing method

Abstract

To provide a data analysis system and a classification processing method that automatically and efficiently perform grouping of page information.SOLUTION: A data analysis system has: a classification device having classification processing means that receives page information linked to a web page in a web site as input of a classification model generated through machine learning using a plurality of pieces of page information in the past, and classifies the pieces of page information received as input into any one of a plurality of groups based on the similarity in properties; and an analyzer that executes analysis processing on data obtained through the classification performed by the classification device. The classification device has preprocessing means that generates, from display related data including information on objects in the web page, for every object, identification information obtained by combining type information indicating the type of the object and arrangement information indicating a position and a size of the object, and generates a frame feature quantity as page information on the basis of the generated identification information for every object.SELECTED DRAWING: Figure 6

Description

[Technical field]

[0001] The present invention relates to a data analysis system and a classification processing method for classifying page information associated with web pages on a website. [Background technology]

[0002] Websites are used not only to advertise products and services handled by companies, but also to transact those products and services. For this reason, website operators analyze users' browsing behavior on each webpage of the website from information linked to each webpage, and identify issues from the analysis results and make various improvements. However, since such information is generated in large quantities, grouping (classification processing) of the information for analysis is required to improve the accuracy and speed of the analysis. For example, the analysis system of Patent Document 1 performs a predetermined analysis process after grouping pages that exhibit similar characteristics or similar search keywords. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2016-201080 A Summary of the Invention [Problem to be solved by the invention]

[0004] However, in the conventional configuration as disclosed in Patent Document 1, the settings related to the similarity of pages and keywords must be manually configured, and therefore the reality is that this is not suitable for grouping a large amount of generated page information.

[0005] The present invention has been made to solve the above-mentioned problems, and has an object to provide a data analysis system and a classification processing method that automatically and efficiently group page information. [Means for solving the problem]

[0006] A data analysis system according to one embodiment of the present invention includes a classification device having a classification processing means that uses page information linked to a webpage on a website as input to a classification model generated by machine learning using multiple pieces of past page information and classifies the input page information into one of multiple groups based on similarity in properties, and an analysis device that performs analytical processing on data obtained by classification by the classification device, wherein the classification device has a preprocessing means that generates identification information for each object from display-related data including information of each object on the web page by combining type information indicating the type of the object and placement information indicating the position and size of the object, and generates frame features as page information based on the generated identification information for each object.

[0007] A data analysis system according to one embodiment of the present invention includes a classification device having a classification processing means that uses page information linked to web pages on a website as input to a classification model generated by machine learning using multiple pieces of past page information and classifies the input page information into one of multiple groups based on similarity in properties, and an analysis device that performs analytical processing on data obtained by classification by the classification device, in which the classification processing means includes address classification means that extracts address features from address information indicating the location of the web page as input to the address classification model generated by machine learning using multiple pieces of past address information, text classification means that extracts text features from text data, which is information about text in the web page, as input to the text classification model generated by machine learning using multiple pieces of past text data, and an overall classification means that inputs the address features and text features for each web page into the classification model as page information and groups the web pages.

[0008] A data analysis system according to one embodiment of the present invention includes a classification device having a classification processing means that uses page information linked to a webpage on a website as input to a classification model generated by machine learning using multiple pieces of past page information, and classifies the input page information into one of multiple groups based on similarity in properties, and an analysis device that performs analytical processing on data obtained by classification by the classification device, in which the classification processing means uses address information indicating the location of the webpage and text data that is information on the text within the webpage as page information as input to the classification model, and classifies the input page information into one of the multiple groups.

[0009] In a classification processing method according to one embodiment of the present invention, a classification device that classifies page information linked to web pages on a website generates, for each object, identification information from display-related data including information on each object on the web page by combining type information indicating the type of the object and placement information indicating the position and size of the object, generates frame features as page information based on the generated identification information for each object, uses the generated frame features as input to a classification model generated by machine learning using a plurality of past frame features, classifies the input frame features into one of a plurality of groups based on similarity in properties, and an analysis device performs analytical processing on the data obtained by classification by the classification device. Effect of the Invention

[0010] The present invention classifies page information that is input to a classification model based on a plurality of pieces of page information in the past into one of a plurality of groups based on the similarity of properties. Therefore, the grouping of page information linked to a web page can be performed automatically without manual work, thereby making the process faster and more efficient. [Brief description of the drawings]

[0011] [Figure 1]1 is a block diagram showing a configuration example of a data analysis system according to a first embodiment of the present invention and its peripheral devices, etc.; [Diagram 2] 2 is a block diagram illustrating a functional configuration of a management device in FIG. 1; [Diagram 3] 2 is a block diagram illustrating a functional configuration of the terminal device of FIG. 1. [Figure 4] 2 is a block diagram illustrating an example of a functional configuration of the data analysis system of FIG. 1. [Diagram 5] 2 is a flowchart illustrating the operation of the data analysis system of FIG. 1. [Figure 6] FIG. 2 is a block diagram showing a configuration example of a data analysis system and its peripheral devices according to a second embodiment (3 to 5) of the present invention. [Figure 7] 7 is a block diagram illustrating an example of a functional configuration of the data analysis system of FIG. 6. [Figure 8] FIG. 11 is a block diagram illustrating a functional configuration of a data analysis system according to a third embodiment of the present invention. [Figure 9] 7 is an explanatory diagram illustrating an example of a web page displayed on a display unit or the like of the terminal device of FIG. 6. [Figure 10] 10 is an explanatory diagram illustrating an example of a wireframe image in which only part information corresponding to each object in the web page of FIG. 9 is extracted as an image. [Figure 11] FIG. 11 is a block diagram illustrating a functional configuration of a data analysis system according to a fourth embodiment of the present invention. [Figure 12] 12 is an explanatory diagram illustrating an example of a data flow in the classification device of FIG. 11 in association with each step of a classification processing method. [Figure 13] FIG. 13 is a block diagram illustrating a functional configuration of a data analysis system according to a modification 4a of the fourth embodiment of the present invention. [Figure 14] FIG. 13 is a block diagram illustrating a functional configuration of a data analysis system according to a modification 4b of the fourth embodiment of the present invention. [Figure 15] FIG. 13 is a block diagram illustrating a functional configuration of a data analysis system according to a modification 4c of the fourth embodiment of the present invention. [Figure 16] FIG. 13 is a block diagram illustrating a functional configuration of a data analysis system according to a fifth embodiment of the present invention. [Figure 17] FIG. 13 is a block diagram illustrating a functional configuration of a data analysis system according to a modification 5a of the fifth embodiment of the present invention. [Figure 18] FIG. 13 is a block diagram illustrating a functional configuration of a data analysis system according to a modification 5b of the fifth embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] Embodiment 1 A configuration example of the data analysis system and its peripheral devices in the present embodiment 1 will be described with reference to Fig. 1. As shown in Fig. 1, the data analysis system 100 has a classification device 10 and an analysis device 20. The data analysis system 100, a management device 30, a service providing system 40, an analysis system 50, and a terminal device 80 are communicatively connected via a network N such as the Internet. In other words, the data analysis system 100 can communicate with the management device 30, the service providing system 40, the analysis system 50, and the terminal device 80 via wired or wireless communication.

[0013] The management device 30 is a PC (Personal Computer) managed by a website operator, that is, an operator who sells products and services using the service providing system 40. PCs include tablet PCs, notebook PCs, desktop PCs, and the like. However, the management device 30 may also be a mobile terminal such as a smartphone or a tablet terminal.

[0014] The service providing system 40 is a web server that provides information on products and services to a terminal device 80 or the like via a website such as an EC site (Electronic Commerce site). The website provided by the service providing system 40 can be viewed from the terminal device 80 or the like via a network N. The service providing system 40 is configured by a cloud server based on cloud computing, a physical server, or the like.

[0015] The analysis system 50 has a function for performing website access analysis, and is, for example, a server for Google Analytics (GA), an access analysis tool provided by Google (registered trademark). Access analysis is the analysis of the characteristics and behavior of users who visit a website in order to increase the number of conversions (CV) on the website. Conversion is the goal of the website, that is, the behavior that users are desired to achieve on the website, and refers to registration, purchase of a product or service, application, request for information, making a reservation, etc.

[0016] The analysis system 50 sequentially or periodically collects behavioral data indicating user behavior via the network from the service providing system 40, etc. Then, the analysis system 50 of the first embodiment performs access analysis on the collected behavioral data to generate and accumulate address information indicating the location of information resources present on the network N. The analysis system 50 returns the address information in response to a request from the data analysis system 100. The analysis system 50 is configured by a cloud server based on cloud computing, a physical server, or the like.

