JP2017037577A - Apparatus, method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To output an appropriate commodity or service.SOLUTION: An apparatus includes: an individual attribute acquisition unit which acquires individual attribute data including a plurality of attribute values on a plurality of attributes of an individual; an object attribute generation unit which generates attribute data of each of attributes in object attribute data, on the basis of a distribution of attribute values of attributes in multiple pieces of individual attribute data for a plurality of individuals relating to an object; an attribute comparison unit which compares attribute data of one object with at least one of attribute data of one individual and object attribute data of another object, for each of a plurality of attributes; and a similarity calculation unit which calculates at least one of similarity between one object and one individual and similarity between one object and another object, on the basis of a comparison result of the attribute comparison unit.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an apparatus, a method, and a program.

A method has been proposed in which customer attribute data stored in a database is converted into a waveform, and the degree of approximation of a plurality of waveforms is calculated and used for market research (see Patent Document 1).
[Patent Document 1] Japanese Patent Application Laid-Open No. 2014-219785

  However, appropriate products or services may not be output.

  In the first aspect of the present invention, a personal attribute acquisition unit that acquires personal attribute data including a plurality of attribute values possessed by an individual for a plurality of attributes, and a plurality of personal attribute data for a plurality of individuals related to a certain object A target attribute generating unit that generates attribute data of each of the plurality of attributes in the target attribute data of the target based on a distribution of attribute values for each of the plurality of attributes, target attribute data of one target, Based on the comparison result by the attribute comparison unit and the attribute comparison unit that compares at least one of the individual attribute data of one individual and the target attribute data of another target for each of the plurality of attributes, And a similarity calculation unit that calculates the similarity of at least one of one object and another object.

  The summary of the invention does not enumerate all the features of the present invention. In addition, a sub-combination of these feature groups can also be an invention.

1 shows a system 10 according to the present embodiment. An example of the personal attribute data in this embodiment is shown. An example of object attribute data in this embodiment is shown. The 1st processing flow of the apparatus 100 which concerns on this embodiment is shown. An example of personal attribute data and object attribute data is shown. Another example of personal attribute data and object attribute data is shown. The 2nd processing flow of the apparatus 100 which concerns on this embodiment is shown. Another example of personal attribute data and object attribute data is shown. The approximate waveform of the distribution of the attribute which concerns on the object 1 and the object 3 is shown. The 3rd processing flow of the apparatus 100 which concerns on this embodiment is shown. An example of the output of the apparatus 100 is shown. The 4th processing flow of the apparatus 100 concerning this embodiment is shown. The approximate waveform of the distribution of the attribute which concerns on the object 1 and the object 2 is shown. An example of the similarity of the object 1 and the object 2 is shown. An example of the output of the apparatus 100 is shown. 10 shows a fifth processing flow of the apparatus 100 according to the present embodiment. An example of a selection instruction from the user and an output of the apparatus 100 is shown. An example of prioritization of objects by the recommendation processing unit 156 in FIG. 17 is shown. 6 shows a sixth processing flow of the apparatus 100 according to the present embodiment. An example of the hierarchical structure of a selection object is shown. An example of the display screen in a 1st hierarchy is shown. An example of the display screen after inputting a selection instruction is shown. 2 shows an exemplary configuration of a computer 1900 according to the present embodiment.

  Hereinafter, the present invention will be described through embodiments of the invention, but the following embodiments do not limit the invention according to the claims. In addition, not all the combinations of features described in the embodiments are essential for the solving means of the invention.

  FIG. 1 shows a system 10 according to this embodiment. The system 10 generates target attribute data based on the distribution of individual attributes and outputs a target having a high degree of similarity to the user attributes. In addition, when a certain target is selected according to a user's selection instruction or the like, the system 10 outputs another target having a high similarity to the target attribute.

  Here, the “target” refers to information to which awareness or action such as recognition, will, or desire is directed, and is preferably information that can be selected by the user via the user terminal 20. Such information may be, for example, a product, a service, a store, a person, a work, and / or an item on a website, and is also referred to as an item. However, the “target” is not limited to these, and a group of people related to these “target”, for example, a group of people who visit a certain store may be further included in the “target”. The system 10 includes a user terminal 20, a search server 30, a store terminal 40, a history management server 46, a personal attribute database 52, a target attribute database 54, and the device 100.

  One or a plurality of user terminals 20 are connected to the apparatus 100 via the network 22, and a selection instruction for selecting an object is input from the user by a button, a keyboard, a touch panel, and / or voice input, and the selection instruction is input to the apparatus. To 100. The user terminal 20 receives the processing result corresponding to the selection instruction from the device 100 and displays it on the display screen. The user terminal 20 may be a personal computer, a notebook computer, a mobile phone, a smartphone, or other terminal device capable of data input / output.

  The network 22 includes various networks such as the Internet, a wide area network (WAN), a local area network, or a combination thereof. The network 22 may include a connection by at least one of a wired network and a wireless network at each connection point. For example, the network 22 may be realized by the Internet.

  The search server 30 is connected to the apparatus 100 via the network 22, receives a search key from the apparatus 100, extracts one or a plurality of objects that hit the search key from the search database 32, and extracts the extracted results from the apparatus 100. Send to. The search server 30 may be realized by a generally available search engine or the like.

  One or a plurality of store terminals 40 are connected to the apparatus 100 via the network 42, and are POS terminals provided at various stores or computers having equivalent functions. The store terminal 40 transmits purchase data of the target person to the history management server 46 when the target person purchases a product or service (hereinafter may be collectively referred to as “product or the like”).

  The network 42 includes various networks such as the Internet, a wide area network (WAN), a local area network, or a combination thereof. The network 42 may include a connection by at least one of a wired network and a wireless network at each connection point. For example, the network 42 may be realized by a dedicated line separated from the network 22. The network 22 and the network 42 may be the same network (for example, the Internet).

  The history management server 46 is a computer such as a server computer, and is connected to the store terminal 40 and the device 100 via the network 42 and manages one or a plurality of store terminals 40 provided in each store. The history management server 46 may be a distributed system including a plurality of computers. The history management server 46 performs purchase processing (sales processing) based on purchase data received from the store terminal 40 and records user purchase data in the purchase history database 44. Further, the history management server 46 provides user purchase data to the device 100 in response to a request from the device 100.

  The apparatus 100 is a computer such as a server computer, and receives a selection instruction from the user terminal 20 and outputs a target corresponding to the selection instruction based on a comparison result obtained by comparing the target and the user attributes. The apparatus 100 includes an instruction acquisition unit 110, a personal attribute acquisition unit 120, a target attribute generation unit 130, an estimation unit 140, and an output processing unit 150 that are connected to each other via a bus. The apparatus 100 may be a distributed system that processes each of these configurations or functions, or a part of each configuration or function by a plurality of computers.

  The instruction acquisition unit 110 acquires from the user terminal 20 a selection instruction by the user for selecting some of the plurality of targets. For example, the instruction acquisition unit 110 acquires a search key (for example, a search keyword) for searching for a part of a plurality of objects as a selection instruction. In addition, for example, the instruction acquisition unit 110 may input, as a selection instruction, an input indicating a selection target selected by the user from two or more selection targets included in the screen presented to the user (for example, a text with a hyperlink attached). Clicks, etc.). In addition, the instruction acquisition unit 110 may select a part of the target without depending on the selection instruction for the user terminal 20. For example, the instruction acquisition unit 110 may acquire the purchase history of the user stored in the history management server 46 as a selection instruction, and may select a target (such as a product) purchased most recently, or the search server 30 or the user The user's search history stored in the terminal 20 may be acquired as a selection instruction, and the most recently searched object may be selected. The instruction acquisition unit 110 may supply the acquired selection instruction to the target attribute generation unit 130, for example.

  In addition, the instruction acquisition unit 110 may separately acquire personal identification information (for example, a login ID) for identifying the user in order to specify the user who has input the selection instruction. The instruction acquisition unit 110 may supply the acquired personal identification information to the personal attribute acquisition unit 120, for example.

  The personal attribute acquisition unit 120 acquires personal attribute data including a plurality of attribute values possessed by an individual for a plurality of attributes. For example, the personal attribute acquisition unit 120 acquires personal identification information of a user who has input a selection instruction such as a search key from the instruction acquisition unit 110, and converts the personal attribute data including the personal identification information into personal attribute data of a plurality of individuals. Is obtained by searching from the personal attribute database 52 in which is stored. Thereby, the personal attribute acquisition unit 120 acquires the personal attribute data of the user who has input the selection instruction. The personal attribute acquisition unit 120 may generate personal attribute data of a plurality of users from the purchase histories of the plurality of users stored in the purchase history database 44 and store the personal attribute data in the personal attribute database 52. Details of the personal attribute data will be described later.