[0017] The terminal device 80 is a mobile terminal or a PC used by a user. The terminal device 80 accesses and logs into a website in response to a user's operation, displays various information on the website, and relays buying and selling transactions. The user uses one or more terminal devices 80 to access the service providing system 40 via the network N.

[0018] The data analysis system 100 acquires address information indicating the location of a web page as page information linked to the web page on a website, performs a classification process on the acquired address information, and then performs a predetermined analysis process. The data analysis system 100 is configured with a cloud server based on cloud computing, a physical server, or a system that combines these. The data analysis system 100 may be configured with a PC. Hereinafter, a website is also referred to as a "site", and a web page is also referred to as a "page".

[0019] Next, the functional configuration of the management device 30 will be described with reference to Fig. 2. The management device 30 has a management communication unit 31, a management control unit 32, a management storage unit 33, an input unit 34, and a display unit 35. The management storage unit 33 stores various information in addition to the operating program of the management control unit 32. The management storage unit 33 is composed of a RAM (Random Access Memory) and a ROM (Read Only Memory), a PROM (Programmable ROM) such as a flash memory, or an HDD (Hard Disk Drive), etc.

[0020] The input unit 34 includes, for example, a keyboard and a pointing device such as a mouse or a trackball. The input unit 34 accepts an input operation by a user and transmits an operation signal according to the content of the input operation to the management control unit 32. The display unit 35 includes, for example, a liquid crystal display (LCD), and displays various information according to instructions from the management control unit 32.

[0021] The management control unit 32 has an information processing means 32a and a display processing means 32b. When the information processing means 32a accepts an operation related to display on the display unit 35, it outputs a control signal corresponding to the operation to the display processing means 32b. The display processing means 32b displays characters, images, and the like on the display unit 35 in response to the control signal from the information processing means 32a.

[0022] The management control unit 32 can be configured by a calculation device such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit) and an operation program that cooperates with the calculation device to realize the above-mentioned various functions. Note that the management device 30 may have a touch panel including a display panel that displays characters or images, and a detection means that is stacked on the display panel and detects touch operations, instead of the input unit 34 and the display unit 35. The management device 30 may have both a touch panel and a mouse or a keyboard.

[0023] Next, the functional configuration of the terminal device 80 will be described with reference to Fig. 3. The terminal device 80 has a terminal communication unit 81, a terminal control unit 82, a terminal storage unit 83, an input unit 84, and a display unit 85. The terminal storage unit 83 stores various information in addition to the operation program of the terminal control unit 82. The terminal storage unit 83 is composed of RAM and ROM, a PROM such as a flash memory, an HDD, or the like.

[0024] The input unit 84 includes, for example, a keyboard and a pointing device such as a mouse or a trackball. The input unit 84 accepts an input operation by a user and transmits an operation signal according to the content of the input operation to the terminal control unit 82. The display unit 85 includes, for example, a liquid crystal display, and displays various information according to instructions from the terminal control unit 82.

[0025] The terminal control unit 82 has an information processing means 82a and a display processing means 82b. When the information processing means 82a accepts an operation related to display on the display unit 85, it outputs a control signal corresponding to the operation to the display processing means 82b. The display processing means 82b displays characters, images, and the like on the display unit 85 in response to the control signal from the information processing means 82a.

[0026] The terminal control unit 82 can be configured by a calculation device such as a CPU or GPU, and an operation program that cooperates with the calculation device to realize the above-mentioned various functions. The terminal device 80 may have a touch panel including a display panel that displays characters or images, and a detection means that is stacked on the display panel and detects touch operations, instead of the input unit 84 and the display unit 85. The terminal device 80 may have both a touch panel and a mouse or a keyboard.

[0027] Next, the functional configuration of the data analysis system 100 will be described with reference to Fig. 4. As shown in Fig. 4, the classification device 10 classifies page information linked to pages on a site. The classification device 10 has a communication unit 11, a control unit 12, and a classification storage unit 13. The communication unit 11 is an interface that allows the control unit 12 and the analysis processing unit 22 to communicate with devices connected to the network N via wired or wireless communication.

[0028] The classification storage unit 13 stores various data related to the classification process of page information, in addition to the operation program of the control unit 12 including the classification process program P1. For example, the classification storage unit 13 stores page information reflecting the browsing behavior of a user. The classification storage unit 13 stores an address classification model M1 for classifying (grouping) page information.

[0029] The control unit 12 has a collection processing means 12a, a learning processing means 12c, and a classification processing means 12d. The collection processing means 12a periodically transmits a request signal requesting transmission of page information to the analysis system 50, and stores a plurality of pieces of page information returned from the analysis system 50 in the classification storage unit 13. In the first embodiment, the collection processing means 12a sequentially or periodically collects address information indicating the position of a page in a site as page information, and stores the address information in the classification storage unit 13. Here, the information resources include pages in the site, and the address information corresponds to a URL (Uniform Resource Locator).

[0030] The learning processing means 12c generates an address classification model M1 that performs classification settings for page information by machine learning using a plurality of page information stored in the classification storage unit 13. That is, the learning processing means 12c constructs an address classification model M1 that performs grouping based on the similarity of the properties of page information by machine learning using a plurality of past page information as input. The address classification model M1 is a trained model that classifies input page information into one of a plurality of groups and assigns a classification tag, which is a unique identification tag, to the group to which the page information is classified. A group of page information is a group to which page information showing similar properties belongs. The similarity of the properties of page information corresponds to the similarity of the properties of the pages. That is, the groups of page information include a group of product details pages, a group of company information pages, a group of information request pages, a group of product details pages, and the like.

[0031] Here, the collection processing means 12a is configured to collect a plurality of pieces of page information over time and store them in the classification storage unit 13. Then, the learning processing means 12c acquires a plurality of pieces of page information newly stored in the classification storage unit 13 at a set timing, and executes an update process of the address classification model M1 using the acquired plurality of pieces of page information as an input.

[0032] In the first embodiment, the learning processing means 12c generates the address classification model M1 by supervised learning using a DNN (Deep Neural Network). The learning processing means 12c may generate the address classification model M1 by unsupervised learning or semi-supervised learning. The learning processing means 12c may generate the address classification model M1 by machine learning using a GBDT (Gradient Boosting Decision Tree). The learning processing means 12c may generate the address classification model M1 by machine learning using a character-level CNN (Character-level Convolutional Neural Network). In addition, the learning processing means 12c may generate the address classification model M1 by a regression method such as linear regression, logistic regression, or decision tree. Furthermore, the learning processing means 12c may generate the address classification model M1 by a classification method such as a random forest or a support vector machine. However, the learning processing means 12c may generate the address classification model M1 by a method that combines the above-mentioned multiple machine learning methods.

[0033] The classification processing means 12d inputs page information associated with a page in a site to an address classification model M1 generated by machine learning using a plurality of pieces of page information from the past. The classification processing means 12d then classifies the page information input to the address classification model M1 into one of a plurality of groups based on the similarity of properties. More specifically, the classification processing means 12d sequentially inputs a plurality of pieces of address information to be classified into the address classification model M1. The classification processing means 12d then classifies the input address information into one of a plurality of groups, and generates classification data by attaching the classification tag of the group to the address information. That is, the classification processing means 12d attaches a common classification tag to page information showing similar properties, and attaches different classification tags to page information showing different properties. The classification processing means 12d may store the classification data generated by attaching the classification tag to the address information in the analysis storage unit 23, or may output it to the analysis processing unit 22. The classification data in the first embodiment is also called "address classification data".

[0034] (How to get address information) Here, the acquisition pattern of address information as page information by the classification device 10 will be described. When the classification device 10 is linked with a site provided by the service providing system 40 via the analysis system 50 (for example, GA linkage with the site: hereinafter referred to as system linkage), the classification device 10 can use the API (Application Programming Interface) of the analysis system 50 to enumerate and acquire the URLs of pages included in the site. At that time, the classification device 10 can also acquire information such as the number of PVs (number of page views) of each page, and can use such information for grouping.

[0035] On the other hand, even if the systems are not linked, the classification device 10 can directly collect the URLs of the pages by, for example, the following two methods. (1) Recursively crawling pages on a site: For example, address information can be collected by crawling recursively, such as crawling the top page of a site first, collecting links within the site contained therein, and crawling those links as well. When adopting such a method, the collection processing means 12a may be provided with a crawling function for recursively crawling pages within the site, or the collection processing means 12a may collect address information after crawling from an external server having a crawling function. (2) Using a sitemap: Some sites may have URLs of pages in the site listed in an XML sitemap (sitemap.xml), in other words, in such a case, the collection processing means 12a can collect address information from the XML sitemap.

[0036] In this way, even if the classification device 10 is not linked to the system, it can obtain data that is publicly available on the Internet. However, in this case, it cannot obtain information such as the number of page views of each page. In other words, the classification device 10 cannot use information such as the number of page views of each page in the learning process by the learning process means 12c and the classification process by the classification process means 12d. Therefore, from the viewpoint of the accuracy of grouping, it is preferable that the system is linked.

[0037] 4, the analysis device 20 performs analysis processing on the classified data obtained by classification by the classification device 10. The analysis device 20 has an analysis processing unit 22 and an analysis memory unit 23. The analysis memory unit 23 stores various data related to the analysis processing of the classified data, in addition to the operation program of the analysis processing unit 22 including the analysis processing program 23p. The analysis processing unit 22 executes various analysis processing such as an entry analysis processing on the classified data generated by the classification processing means 12d.

[0038] The entrance analysis process is a process of comparing the CVR (Conversion Rate) between entrance pages at the same inflow source and evaluating the importance of the entrance page. An entrance page is the first page viewed by a user who visits the site to be analyzed. An entrance source is the source through which a user visits the entrance page of the site to be analyzed. The inflow source shows the approximate characteristics of a user. The inflow sources include via e-mail, via Facebook (registered trademark), via search engines, via links on other sites, via bookmarks in a browser, and directly typing in a URL. CVR is an index that indicates the degree to which site visitors take actions that contribute to achieving the site's goals, and is also called the conversion rate.

[0039] The analysis processing unit 22 can also perform entrance analysis processing on a group basis for entrance pages. For example, the analysis processing unit 22 distinguishes between entrance pages with relatively high CVR and entrance pages with relatively low CVR within the same group of inflow sources. Then, when a user can set a link to an entrance page, the analysis processing unit 22 transmits recommendation information recommending guidance to an entrance page with a higher CVR to the management device 30. The analysis processing unit 22 may transmit improvement information indicating the CVR growth margin when the settings are changed according to the recommendation information to the management device 30 together with the recommendation information.

[0040] On the other hand, if the user cannot set a link to the entry page (for example, if the entry page is determined by an external site), the analysis processing unit 22 transmits proposal information relating to a proposal to bring an entry page with a relatively low CVR closer to the form of an entry page with a relatively high CVR to the management device 30. The analysis processing unit 22 may transmit improvement information indicating the CVR growth potential and the like when the settings are changed according to the proposal information to the management device 30 together with the proposal information.

[0041] The display processing means 32b of the management device 30 causes information based on the recommended information or proposed information to be displayed on the display unit 35. This allows the administrator or the like to take measures to improve the CVR or the like. When improvement information is transmitted from the analysis processing unit 22, the display processing means 32b may cause the display unit 35 to display information based on the improvement information.

[0042] The classification storage unit 13 and the analysis storage unit 23 can be configured with RAM and ROM, PROM such as flash memory, or HDD. The control unit 12 can be configured with a calculation device such as a CPU or GPU, and a classification processing program P1 that cooperates with the calculation device to realize the various functions described above. That is, the classification processing program P1 is a program that causes the control unit 12 as a computer and the classification storage unit 13 to function as the collection processing means 12a, the learning processing means 12c, and the classification processing means 12d. Similarly, the analysis processing unit 22 can be configured with a calculation device such as a CPU or GPU, and an analysis processing program 23p that cooperates with the calculation device to realize the various functions described above.

[0043] In FIG. 2, in order to clearly separate and explain each function, the classification memory unit 13 and the analysis memory unit 23 are shown as separate configurations, and the control unit 12 and the analysis processing unit 22 are shown as separate configurations, but this is not limited thereto. For example, the control unit 12 and the analysis processing unit 22 may be configured as a single arithmetic device, or may be formed as a single chip as a SoC (System-on-a-Chip). Similarly, the classification memory unit 13 and the analysis memory unit 23 may be integrated into one configuration. The same applies to the following figures and their corresponding explanations.

[0044] Next, a classification processing method by the classification device 10 will be described with reference to Fig. 5. Here, an example of the classification processing step performed by the classification processing means 12d will be described. That is, in the classification processing method, the learning processing step performed by the learning processing means 12c is a premise for the processing in Fig. 5.

[0045] First, the classification processing means 12d acquires page information stored in the classification storage unit 13 as a classification target (step S101). Next, the classification processing means 12d sets the acquired page information as an input of the address classification model M1 (step S102). Then, the classification processing means 12d generates classification data by attaching an identification tag of the group to which the input page information belongs to the page information using the address classification model M1, and stores the generated classification data in the classification storage unit 13 or the analysis storage unit 23 (step S103).

[0046] The classification processing means 12d judges whether or not the page information to be classified remains in the classification storage unit 13 (step S104). If the page information to be classified exists (step S104 / Yes), the classification processing means 12d proceeds to the process of step S101. If the page information to be classified does not exist (step S104 / No), the classification processing means 12d ends the classification process.

[0047] Here, a supplementary explanation will be given regarding the outline of the classification processing method in the present embodiment 1. The classification processing method includes a learning processing step of generating an address classification model M1 by machine learning using a plurality of pieces of page information from the past, and a classification processing step of inputting page information to be classified into the address classification model M1 and classifying the input page information into one of a plurality of groups based on the similarity of properties.

[0048] As described above, the classification device 10 of the first embodiment has a classification processing means 12d that classifies the input page information into one of a plurality of groups based on the similarity of properties by inputting page information linked to a page in a site into the address classification model M1. The classification processing means 12d attaches an identification tag of the group to which the page information is to be classified to the page information input into the address classification model M1. Therefore, the grouping of page information linked to the page to be analyzed can be performed automatically and efficiently, thereby making it possible to speed up and streamline the process.

[0049] Moreover, the page information in the present embodiment 1 is information indicating the location of information resources present on the network N, and is address information equivalent to a URL. Here, the address information can be easily and quickly acquired, particularly in cases where systems are linked, or where the URLs of pages within a site are listed in an XML site map (sitemap.xml). Due to the data configuration, the classification process of the address information can be performed in a relatively short time. Note that the control unit 12 does not need to have the learning processing means 12c. In this case, the classification device 10 may store an address classification model M1 constructed externally in the classification storage unit 13, and have the classification processing means 12d use this.

[0050] Embodiment 2 6 and 7, a configuration example of a data analysis system 200 and its peripheral devices in the second embodiment will be described. The same configurations and components as those in the first embodiment will be denoted by the same reference numerals and the description will be omitted.

[0051] As shown in Fig. 6, the data analysis system 200 of the second embodiment communicates with the crawling system 60 via a network N. The crawling system 60 of the second embodiment has a function of acquiring text data of a page by crawling the page in the site. When the site is composed of a plurality of pages, the crawling system 60 acquires text data for each page.

[0052] 7, the control unit 112 of the classification device 110 in the second embodiment has a collection processing means 112a, a preprocessing means 112b, a learning processing means 112c, and a classification processing means 112d. The collection processing means 112a sequentially or periodically collects text data, which is information on text within pages of a site, from the crawling system 60, and stores the collected text data in the classification storage unit 13.

[0053] The preprocessing means 112b performs an analysis process on text data, which is information on text within a page on a site, to generate analysis data as page information. The preprocessing means 112b in the fourth embodiment performs a morpheme analysis process on the text data stored in the classification storage unit 13 to generate analysis data as page information. Morphological analysis is to divide text data written in a natural language into morphemes, which are the smallest units in language, and to determine the parts of speech and changes of each morpheme. That is, the preprocessing means 112b stores in the classification storage unit 13 analysis data that has been organized by attaching information on parts of speech, such as verbs, nouns, and adjectives, to each of the morphemes divided from the text data. When attaching and organizing information on parts of speech to each morpheme, the preprocessing means 112b may perform processes such as conversion to vector representation and weighting based on frequently used words registered in advance.

[0054] The learning processing means 112c generates a text classification model M2 that performs classification setting for analysis data by machine learning using a plurality of analysis data stored in the classification storage unit 13. That is, the learning processing means 112c constructs a text classification model M2 that performs grouping based on the similarity of the properties of analysis data by machine learning using a plurality of past analysis data as input. The learning processing means 112c executes an update process for the text classification model M2, similar to the above-mentioned learning processing means 12c. The learning processing means 112c can generate the text classification model M2 by various machine learning methods such as learning using DNN, or by a method combining a plurality of machine learning methods, similar to the learning processing means 12c of the first embodiment.