  The target attribute generation unit 130 generates target attribute data indicating a plurality of attribute data of each target for each of the plurality of attributes in the target attribute data, corresponding to each of the plurality of targets. For example, the target attribute generation unit 130 may generate target attribute data for each target based on a plurality of personal attribute data for a plurality of individuals related to each target, and store the target attribute data in the target attribute database 54. Here, “related to a certain object” means being interested in the object, for example, tasting the object. As an example, when an individual purchases, sells, uses and / or rents an object, when an individual bookmarks and / or collects information about an object, and when an individual sends or posts a message about the object When doing so, the individual can be said to be “related to a certain object”.

  In addition, the target attribute generation unit 130 acquires target target attribute data that matches a selection instruction such as a search key acquired by the instruction acquisition unit 110 from the target attribute database 54 that stores a plurality of target target attribute data. Details of the target attribute data will be described later.

  When the attribute value for at least one attribute is missing among a plurality of attributes in one individual's personal attribute data, the estimating unit 140 determines the attribute value missing in the one individual's personal attribute data. presume. Further, the estimation unit 140 stores the estimated attribute value in association with the data missing attribute of one individual whose data is missing among the plurality of personal attribute data stored in the personal attribute database 52. For example, when a part of items such as a registration form for collecting basic attribute data or a questionnaire for collecting research data is missing and an attribute value of an attribute corresponding to the item cannot be obtained, an estimation unit 140 may predict the attribute value of the attribute from the attribute values of other attributes. For example, even if the information on whether or not a user has a driver's license is not obtained in the research data, the estimation unit 140 has a high probability (for example, 95) if the purchase history data includes information on car purchase. %), It can be estimated that the user has a driver's license.

  Further, for example, the estimation unit 140 may generate a probability inference algorithm from learning data from which attribute value data of all attributes is obtained, and predict missing attribute values based on the probability inference algorithm. The personal attribute acquisition unit 120 may use various methods such as a Bayesian network, a neural network, and a random forest as a probability inference algorithm. Moreover, the estimation part 140 may estimate the missing attribute value by calculating the waveform of the user's personal attribute data with missing attribute values by multiple regression analysis.

  In addition, for example, the estimation unit 140 includes a plurality of attributes included in target attribute data of a target whose similarity with respect to the personal attribute data of one individual whose attribute values are missing for some attributes is equal to or higher than the reference similarity. Of the attribute values, the missing attribute values may be estimated based on the distribution of the attribute values for some of the attributes. Details of such an estimation method will be described later.

  The output processing unit 150 prioritizes each target selected by a selection instruction among a plurality of targets based on a comparison result comparing the personal attribute data and the target attribute data, or a comparison result comparing the target attribute data. And output. Here, “each target selected by the selection instruction” may be a target directly selected by the selection instruction. For example, the output processing unit 150 prioritizes and outputs each target that hits the search key among a plurality of targets based on the comparison result. Further, “each target selected by the selection instruction” may be another target derived from the target selected by the selection instruction. For example, the output processing unit 150 may prioritize and output each target to be included in the next screen to be presented to the user according to the target selected by the selection instruction based on the comparison result. The output processing unit 150 includes a target suitability calculation unit 152, a selection unit 153, an attribute comparison unit 154, a similarity calculation unit 155, and a recommendation processing unit 156.

  The target suitability calculation unit 152 calculates a target suitability indicating the suitability between the selection instruction and the target selected by the selection instruction. For example, the target suitability calculation unit 152 requests the search server 30 to search for a target that matches at least a part of the search key, receives a search result including one or a plurality of targets from the search server 30, and receives each received target. The degree of fitness for the search key is calculated. The target fitness calculation unit 152 may supply the calculated target fitness of each target to the recommendation processing unit 156, for example.

  The selection unit 153 selects an attribute to be compared when the similarity between the target attribute data of one target and the personal attribute data of one individual is calculated by the attribute comparison unit 154 and the similarity calculation unit 155 To do. For example, the selection unit 153 selects an attribute whose influence on the presence / absence of the relationship with respect to one target is greater than or equal to a reference from among a plurality of attributes. Further, the selection unit 153 should be compared when the similarity between the target attribute data of one target and the target attribute data of another target is calculated by the attribute comparison unit 154 and the similarity calculation unit 155. Select an attribute. For example, the selection unit 153 selects, from among a plurality of attributes, an attribute whose influence on the presence / absence of association with one target is higher than a reference and an attribute whose influence on the presence / absence of association with another target is higher than a reference To do. The selection unit 153 may supply the attribute selection result to the attribute comparison unit 154, for example.