[0055] The classification processing means 112d inputs the analysis data to the text classification model M2 and classifies the input analysis data into one of a plurality of groups based on the similarity of properties. That is, the classification processing means 112d sequentially inputs a plurality of analysis data to be classified into the text classification model M2. The classification processing means 112d then classifies the input analysis data into one of a plurality of groups and generates classified data by attaching a classification tag of the group to which the data is classified. The classified data of the second embodiment is also referred to as "text classified data".

[0056] The classification storage unit 13 of the second embodiment stores various data related to classification processing of the analysis data as page information, in addition to the operation program of the control unit 112 including the classification processing program P2. The classification storage unit 13 also stores a text classification model M2 for classifying the analysis data. The classification processing program P2 is a program for causing the control unit 112 as a computer and the classification storage unit 13 to function as the collection processing means 112a, the preprocessing means 112b, the learning processing means 112c, and the classification processing means 112d. Other configurations and alternative configurations are the same as those of the first embodiment described above. In addition, among the classification processing method by the classification device 110, the learning processing step performed by the learning processing means 112c and the classification processing step performed by the classification processing means 112d are the same as those described with reference to FIG. 5.

[0057] As described above, the classification device 110 of the second embodiment has a classification processing means 112d that inputs the analysis data as page information to the text classification model M2 and classifies the input analysis data into one of a plurality of groups. Therefore, the page information associated with the page to be analyzed can be automatically and efficiently grouped, and therefore the processing can be speeded up and made more efficient. Moreover, the page information of the second embodiment is text data, which is information on the text in the webpage, and more information useful for grouping the webpages can be obtained from the text data than from the address information. Therefore, the grouping of the webpages can be performed with higher accuracy. Other effects and the like are the same as those of the first embodiment described above.

[0058] Incidentally, in the above description based on FIG. 7, an example in which the control unit 112 has the preprocessing means 112b has been described, but the control unit 112 may be configured not to have the preprocessing means 112b. The learning processing means 112c may construct the text classification model M2 by machine learning using a plurality of past text data as input. In this case, the learning processing means 112c generates the text classification model M2 that performs grouping based on the similarity of the properties of the text data by machine learning using, for example, character-level CNN. Therefore, the classification processing means 112d sequentially inputs a plurality of text data to be classified into the text classification model M2. Then, the classification processing means 112d classifies the input text data into one of a plurality of groups, and generates classified data by attaching a classification tag of the group to which the data is classified.

[0059] The preprocessing means 112b may also generate analysis data by deleting unnecessary data from the text data. The unnecessary data is, for example, characters such as numbers or line breaks that are not necessary for training the text classification model M2 and prediction by the model. In this case, the learning processing means 112c constructs the text classification model M2 by machine learning using multiple past analysis data as input. The classification processing means 112d sequentially inputs multiple analysis data to be classified into the text classification model M2, classifies the input analysis data into one of multiple groups, and generates classified data by attaching a classification tag of the group to which the data is classified.

[0060] Embodiment 3 A configuration example of a data analysis system 300 and its peripheral devices in the third embodiment will be described with reference to Fig. 6 and Fig. 8 to Fig. 10. The same configurations and components as those in the first and second embodiments are denoted by the same reference numerals and will not be described.

[0061] The crawling system 60 of the third embodiment has a function of crawling pages in a site to acquire display-related data including information on each object in the page. When a site is composed of multiple pages, the crawling system 60 acquires display-related data for each page. Here, an object is some kind of entity that is the subject of an operation or process on a computer. Objects include data sets and operation elements displayed in the form of icons or the like on a display screen, and individual figures arranged on a display screen. For example, an object may be an element of a Document Object Model (DOM).

[0062] 8, the control unit 212 of the classification device 210 in the third embodiment has a collection processing means 212a, a preprocessing means 212b, a learning processing means 212c, and a classification processing means 212d. The collection processing means 212a sequentially or periodically collects display-related data, which is information on the contents of pages on a site and includes information on each object on the page, from the crawling system 60, and stores the collected display-related data in the classification storage unit 13.

[0063] The pre-processing means 212b generates identification information for each object from the display-related data including information on each object on a page of the site by combining type information indicating the type of the object with layout information indicating its position and size. That is, the pre-processing means 212b generates identification information for each object in the display-related data by combining type information and layout information. For example, if there are 26 types of object types and 1024 types of layout information indicating their positions and sizes for objects on a web page, there will be 26624 (26×1024) types of object identification information. However, the types of objects may be set to 27 or more or 25 or less, and the layout information may be set to 1025 or more or 1023 or less.

[0064] Then, the preprocessing means 212b generates frame features (wireframe features) as page information based on the generated identification information for each object. As an example, the preprocessing means 212b converts each generated identification information into a vector representation such as BoW (Bag-of-Words) and reduces the dimension of the converted data using, for example, LDA (Latent Dirichlet Allocation), which is a type of topic model, to generate frame features. However, the preprocessing means 212b is not limited to the above, and may employ, for example, unsupervised learning such as an autoencoder as a method for reducing the dimension of the data converted into a vector representation.

[0065] Here, the frame feature generation process will be conceptually explained with reference to Fig. 9 and Fig. 10. For example, the display screen 85x in Fig. 9 illustrates an object 91a of a photo as a background, an object 91b of a setting in which a plurality of photos are switched in a slideshow format, and an object 91c of a photo with a link. The display screen 85x also illustrates an object 92 of simple characters, an object 93 of linked characters or logos, an object 94 of a plurality of linked characters arranged, and an object 95 of a plurality of simple characters arranged. In addition, the display screen 85x illustrates an object 96 in which a plurality of characters are arranged in a table format, and an object 97 of a bar including characters.

[0066] In Fig. 9, the object in the photo is shaded, the object to which the link is attached is shown as a rectangle, and the object to which the link is attached is shown as a rounded rectangle. Object 94 is a so-called menu bar, etc. In Fig. 9, the object 96 that is colored is designated as object 96a. There are various types of objects other than those described above, and the object types may be set more finely or more roughly.

[0067] When only the parts information corresponding to the type of each object is extracted as an image from a web page such as that shown in Fig. 9, a wireframe image such as that shown in Fig. 10 is obtained. The parts information is information including at least one tag included in the source code of the web page, and corresponds to type information indicating the type of the object. That is, the preprocessing means 212b converts the web page into a wireframe image, and then generates frame features, which are embedded representations (vector representations) of fixed-length vectors, from the wireframe image as page information.

[0068] The learning processing means 212c generates a page classification model M3 that performs classification settings for frame features by machine learning using a plurality of frame features stored in the classification storage unit 13. That is, the learning processing means 212c constructs a page classification model M3 that performs grouping based on the similarity of the properties of the frame features by machine learning using a plurality of past frame features as input. The learning processing means 212c executes an update process for the page classification model M3, similar to the above-mentioned learning processing means 12c. The learning processing means 212c can generate the page classification model M3 by various machine learning methods such as learning using DNN, or by a method that combines a plurality of machine learning methods, similar to the learning processing means 12c in the first embodiment.

[0069] The classification processing means 212d inputs the frame features to a page classification model M3 generated by machine learning using multiple past frame features, and classifies the input frame features into one of multiple groups based on the similarity of properties. That is, the classification processing means 212d sequentially inputs multiple frame features to be classified into the page classification model M3. The classification processing means 212d then classifies the input frame features into one of multiple groups, and generates classification data by attaching a classification tag of the group to which the frame features are to be classified. The classification data of the third embodiment is also referred to as "page classification data".

[0070] The classification storage unit 13 of the third embodiment stores various data related to classification processing of frame features as page information, in addition to the operation program of the control unit 212 including the classification processing program P3. The classification storage unit 13 also stores a page classification model M3 for classifying frame features. The classification processing program P3 is a program for causing the control unit 212 as a computer and the classification storage unit 13 to function as the collection processing means 212a, the preprocessing means 212b, the learning processing means 212c, and the classification processing means 212d. Other configurations and alternative configurations are the same as those of the first and second embodiments described above. In addition, among the classification processing method by the classification device 210, the learning processing step performed by the learning processing means 212c and the classification processing step performed by the classification processing means 212d are the same as those described with reference to FIG. 5.

[0071] As described above, the classification device 210 of the third embodiment has the classification processing means 212d that inputs frame features as page information to the page classification model M3 and classifies the input frame features into one of a plurality of groups. Therefore, the grouping of page information linked to the page to be analyzed can be performed automatically and efficiently, and therefore the processing can be speeded up and made more efficient.