  Here, the degree of influence on whether or not a certain attribute is related to a target is the attribute value of the attribute of the personal attribute data between a plurality of individuals related to the target and a plurality of individuals not related to the target. The degree of distribution is different. As an aspect of distribution, for example, average, dispersion, skewness, kurtosis, or the like, or a combination thereof can be used.

  For example, when the target is “cosmetics”, the set of individuals (for example, purchasers) related to “cosmetics” includes many women. Therefore, the degree of difference in the distribution of attribute values of “sex” (statistical probability that the user is a woman) is large between a plurality of individuals related to “cosmetics” and a plurality of individuals not related. That is, the degree of influence of the attribute “sex” on the presence / absence of association with the target “cosmetics” is large. In such a case, the attribute “sex” is an attribute to be compared in order to calculate the similarity between the object “cosmetics” and one individual, or the similarity between the object “cosmetics” and another object. Selected as. As a result, the comparison is performed with the attribute having a large influence on the presence / absence of the relationship with the object “cosmetics”, that is, the attribute having a characteristic distribution in the object attribute data of “cosmetics”. Therefore, the reliability of similarity is improved.

  For example, when the target is “umbrella”, there is no gender bias in the set of individuals (for example, purchasers) related to “umbrella”. Therefore, the degree of difference in the distribution of attribute values of “gender” is small between a plurality of individuals related to “umbrella” and a plurality of unrelated individuals. That is, the degree of influence of the attribute “gender” on the presence / absence of association with the target “umbrella” is small. In such a case, the attribute “gender” is not selected as an attribute having a large influence on the presence / absence of association with the target “umbrella”. Therefore, when calculating the similarity between the target “umbrella” and one individual, the attribute “sex” is not selected as the attribute to be compared. In addition, when calculating the similarity between the target “umbrella” and another target, the attribute “gender” is selected unless “gender” is selected as an attribute having a large influence on the presence / absence of association with the other target. It is not selected as an attribute to be compared. This prevents a comparison with an attribute having a small influence on the presence / absence of association with the target “umbrella”, that is, an attribute having no characteristic distribution in the target attribute data of “umbrella”. Therefore, the reliability of similarity is improved.

  The attribute comparison unit 154, when the similarity between the target attribute data of one target and the personal attribute data of one individual is calculated by the similarity calculation unit 155, the target attribute data of one target, Individual personal attribute data is compared for each of a plurality of attributes. For example, the attribute comparison unit 154 compares the target attribute data of one target with the personal attribute data of one individual for each attribute selected by the selection unit 153 from among a plurality of attributes, and compares the comparison result with the similarity. It supplies to the calculation part 155.

  Further, the attribute comparison unit 154 determines that when the similarity between the target attribute data of one target and the target attribute data of another target is calculated by the similarity calculation unit 155, the target attribute data of one target The target attribute data of other targets is compared for each of the plurality of attributes. For example, the attribute comparison unit 154 compares the target attribute data of one target with the target attribute data of another target for each attribute selected by the selection unit 153 from among a plurality of attributes, and compares the comparison result with the similarity. It supplies to the calculation part 155. As an example, the attribute comparison unit 154 has, for each of the attributes whose influence on the presence / absence of the relationship with respect to one object is higher than the reference and the attributes whose influence on the presence / absence of association with other objects is higher than the reference, The target attribute data of one target is compared with the target attribute data of another target.

  The similarity calculation unit 155 calculates the similarity between one target and one individual based on the comparison result by the attribute comparison unit 154 for the target attribute data of one target and the personal attribute data of one individual. To do. For example, the similarity calculation unit 155 compares the personal attribute data of the user who has operated the user terminal 20 and the target attribute data of each target included in the search result received from the search server 30 by the target suitability calculation unit 152. Is calculated. Also, the similarity calculation unit 155 determines the similarity between the one target and the other target based on the comparison result by the attribute comparison unit 154 for the target attribute data of the one target and the target attribute data of the other target. Calculate the degree. For example, the similarity calculation unit 155 calculates the similarity between the target attribute data of the target displayed by the user on the browser of the user terminal 20 and the target attribute data of another target. The similarity calculation unit 155 may supply the calculated similarity to the recommendation processing unit 156.

  The recommendation processing unit 156 determines whether or not to recommend the one target to the user based on the target attribute data for one target, for example, one product or service, and the personal attribute data of the user, The target judged to be recommended is output. For example, based on the similarity calculated by the similarity calculation unit 155, the recommendation processing unit 156 determines that some of the targets are recommended, selects them, outputs them with priority.

  Also, the recommendation processing unit 156 determines whether or not to recommend one target to the user based on the target attribute data for one target and the attribute data for another target, and determines to recommend it. The target is output. For example, the recommendation processing unit 156 determines and recommends some of the plurality of targets based on the similarity calculated by the similarity calculation unit 155, and prioritizes and outputs the selected target. As an example, when a search key is used as a selection instruction, the recommendation processing unit 156 selects a plurality of targets selected according to the selection instruction, the target fitness calculated by the target fitness calculation unit 152, and the similarity. Ranking based on the similarity calculated by the calculation unit 155, and the output result to the user terminal 20 that has input the selection instruction so that the target matching degree and the similarity are comprehensively displayed first. Supply.

  Note that the recommendation processing unit 156 may cause the target advertisement to be output on the display screen of the user terminal 20, or may send a direct mail including a privilege about the target to the user. By performing at least one of these, the recommendation processing unit 156 recommends a target (for example, a product or a service) to the user.

  As described above, the apparatus 100 according to the present embodiment acquires personal attribute data including a plurality of attribute values that an individual has for a plurality of attributes, and attribute values for each attribute in the personal attribute data of a plurality of individuals related to a certain target. Based on the distribution, the attribute data of each attribute in the target attribute data of the target is generated. Then, the target attribute data of one target and at least one of the personal attribute data of one individual and the target attribute data of another target are compared for each attribute, and the similarity is calculated based on the comparison result. As a result, it is possible to output one object having a high degree of similarity to one individual or another object. Therefore, a more appropriate target can be output.

  Note that a plurality of elements of the system 10 may be combined to form one element. For example, the personal attribute database 52 and the target attribute database 54 may be realized by a single database. The device 100 may include elements other than the device 100 (for example, the personal attribute database 52), or the elements in the device 100 (for example, the personal attribute acquisition unit 120) may be implemented by an external server device or the like. Good.

FIG. 2 shows an example of personal attribute data in the present embodiment. As shown in the figure, personal attribute data includes a plurality of attribute values possessed by each individual for a plurality of attributes, and is represented as a waveform on a graph. FIG. 2 shows the personal attribute data of the person 1 represented by a solid line and the personal attribute data of the person 2 represented by a broken line. For example, the attribute x ₁ (marriage) of the individual 1 is 100%, and the attribute x ₂ (child) is 59%. This may mean that the statistical probability that the individual 1 is married is 100% (ie, surely married) and the probability that the individual 1 has children is estimated to be 60%. Further, for example, the attribute x ₃ (driving) of the individual 1 is 48%, and the attribute x ₄ (smoking) is 20%. This is about 48 on a scale of 0-100 for the driving of the individual 1 (for example, driving frequency of about twice a month), and about 20 on a scale of 0-100 for the smoking of the individual 1 ( For example, it means smoking amount of about one box per month).

  The apparatus 100 generates the personal attribute data of the individual 1 that forms the waveform as shown in FIG. 2 by specifying the attribute values of the N attributes of the individual 1. In this way, the apparatus 100 generates personal attribute data for each of a plurality of individuals. In addition to / instead of generating the personal attribute data for each individual, the apparatus 100 also represents personal attribute data representing a plurality of individuals belonging to the same group or similar groups (for example, the personal attribute data of all family members belonging to a specific household) Alternatively, personal attribute data corresponding to all members of a specific group may be generated.

FIG. 3 shows an example of target attribute data in the present embodiment. The target attribute data may be generated based on personal attribute data of a plurality of individuals related to each target, and may include a distribution of attribute values of each attribute of the individual who selected each target 1. For example, FIG. 3 shows target attribute data generated based on the personal attribute data of a plurality of individuals who purchased the target 1 which is one product. As shown, the distribution ratio of the individual attribute value of the attribute x ₁ 90% or more of the plurality of individuals who bought target 1 is 10%, the distribution ratio of the individual attribute values less than 80% 90% Is 5%, the distribution ratio of individuals whose attribute value is 70% or more and less than 80% is 0%,..., The distribution ratio of individuals whose attribute value is 0% or more and less than 10% is 5%. Further, the plurality of individuals who bought target 1, of the attribute value of the attribute x _1, the most occurrences large distribution segment is less than 30% to 40% of the frequency, the attribute value of the maximum frequency, that is the mode that It indicates that the value is within this range (for example, 35% which is the median value within the range). The average of the attribute values of the attribute x ₁ of a plurality of individuals who bought subject 1 is 41%. In this way, the target attribute data is represented as distribution data of individual attribute values related to the target. The target attribute data may be represented not by the distribution ratio (%) of individuals related to the target but by the total number of individuals related to the target.