[0072] Incidentally, display-related data indicating the contents of a web page provides more information useful for grouping web pages than address information. However, since web pages are generally variable length, it is difficult to use display-related data for grouping as is. This is because machine learning methods that use images as input generally have a fixed input image size, and few methods support variable-length images. In addition, if an image is forcibly reduced to a fixed size, information that is essentially necessary for the classification process may be lost, and the performance of the classification process may deteriorate. In other words, for these reasons, display-related data has not been used in page grouping so far.

[0073] In this regard, the preprocessing means 212b of the third embodiment generates identification information for each object by combining type information and arrangement information from the display-related data corresponding to the variable-length webpage to be analyzed. Then, the preprocessing means 212b generates frame features based on the generated identification information for each object. That is, the preprocessing means 212b has a function of converting the display-related data into a fixed-length vector representation so that the data can be used as input data for machine learning. That is, the preprocessing means 212b generates frame features by performing dimensional compression (dimension reduction) on the display-related data corresponding to the variable-length webpage into a vector representation. Therefore, it is possible to further improve the accuracy of grouping of webpages.

[0074] Embodiment 4 A configuration example of a data analysis system 400 and its peripheral devices in the fourth embodiment will be described with reference to Fig. 11. The same configurations and components as those in the first to third embodiments described above will be denoted by the same reference numerals and description thereof will be omitted.

[0075] 11, the control unit 312 of the classification device 310 in the fourth embodiment has a collection processing means 312a, a preprocessing means 312b, a learning processing means 312c, and a classification processing means 312d. The collection processing means 312a sequentially or periodically collects address information, text data, and display related data as page information, and stores the collected information in the classification storage unit 13.

[0076] The pre-processing means 312b has a text pre-processing means B2 and a page pre-processing means B3. The text pre-processing means B2 functions in the same manner as the pre-processing means 112b of the above-mentioned embodiment 2. That is, the text pre-processing means B2 performs an analysis process on the text data to generate analysis data as page information. The page pre-processing means B3 functions in the same manner as the pre-processing means 212b of the above-mentioned embodiment 3. That is, the page pre-processing means B3 generates identification information for each object by combining type information and layout information from the display-related data corresponding to the page to be classified, and generates frame features as page information based on the generated identification information for each object.

[0077] The learning processing means 312c has an address learning means C1, a text learning means C2, a page learning means C3, and a general learning means Cx. The address learning means C1 generates and updates a learned model in the same manner as the learning processing means 12c of the above-mentioned first embodiment. That is, the address learning means C1 constructs an address classification model M10 that extracts and outputs address features from the input address information by machine learning using a plurality of address information in the past as input. The address features may be address classification data in which a classification tag is attached to the address information, or may be information in a form different from the address classification data.

[0078] The address learning means C1 may extract, for example, an embedded representation of an output vector of a hidden layer (middle layer) of the DNN as an address feature in a form different from the address classification data, and have the general learning means Cx use this. In this case, the address learning means C1 extracts an address feature in an embedded representation of an output vector of a hidden layer of the DNN in the address classification model M10, one layer before the output layer, etc. In addition, if a SoftMax function, which is a type of activation function, is adopted in the output layer of the DNN in the address classification model M10, the probability for each group is output by the SoftMax function. Therefore, in this case, the address learning means C1 may extract the probability vector of each group as an address feature in a form different from the address classification data.

[0079] Furthermore, in the case where the address learning means C1 is configured to use a character-level CNN, for example, the address learning means C1 may extract an embedded representation of the output vector of the hidden layer of the character-level CNN as an address feature quantity in a form different from the address classification data. In addition, in the case where a SoftMax function is adopted in the output layer of the character-level CNN in the address classification model M10, the address learning means C1 may extract the probability vector of each group as an address feature quantity in a form different from the address classification data.

[0080] The text learning means C2 generates and updates a trained model in the same manner as the learning processing means 112c of the above-mentioned second embodiment. That is, the text learning means C2 constructs a text classification model M20 that extracts and outputs text features from the input analysis data by machine learning using a plurality of past analysis data as input. The text features may be text classification data in which classification tags are added to the analysis data, or may be information of an intermediate expression such as an embedded expression (vector expression) during the process of generating the text classification data.

[0081] The page learning means C3 generates and updates a trained model in the same manner as the learning processing means 212c of the above-mentioned third embodiment. That is, the page learning means C3 constructs a page classification model M30 that extracts and outputs page features from the input frame features by machine learning using a plurality of past frame features as input. The page features may be page classification data in which classification tags are attached to the frame features, or may be information of an intermediate expression such as an embedded expression (vector expression) during the process of generating the page classification data.

[0082] The general learning means Cx generates a general classification model M4 that classifies web pages corresponding to each piece of input information by machine learning using the address features generated by the address classification means D1, the text features generated by the text classification means D2, and the page features generated by the page classification means D3 as input. In the fourth embodiment, the general learning means Cx generates the general classification model M4 by machine learning using the GBDT.

[0083] The classification processing means 312d has an address classification means D1, a text classification means D2, a page classification means D3, and a general classification means Dx. The address classification means D1 executes classification processing in the same manner as the classification processing means 12d of the above-mentioned embodiment 1. That is, the address classification means D1 extracts address features using address information corresponding to a page to be classified as an input of an address classification model M10.

[0084] The text classification means D2 executes classification processing in the same manner as the classification processing means 112d of the above-mentioned embodiment 2. That is, the text classification means D2 extracts text features using analysis data corresponding to the page to be classified as an input to the text classification model M20. The page classification means D3 executes classification processing in the same manner as the classification processing means 212d of the above-mentioned embodiment 3. That is, the page classification means D3 extracts page features using frame features corresponding to the page to be classified as an input to the page classification model M30.

[0085] The general classification means Dx groups each webpage using the address feature, text feature, and page feature of each webpage. More specifically, the general classification means Dx inputs the address feature, text feature, and page feature associated with the webpage to be classified to the general classification model M4, and classifies the webpage into one of a plurality of groups based on the similarity of properties. That is, the general classification means Dx inputs each feature associated with the plurality of webpages to be classified to the general classification model M4 in sequence. Then, the general classification means Dx classifies the webpage associated with each input feature into one of a plurality of groups, and generates classification data by attaching the classification tag of the group to be classified to information indicating the webpage. The classification data of the fourth embodiment is also called "general classification data".

[0086] Incidentally, since the text classification means D2 and the page classification means D3 use data that has been crawled and preprocessed, the processing time per page by the text classification means D2 or the page classification means D3 is longer than the processing time per page by the address classification means D1. In addition, since the processing contents performed by the address classification means D1, the text classification means D2, and the page classification means D3 are different from each other, the time required for classification processing for the same page is also different. The time required for classification processing per page is generally longest for the address classification means D1, followed by the text classification means D2 and the page classification means D3. Therefore, if the overall classification processing is performed in accordance with the processing speed of the address classification means D1, the data corresponding to the delay in the text classification means D2 and the page classification means D3 becomes unprocessed missing data (missing values).

[0087] Therefore, in the fourth embodiment, a conditional branch of "missing data or not?" is set in a plurality of decision trees included as weak learners in the GBDT-based overall classification model M4. Therefore, the overall classification means Dx can also use the missing data in the text classification means D2 and the page classification means D3 as input to the overall classification model M4.

[0088] The classification storage unit 13 of the fourth embodiment stores various data related to classification processing of frame features as page information, in addition to the operation program of the control unit 312 including the classification processing program P4. The classification storage unit 13 also stores an address classification model M10 for extracting address features from address information, a text classification model M20 for extracting text features from analysis data, etc., and a page classification model M30 for extracting page features from frame features. The classification storage unit 13 also stores a general classification model M4 that generates classification data using address features, text features, and page features as inputs. The classification processing program P4 is a program for causing the control unit 312 and the classification storage unit 13 as a computer to function as a collection processing means 312a, a preprocessing means 312b, a learning processing means 312c, and a classification processing means 312d. Other configurations and alternative configurations are the same as those of the first to third embodiments described above. In addition, in the classification method by the classification device 310, the basic flow of the learning process performed by the learning process means 312c and the classification process performed by the classification process means 312d is the same as that explained with reference to FIG.

[0089] Next, each step of the classification processing method in the fourth embodiment will be described with reference to Fig. 12. Here, the learning processing step of generating each classification model by the learning processing means 312c will be omitted. Note that each step in Fig. 12 is performed in parallel.

[0090] [Collection and processing process] 12, the collection processing means 312a sequentially or periodically collects address information, text data, and display-related data from the outside, and stores each collected data in the classification storage unit 13. At that time, the collection processing means 312a stores in the classification storage unit 13 the address information, text data, and display-related data corresponding to the same web page in association with each other.