  FIG. 4 shows a first processing flow of the apparatus 100 according to the present embodiment. The apparatus 100 generates personal attribute data and target attribute data by executing the processes of S110 to S150.

  First, the personal attribute acquisition unit 120 acquires basic attribute data for a plurality of individuals (S110). For example, the personal attribute acquisition unit 120 provides basic attribute information (for example, age, gender, address, residence area, occupation, etc.) provided to each store by a plurality of individuals from the store terminal 40 at the time of membership registration. Married and unmarried, and / or commuting school area, etc.) are acquired as basic attribute data in association with the individual identification information of each individual. The personal attribute acquisition unit 120 may acquire basic attribute data directly from the user via the user terminal 20 in addition to / in place of the store terminal 40.

  Next, the personal attribute acquisition unit 120 acquires purchase history data for a plurality of individuals (S120). For example, the personal attribute acquisition unit 120 purchases a plurality of individuals linked with personal identification information from the history management server 46 (purchased product / service name, purchased product / service category, purchased product / service characteristics, Purchase price / price range, purchase date / time, purchase place / store, and / or payment means, etc.). The personal attribute acquisition unit 120 may acquire the behavior history of each individual on a communication network such as the Internet in addition to / in place of the purchase history. For example, the personal attribute acquisition unit 120 may acquire the name of the website accessed by each individual, the category of the website, the estimated staying time, the migratory behavior style, and the like as the action history.

  Next, the personal attribute acquisition unit 120 acquires research data about a plurality of individuals (S130). For example, the personal attribute acquisition unit 120 obtains questionnaire response results obtained from a plurality of individuals (for example, whether married or unmarried, presence of children, residence status, income, presence or absence of a driver's license, credit status, hobby preference, Behavioral style, luxury orientation, price orientation, traditional orientation, and / or innovation orientation, etc.) may be associated with personal identification information and acquired as research data. The research data may be data in which at least a part of information overlaps with the basic attribute data.

Next, the personal attribute acquisition unit 120 generates personal attribute data (S140). The personal attribute acquisition unit 120 assigns attribute values of each attribute to a plurality of individuals based on at least one of the acquired basic attribute data, purchase history data, behavior history data, and research data. For example, the personal attribute acquiring unit 120, to indicate that the basic attribute data is married, may allocate 100% of the attribute value of the attribute x ₁ regarding Marriage. Further, for example, if 5 out of upscale of value 0 included in the research system data, may allocate 50% of the attribute value of the attribute x _N indicating the upscale properties.

  Here, the personal attribute acquisition unit 120 stores the personal attribute data generated for each individual in the personal attribute database 52 in association with the personal identification information.

  Next, the target attribute generation unit 130 determines the plurality of attributes in the target attribute data of the target based on the distribution of each of the plurality of attributes in the plurality of personal attribute data for the plurality of individuals related to the target. Each attribute data is generated (S150). For example, when the target is a product or service, the target attribute generation unit 130 purchases, sells, uses, rents, bookmarks, collects information about the target, or sends / posts a message about the target for a plurality of targets. The personal identification information of a plurality of individuals who have performed etc. is acquired from the history management server 46. Thereafter, the target attribute generation unit 130 acquires personal attribute data corresponding to the acquired personal identification information from the personal attribute database 52, aggregates attribute values of each attribute, and creates a distribution of attribute values of each attribute. Then, the target attribute generation unit 130 uses the attribute value distribution itself of each attribute and / or a numerical value representing the distribution as target attribute data for the target based on the distribution of attribute values for each of the plurality of attributes. Generate.