[0091] [Pretreatment process] The pre-processing means 312b generates analyzed data based on a morphological analysis of the text data, or generates analyzed data by deleting unnecessary data from the text data (text pre-processing step). The pre-processing means 312b generates identification information for each object by combining type information and placement information from the display-related data corresponding to the page to be classified, and generates frame features based on the generated identification information for each object (page pre-processing step).

[0092] [Classification process] The classification processing means 312d extracts address features using address information corresponding to the page to be classified by the address classification means D1 as an input to the address classification model M10 (address classification step). The classification processing means 312d uses the analysis data corresponding to the page to be classified by the text classification means D2 to input the text classification model M20 to extract text features (text classification step). The classification processing means 312d uses the frame feature corresponding to the page to be classified by the page classification means D3 to extract page feature (page classification step) using the frame feature corresponding to the page to be classified as an input to the page classification model M30.

[0093] The classification processing means 312d sequentially inputs address features, text features, and page features associated with the multiple web pages to be classified into the general classification model M4. The classification processing means 312d then classifies the web pages associated with each input feature into one of multiple groups, and generates classification data by attaching the classification tag of the group to information indicating the web pages (general classification process).

[0094] As described above, the classification device 310 of the fourth embodiment can also automatically and efficiently group the web pages to be analyzed, so that the processing can be made quicker and more efficient. The classification device 310 has an address classification means D1 that extracts address features using address information as an input to the address classification model M10, a text classification means D2 that extracts text features using analysis data as an input to the text classification model M20, and a page classification means D3 that extracts page features using frame features as an input to the page classification model M30. The classification device 310 also has a general classification means Dx that groups each web page using the address features, text features, and page features of each web page. That is, the classification device 310 can reflect information that cannot be obtained by a configuration using one piece of browsing data in the machine learning and classification processing by a two-stage classification processing using three pieces of information, namely, address information, text data, and display-related data. Therefore, the accuracy of grouping of web pages can be improved.

[0095] In the above description, an example has been shown in which the general learning means Cx generates the general classification model M4 by machine learning using GBDT, but this is not limiting. The general learning means Cx may generate the general classification model M4 by various machine learning methods such as learning using DNN, or by a method combining multiple machine learning methods.

[0096] Incidentally, the classification device 310 may be configured without the text preprocessing means B2. In this case, the text learning means C2 constructs a text classification model M20 that extracts text features from input text data by machine learning using a plurality of past text data as input. That is, the text classification means D2 generates text features using text data corresponding to the page to be classified as input to the text classification model M20.

[0097] <Variation 4a> With reference to FIG. 13, a configuration example of a data analysis system 400A in a modification 4a of the fourth embodiment will be described. The control unit 312 in this modification 4a has a collection processing means 312a, a text pre-processing means B2, a learning processing means 312c, and a classification processing means 312d. The learning processing means 312c in this modification 4a has an address learning means C1, a text learning means C2, and a general learning means Cxa. The classification processing means 312d in this modification 4a has an address classification means D1, a text classification means D2, and a general classification means Dxa. The collection processing means 312a in this modification 4a sequentially or periodically collects address information and text data as page information, and stores each of the collected information in the classification storage unit 13.

[0098] The general learning means Cxa generates a general classification model M41 that classifies web pages corresponding to each input feature by machine learning using the address features generated by the address classification means D1 and the text features generated by the text classification means D2 as input. The general classification means Dxa sequentially inputs the address features and text features linked to the multiple web pages to be classified into the general classification model M41. The general classification means Dxa then classifies the web pages linked to each input feature into one of multiple groups, and generates classification data by attaching the classification tag of the group to information indicating the web pages.

[0099] As described above, the classification device 310 of the present modification 4a can reflect information that cannot be obtained by a configuration using one piece of browsing data in the machine learning and classification processing by performing a two-stage classification process using two pieces of information, namely address information and text data. Therefore, it is possible to improve the accuracy of grouping of web pages. Meanwhile, the classification device 310 may be configured without the text preprocessing means B2. In this case, the text learning means C2 constructs a text classification model M20 that extracts text features from input text data by machine learning using a plurality of past text data as input.

[0100] <Variation 4b> With reference to FIG. 14, a configuration example of a data analysis system 400B in a modification 4b of the fourth embodiment will be described. The control unit 312 in this modification 4b has a collection processing means 312a, a page pre-processing means B3, a learning processing means 312c, and a classification processing means 312d. The learning processing means 312c in this modification 4b has an address learning means C1, a page learning means C3, and a general learning means Cxb. The classification processing means 312d in this modification 4b has an address classification means D1, a page classification means D3, and a general classification means Dxb. The collection processing means 312a in this modification 4b sequentially or periodically collects address information and display related data as page information, and stores each of the collected information in the classification storage unit 13.

[0101] The general learning means Cxb uses machine learning to input the address features generated by the address classification means D1 and the page features generated by the page classification means D3, and generates a general classification model M42 that classifies web pages corresponding to each of the input features. The general classification means Dxb sequentially inputs the address features and page features associated with the multiple web pages to be classified into the general classification model M42. The general classification means Dxb then classifies the web pages associated with each of the input features into one of multiple groups, and generates classification data by attaching the classification tag of the group to information indicating the web pages.

[0102] As described above, the classification device 310 of the present modification 4b can reflect information that cannot be obtained by a configuration that uses one piece of browsing data in the machine learning and classification process by performing a two-stage classification process using two pieces of information, namely, address information and display-related data, thereby improving the accuracy of grouping of web pages.

[0103] <Variation 4c> With reference to FIG. 15, a configuration example of a data analysis system 400C in a modification 4c of the fourth embodiment will be described. The control unit 312 in this modification 4c has a collection processing means 312a, a pre-processing means 312b, a learning processing means 312c, and a classification processing means 312d. The learning processing means 312c in this modification 4c has a text learning means C2, a page learning means C3, and a general learning means Cxc. The classification processing means 312d in this modification 4b has a text classification means D2, a page classification means D3, and a general classification means Dxc. The collection processing means 312a in this modification 4c collects text data and display-related data sequentially or periodically, and stores each of the collected information in the classification storage unit 13.

[0104] The general learning means Cxc generates a general classification model M43 that classifies web pages corresponding to each input feature by machine learning using the text features generated by the text classification means D2 and the page features generated by the page classification means D3 as input. The general classification means Dxc sequentially inputs the text features and page features associated with the multiple web pages to be classified into the general classification model M43. The general classification means Dxc then classifies the web pages associated with each input feature into one of multiple groups, and generates classification data by attaching the classification tag of the group to information indicating the web pages.

[0105] As described above, the classification device 310 of the present modification 4c can reflect information that cannot be obtained by a configuration using one piece of browsing data in the machine learning and classification processing by performing a two-stage classification process using two pieces of information, namely, text data and display-related data. Therefore, it is possible to improve the accuracy of grouping of web pages. Incidentally, the classification device 310 may be configured without the text preprocessing means B2. In this case, the text learning means C2 constructs a text classification model M20 that extracts text features from input text data by machine learning using a plurality of past text data as input.

[0106] Embodiment 5. A configuration example of a data analysis system 500 and its peripheral devices in the fifth embodiment will be described with reference to Fig. 16. The same configurations and components as those in the first to fourth embodiments described above will be denoted by the same reference numerals and description thereof will be omitted.

[0107] As shown in FIG. 16, a control unit 412 of a classification device 410 in the fifth embodiment has a collection processing means 312a, a preprocessing means 312b, a learning processing means 412c, and a classification processing means 412d.

[0108] The learning processing means 412c constructs a classification model M5 that classifies web pages associated with each piece of input information and generates classification data by machine learning using address information indicating the positions of web pages, analysis data corresponding to the web pages, and frame features corresponding to the web pages from the past as input. The learning processing means 412c can generate the classification model M5 by various machine learning methods such as learning using DNN, or by a combination of multiple machine learning methods, like the learning processing means 12c in the first embodiment.

[0109] The classification processing means 412d inputs address information indicating the location of a web page, analysis data corresponding to the web page, and frame features corresponding to the web page as page information to the classification model M5. The classification processing means 412d then classifies the web page associated with the page information input to the classification model M5 into one of a plurality of groups, and generates classification data by attaching the classification tag of the group to information indicating the web page.

[0110] The classification storage unit 13 of the fifth embodiment stores various data related to the classification process of page information, in addition to the operation program of the control unit 412 including the classification process program P5. The classification storage unit 13 also stores a classification model M5 for classifying page information. The classification process program P5 is a program for causing the control unit 412 as a computer and the classification storage unit 13 to function as the collection processing means 312a, the pre-processing means 312b, the learning processing means 412c, and the classification processing means 412d. Other configurations and alternative configurations are the same as those of the first to fourth embodiments described above. In addition, the basic flow of the learning process step performed by the learning process means 412c and the classification process step performed by the classification process means 412d in the classification process method by the classification device 410 is the same as that described with reference to FIG. 5.