  For example, the target attribute generation unit 130 generates attribute data of each attribute in the target attribute data of each target based on the result of averaging a plurality of individual attribute data for each of the plurality of attributes. As a result, the target attribute generation unit 130 generates target attribute data including the average value of the individual attribute values related to the target as the attribute value of each attribute, as shown in the average row of FIG. In addition to the average, the target attribute generation unit 130 uses numerical values (variance, deviation, kurtosis, skewness, type of distribution, KLD, JSD, etc.) regarding the distribution of each attribute of the target attribute data as target attribute data. It may be calculated as at least a part of

  Further, for example, the target attribute generation unit 130 divides the range of values that the attribute value can take into equally spaced sections, and totals the number of people for each section. As a result, the target attribute generation unit 130 generates target attribute data representing the distribution of attribute values in the personal attribute data of a plurality of individuals related to the target, as indicated by the frame in FIG.

  By such an operation, the target attribute generation unit 130 can generate target attribute data representing the attribute value distribution itself, that is, the characteristics of each attribute, rather than a single attribute value such as an average value. The state of each attribute of each target can be reflected in the target attribute data.

  As described above, the apparatus 100 generates the personal attribute data and the target attribute data by the processes of S110 to S150. Here, the apparatus 100 may execute all or a part of the processing of S110 to S150 in advance before the instruction acquisition unit 110 acquires a selection instruction from the user, or the instruction acquisition unit 110 issues a selection instruction from the user. You may perform all or one part of the process of S110-S150 whenever it acquires.

  Note that the apparatus 100 may omit any of the processes of S110 to S130 as necessary. For example, the apparatus 100 may omit the process of S <b> 130 when research data is not necessary for generating personal attribute data. Moreover, if there exists a process required for the production | generation of personal attribute data and object attribute data, it can perform suitably. For example, the apparatus 100 can generate personal attribute data and target attribute data using the technique described in Patent Document 1.

  FIG. 5 shows an example of personal attribute data and target attribute data. FIG. 5 shows an example in which the apparatus 100 generates a waveform of the personal attribute data of the individual 1 and a waveform of the target attribute data of the target 1 and the target 2 by the first processing flow. In this figure, the waveform of the target attribute data of the target 1 and the target 2 is formed from the average of the attribute values of the respective attributes of the individual personal attribute data related to each target. The distance between the waveforms and the similarity of the shape represent the affinity between the individual and the object or between the objects. For example, FIG. 5 shows that the waveform of the target 1 is closer to the waveform of the individual 1 than the waveform of the target 2, and the individual 1 is more likely to be related to the target 1 than the target 2.

  FIG. 6 shows another example of personal attribute data and target attribute data. FIG. 6 shows a case where the apparatus 100 generates the waveform of the personal attribute data of the individual 1 and the distribution of the target attribute data of the target 1 by the first processing flow. The distribution of the target attribute data indicated by hatched hatching corresponds to the target 1 in FIGS. The density of the diagonal lines corresponds to the distribution ratio of the target attribute data. The personal attribute data of the individual 1 indicated by the dotted line and the target attribute data of the target 1 indicated by the solid line are the same as those shown in FIG.

The distribution ratio on each attribute of the target attribute data corresponding to the waveform of the personal attribute data represents the degree of approximation of the individual and the target. The apparatus 100 can evaluate the affinity with an individual more strictly by expressing the target attribute data as a distribution. For example, when the degree of approximation between the attribute value of the target attribute data of the target 1 and the attribute value of the personal attribute data of the individual 1 is evaluated by averaging without considering the distribution in the attribute x ₃ , the dotted line graph and the solid line graph of FIG. As such, both are evaluated as being relatively approximate.

However, the distribution of the target attribute data in the attribute x ₃ is sparse compared with the surrounding area at a value (around 50%) in the vicinity of the graph of the personal attribute data, and actually the individual 1 and the target 1 the attribute value of the attribute x ₃ is likely you do not have too much approximation. For example, the individual 1 is an individual who performs driving at a moderate frequency (for example, twice a month), whereas the group of individuals related to the target 1 is a high-frequency driving individual (for example, every day) and the driving frequency. Since they are composed of individuals with low (for example, once in February), it is highly possible that they are different in nature. Since the apparatus 100 of this embodiment handles target attribute data as a distribution, it is possible to estimate the similarity between an individual and a target and the similarity between targets with higher accuracy.

  FIG. 7 shows a second processing flow of the apparatus 100 according to the present embodiment. When the attribute value for one attribute in one individual's personal attribute data is missing, the apparatus 100 executes the processing of S210 to S290, so that one attribute of the one individual's personal attribute data is obtained. Estimate attribute values. Note that the second processing flow may be executed when the first processing flow is performed and the generated personal attribute data is missing.