[0111] As described above, the classification device 410 of the fifth embodiment can also automatically and efficiently group the web pages to be analyzed, thereby making it possible to speed up and streamline the processing. In addition, the classification processing means 412d inputs the address information, analysis data, and frame feature amount for each web page to the classification model M5, and classifies the web pages associated with each piece of input information into one of a plurality of groups. That is, according to the classification device 410, by performing classification processing using three pieces of information, namely, the address information, the text data, and the display-related data, it is possible to reflect information that cannot be obtained by a configuration using one piece of browsing data in the machine learning and classification processing. Therefore, it is possible to improve the accuracy of grouping of web pages.

[0112] Incidentally, the classification device 410 may be configured without the text pre-processing means B2. In this case, the learning processing means 412c constructs a classification model M5 that classifies web pages linked to each piece of input information by machine learning using address information, text data, and frame features of each past web page as input. The classification processing means 412d inputs the address information, text data, and frame features of each web page as page information to the classification model M5, and classifies the web pages linked to each piece of input information into one of a plurality of groups.

[0113] Also, the collection processing means 312a may collect text data and display related data sequentially or periodically, similarly to the modification 4c. In such a configuration, when the classification device 410 has text preprocessing means B2, the learning processing means 412c constructs a classification model M5 that classifies web pages linked to each piece of input information by machine learning using past analysis data and frame features for each web page as input. Then, the classification processing means 412d inputs the analysis data and frame features for each web page as page information to the classification model M5, and classifies the web pages linked to each piece of input information into one of a plurality of groups. In such a configuration, if the classification device 410 does not have the text preprocessing means B2, the learning processing means 412c constructs a classification model M5 that classifies web pages linked to each piece of input information by machine learning using text data and frame features of each past web page as input. The classification processing means 412d inputs the text data and frame features of each web page as page information to the classification model M5, and classifies the web pages linked to each piece of input information into one of a plurality of groups.

[0114] <Variation 5a> A configuration example of a data analysis system 500A in a modification 5a of the present embodiment 5 will be described with reference to FIG. 17. The collection processing means 312a in the present modification 5a collects address information and text data sequentially or periodically, as in the modification 4a. Therefore, the learning processing means 412c constructs a classification model M5 that classifies web pages associated with each piece of input information by machine learning using multiple sets of address information indicating the location of a web page and analysis data corresponding to the web page in the past as input. The classification processing means 412d inputs the address information indicating the location of the web page and the analysis data corresponding to the web page as page information to the classification model M5, and classifies the web page associated with the input page information into one of multiple groups.

[0115] As described above, the classification device 410 of the present modification 5a can reflect information that cannot be obtained by a configuration using one piece of browsing data in machine learning and classification processing by classification processing using two pieces of information, namely address information and text data. Therefore, it is possible to improve the accuracy of grouping of web pages. Meanwhile, the classification device 410 may be configured without the text pre-processing means B2. In this case, the learning processing means 412c constructs a classification model M5 that classifies web pages linked to each piece of input information by machine learning using address information and text data for each past web page as input. The classification processing means 412d inputs the address information and text data for each web page as page information to the classification model M5, and classifies the web pages linked to the input page information into one of a plurality of groups.

[0116] <Variation 5b> With reference to FIG. 18, a configuration example of a data analysis system 500B in a modification 5b of the present embodiment 5 will be described. The collection processing means 312a in this modification 5b sequentially or periodically collects address information and display-related data, similar to the modification 4b. Therefore, the learning processing means 412c constructs a classification model M5 that classifies web pages associated with each piece of input information by machine learning using address information indicating the positions of web pages in the past and frame features corresponding to the web pages as input. The classification processing means 412d inputs the address information and frame features of each web page as page information to the classification model M5, and classifies the web pages associated with the input page information into one of a plurality of groups.

[0117] As described above, the classification device 410 of the present modification 5b can reflect information that cannot be obtained by a configuration that uses one piece of browsing data in the machine learning and classification process by performing classification processing using two pieces of information, namely address information and display-related data. Therefore, it is possible to improve the accuracy of grouping of web pages.

[0118] The above-mentioned embodiments are preferred specific examples of the classification device, data analysis system, classification processing program, and classification processing method, and the technical scope of the present invention is not limited to these aspects. For example, in the above description, an example was shown in which the collection processing means collects text data and display-related data from the crawling system 60, but the present invention is not limited to this. The collection processing means may have a function of crawling a page in a site to obtain at least one of the text data and display-related data of the page. [Explanation of symbols]

[0119] 10, 110, 210, 310, 410 Classification device, 11 Communication unit, 12, 112, 212, 312, 412 Control unit, 12a, 112a, 212a, 312a Collection processing means, 12c, 112c, 212c, 312c, 412c Learning processing means, 12d, 112d, 212d, 312d, 412d Classification processing means, 13 Classification storage unit, 20 Analysis device, 22 Analysis processing unit, 23 Analysis storage unit, 23p Analysis processing program, 30 Management device, 31 Management communication unit, 32 Management control unit, 32a Information processing means, 32b Display processing means, 33 Management storage unit, 34, 84 Input unit, 35, 85 Display unit, 40 Service providing system, 50 Analysis system, 60 Crawling system, 80 Terminal device, 81 terminal communication unit, 82 terminal control unit, 82a information processing means, 82b display processing means, 83 terminal storage unit, 85x display screen, 91a, 91b, 91c, 92, 93, 94, 95, 96, 96a, 97 object, 100, 200, 300, 400, 400A to 400C, 500, 500A, 500B data analysis system, 112b, 212b, 312b, 412b preprocessing means, B2 text preprocessing means, B3 page preprocessing means, C1 address learning means, C2 text learning means, C3 page learning means, Cx, Cxa to Cxc general learning means, D1 address classification means, D2 text classification means, D3 page classification means, Dx, Dxa to Dxc general classification means, M1, M10 Address classification model, M2, M20 text classification model, M3, M30 page classification model, M4, M41, M42, M43 general classification model, M5 classification model, N network, P1~P5 classification processing program.

Claims

1. a classification device having a classification processing means for inputting page information associated with a web page on a website into a classification model generated by machine learning using a plurality of pieces of page information from the past, and classifying the input page information into one of a plurality of groups based on similarity in properties; An analysis device that performs an analysis process on the data obtained by the classification by the classification device, The classification device comprises: a pre-processing means for converting the web page into a wireframe image including part information corresponding to a type of each object using display-related data including information of each object arranged on a display screen in the web page, generating identification information for each object from the wireframe image by combining type information indicating the type of the object and placement information indicating a position and size of the object, and generating frame features which are fixed-length vector representations as the page information based on the generated identification information for each object.

2. a classification device having a classification processing means for inputting page information associated with a web page on a website into a classification model generated by machine learning using a plurality of pieces of page information from the past, and classifying the input page information into one of a plurality of groups based on similarity in properties; An analysis device that performs an analysis process on the data obtained by the classification by the classification device, The classification device comprises: a pre-processing means for converting the web page into a wireframe image including parts information corresponding to a type of each object by using display-related data including information of each object in the web page arranged on a display screen, generating identification information for each object from the wireframe image by combining type information indicating the type of the object and placement information indicating a position and size of the object, and generating a frame feature which is a fixed-length vector representation based on the generated identification information for each object, The classification processing means an address classification means for extracting address features from address information indicating a position of the web page as an input of an address classification model generated by machine learning using a plurality of pieces of past address information; a page classification means for extracting page features from the frame features as an input of a page classification model generated by machine learning using a plurality of past frame features; and a general classification means for inputting the address feature amount and the page feature amount for each of the web pages into the classification model as the page information, and grouping the web pages.

3. a classification device having a classification processing means for inputting page information associated with a web page on a website into a classification model generated by machine learning using a plurality of pieces of page information from the past, and classifying the input page information into one of a plurality of groups based on similarity in properties; An analysis device that performs an analysis process on the data obtained by the classification by the classification device, The classification device comprises: a pre-processing means for converting the web page into a wireframe image including parts information corresponding to a type of each object by using display-related data including information of each object in the web page arranged on a display screen, generating identification information for each object from the wireframe image by combining type information indicating the type of the object and placement information indicating a position and size of the object, and generating a frame feature which is a fixed-length vector representation based on the generated identification information for each object, The classification processing means a text classification means for extracting text features from text data, which is information on text in the webpage, as an input for a text classification model generated by machine learning using a plurality of pieces of the text data in the past; a page classification means for extracting page features from the frame features as an input of a page classification model generated by machine learning using a plurality of past frame features; and an overall classification means for inputting the text feature amount and the page feature amount for each of the web pages into the classification model as the page information, and grouping the web pages.

4. a classification device having a classification processing means for inputting page information associated with a web page on a website into a classification model generated by machine learning using a plurality of pieces of page information from the past, and classifying the input page information into one of a plurality of groups based on similarity in properties; An analysis device that performs an analysis process on the data obtained by the classification by the classification device, The classification device comprises: a text pre-processing means for performing an analysis process on text data, which is information on text within the web page, to generate analysis data; a page pre-processing means for converting the web page into a wireframe image including parts information corresponding to a type of each object by using display-related data including information of each object in the web page arranged on a display screen, generating identification information for each object from the wireframe image by combining type information indicating the type of the object and placement information indicating a position and size of the object, and generating a frame feature which is a fixed-length vector representation based on the generated identification information for each object, The classification processing means a text classification means for extracting text features from the analysis data as an input of a text classification model generated by machine learning using a plurality of pieces of analysis data from the past; a page classification means for extracting page features from the frame features as an input of a page classification model generated by machine learning using a plurality of past frame features; and an overall classification means for inputting the text feature amount and the page feature amount for each of the web pages into the classification model as the page information, and grouping the web pages.

5. a classification device having a classification processing means for inputting page information associated with a web page on a website into a classification model generated by machine learning using a plurality of pieces of page information from the past, and classifying the input page information into one of a plurality of groups based on similarity in properties; An analysis device that performs an analysis process on the data obtained by the classification by the classification device, The classification device comprises: a pre-processing means for converting the web page into a wireframe image including parts information corresponding to a type of each object by using display-related data including information of each object in the web page arranged on a display screen, generating identification information for each object from the wireframe image by combining type information indicating the type of the object and placement information indicating a position and size of the object, and generating a frame feature which is a fixed-length vector representation based on the generated identification information for each object, The classification processing means an address classification means for extracting address features from address information indicating a position of the web page as an input of an address classification model generated by machine learning using a plurality of pieces of past address information; a text classification means for extracting text features from text data, which is information on text in the webpage, as an input for a text classification model generated by machine learning using a plurality of pieces of the text data in the past; a page classification means for extracting page features from the frame features as an input of a page classification model generated by machine learning using a plurality of past frame features; and an overall classification means for inputting the address feature, the text feature, and the page feature for each web page into the classification model as the page information, and grouping each web page.

6. a classification device having a classification processing means for inputting page information associated with a web page on a website into a classification model generated by machine learning using a plurality of pieces of page information from the past, and classifying the input page information into one of a plurality of groups based on similarity in properties; An analysis device that performs an analysis process on the data obtained by the classification by the classification device, The classification device comprises: a text pre-processing means for performing an analysis process on text data, which is information on text within the web page, to generate analysis data; a page pre-processing means for converting the web page into a wireframe image including parts information corresponding to a type of each object by using display-related data including information of each object in the web page arranged on a display screen, generating identification information for each object from the wireframe image by combining type information indicating the type of the object and placement information indicating a position and size of the object, and generating a frame feature which is a fixed-length vector representation based on the generated identification information for each object, The classification processing means an address classification means for extracting address features from address information indicating a position of the web page as an input of an address classification model generated by machine learning using a plurality of pieces of past address information; a text classification means for extracting text features from the analysis data as an input of a text classification model generated by machine learning using a plurality of pieces of analysis data from the past; a page classification means for extracting page features from the frame features as an input of a page classification model generated by machine learning using a plurality of past frame features; and an overall classification means for inputting the address feature, the text feature, and the page feature for each web page into the classification model as the page information, and grouping each web page.

7. a classification device having a classification processing means for inputting page information associated with a web page on a website into a classification model generated by machine learning using a plurality of pieces of page information from the past, and classifying the input page information into one of a plurality of groups based on similarity in properties; An analysis device that performs an analysis process on the data obtained by the classification by the classification device, The classification device comprises: a page pre-processing means for converting the web page into a wireframe image including parts information corresponding to a type of each object by using display-related data including information on each object in the web page arranged on a display screen, generating identification information for each object from the wireframe image by combining type information indicating the type of the object and placement information indicating a position and size of the object, and generating a frame feature which is a fixed-length vector representation based on the generated identification information for each object, The classification processing means The data analysis system inputs address information indicating the location of the web page and the frame features corresponding to the web page as page information to the classification model, and classifies the input page information into one of the multiple groups.

8. a classification device having a classification processing means for inputting page information associated with a web page on a website into a classification model generated by machine learning using a plurality of pieces of page information from the past, and classifying the input page information into one of a plurality of groups based on similarity in properties; An analysis device that performs an analysis process on the data obtained by the classification by the classification device, The classification device comprises: a page pre-processing means for converting the web page into a wireframe image including parts information corresponding to a type of each object by using display-related data including information on each object in the web page arranged on a display screen, generating identification information for each object from the wireframe image by combining type information indicating the type of the object and placement information indicating a position and size of the object, and generating a frame feature which is a fixed-length vector representation based on the generated identification information for each object, The classification processing means The data analysis system inputs text data, which is information about the text in the webpage, and the frame features corresponding to the webpage into the classification model as page information, and classifies the input page information into one of the multiple groups.

9. a classification device having a classification processing means for inputting page information associated with a web page on a website into a classification model generated by machine learning using a plurality of pieces of page information from the past, and classifying the input page information into one of a plurality of groups based on similarity in properties; An analysis device that performs an analysis process on the data obtained by the classification by the classification device, The classification device comprises: a text pre-processing means for performing an analysis process on text data, which is information on text within the web page, to generate analysis data; a page pre-processing means for converting the web page into a wireframe image including parts information corresponding to a type of each object by using display-related data including information of each object in the web page arranged on a display screen, generating identification information for each object from the wireframe image by combining type information indicating the type of the object and placement information indicating a position and size of the object, and generating a frame feature which is a fixed-length vector representation based on the generated identification information for each object, The classification processing means The analysis data corresponding to the web page and the frame features corresponding to the web page are input to the classification model as page information, and the input page information is classified into one of a plurality of groups.

10. a classification device having a classification processing means for inputting page information associated with a web page on a website into a classification model generated by machine learning using a plurality of pieces of page information from the past, and classifying the input page information into one of a plurality of groups based on similarity in properties; An analysis device that performs an analysis process on the data obtained by the classification by the classification device, The classification device comprises: a pre-processing means for converting the web page into a wireframe image including parts information corresponding to a type of each object by using display-related data including information of each object in the web page arranged on a display screen, generating identification information for each object from the wireframe image by combining type information indicating the type of the object and placement information indicating a position and size of the object, and generating a frame feature which is a fixed-length vector representation based on the generated identification information for each object, The classification processing means a classification model that inputs address information indicating the location of the web page, text data that is information about the text in the web page, and the frame features corresponding to the web page as page information, and classifies the input page information into one of a plurality of groups.

11. a classification device having a classification processing means for inputting page information associated with a web page on a website into a classification model generated by machine learning using a plurality of pieces of page information from the past, and classifying the input page information into one of a plurality of groups based on similarity in properties; An analysis device that performs an analysis process on the data obtained by the classification by the classification device, The classification device comprises: a text pre-processing means for performing an analysis process on text data, which is information on text within the web page, to generate analysis data; a page pre-processing means for converting the web page into a wireframe image including parts information corresponding to a type of each object by using display-related data including information of each object in the web page arranged on a display screen, generating identification information for each object from the wireframe image by combining type information indicating the type of the object and placement information indicating a position and size of the object, and generating a frame feature which is a fixed-length vector representation based on the generated identification information for each object, The classification processing means a classification model that inputs address information indicating the location of the web page, the analysis data corresponding to the web page, and the frame features corresponding to the web page as page information, and classifies the input page information into one of a plurality of groups.

12. A classification device that classifies page information associated with a web page on a website, converting the web page into a wireframe image including parts information corresponding to a type of each object by using display-related data including information on each object arranged on a display screen in the web page, and generating identification information for each object from the wireframe image by combining type information indicating the type of the object and placement information indicating the position and size of the object; generating a frame feature, which is a fixed-length vector representation, as the page information based on the generated identification information for each object; The generated frame feature is input to a classification model generated by machine learning using a plurality of past frame features, and the input frame feature is classified into one of a plurality of groups based on similarity of properties; A classification processing method, in which an analysis device performs analysis processing on the data obtained by classification by the classification device.

Images