  First, the estimation unit 140 acquires user personal attribute data (S210). For example, the instruction acquisition unit 110 supplies a login screen to the user terminal 20 via the network 22, and personal identification information (for example, login ID, member ID, name, or email address) is given to the user on the login screen. The personal identification information of the user who operates the user terminal 20 is acquired. The personal attribute acquisition unit 120 receives the user's personal identification information from the instruction acquisition unit 110 and acquires the personal attribute data associated with the personal identification information from the personal attribute database 52 to operate the user terminal 20. Get the personal attribute data of the user. The personal attribute acquisition unit 120 supplies the acquired personal attribute data to the attribute comparison unit 154 of the output processing unit 150. If the estimation unit 140 has acquired the personal attribute data created in the first processing flow from the personal attribute acquisition unit 120, the process of S210 may not be performed. Here, in this processing flow, the attribute value for one attribute in the acquired personal attribute data is missing.

  Next, the target attribute generation unit 130 acquires target attribute data of a plurality of targets (S240). For example, the target attribute generation unit 130 acquires target attribute data for all targets stored in the target attribute database 54. The target attribute generation unit 130 supplies the acquired target attribute data to the selection unit 153 and the attribute comparison unit 154 of the output processing unit 150. If the target attribute generation unit 130 holds the target attribute data created in the first processing flow, the process of S240 may not be performed.

  Next, the selection unit 153 selects an attribute to be compared in the calculation of the similarity between the target attribute data of one target and the personal attribute data of one individual (S252). That is, the selection unit 153 selects an attribute having an influence level greater than or equal to the reference on the presence / absence of association with the target from among a plurality of attributes in the personal attribute data, and outputs the selection result to the attribute comparison unit 154. For example, the selection unit 153 may detect, for each target from which target attribute data is acquired, an attribute whose influence on the presence / absence of the relationship with the target is greater than or equal to a reference, and select and output each detected attribute. As an example, the selection unit 153 includes, for each attribute, a set of personal attribute data of a plurality of individuals related to one target and a set of personal attribute data of a plurality of individuals not related to the one target. An attribute value distribution mode (average, variance, skewness, kurtosis, etc., or a combination thereof) may be calculated, and an attribute having a distribution mode difference degree greater than or equal to a reference may be selected. As a reference for the degree of distribution mode difference (influence), a predetermined percentile rank in a set of a plurality of distribution modes calculated for a plurality of attributes may be used, or a fixed value may be used.

  Next, the attribute comparison unit 154 compares the personal attribute data of the user acquired in S210 and the target attribute data of each target acquired in S240 for each attribute (S254). Based on the comparison result, the user and each target are compared. Is calculated by the similarity calculation unit 155 (S260). For example, the attribute comparison unit 154 may compare the target attribute data for each attribute selected in S252. Here, the similarity calculation unit 155 calculates the similarity on a scale of 0 to 1.

  Here, regarding S254 and S260, first, (1) a method of calculating the similarity when the comparison is performed using the waveform of the target attribute data will be described.

  In this case, in S254, the attribute comparison unit 154 compares the waveform of the user's personal attribute data and the waveform of the target attribute data of each target (also referred to as “both waveforms”) to calculate the difference between the two waveforms. To do. For example, the attribute comparison unit 154 may calculate the distance between both waveforms by comparing both waveforms. As an example, the attribute comparison unit 154 may calculate a difference by comparing the attribute value of each attribute of the personal attribute data with the attribute value (average value) of the corresponding attribute of the target attribute data. In S260, the similarity calculation unit 155 calculates the similarity based on the ratio of attributes whose calculation results are within a predetermined threshold among all the attributes. As an example, when the difference between the attribute values of 50 attributes out of N = 100 attributes is within the threshold of 10%, 0.50 is calculated as the similarity.

Also, for example, in S254, the attribute comparison unit 154 calculates the closeness of the line segment angles of both waveforms by comparing both waveforms. For example, the attribute comparison unit 154 includes line segment angles of two adjacent attributes in the personal attribute data (for example, a line configured on the graph by the attribute value of the attribute x _{1 and} the attribute value of the attribute x ₂ of the personal attribute data). the difference between the minute angle), a line segment angle of the corresponding attribute in the subject attribute data (e.g., a line segment constructed on the graph by the attribute value of the attribute x ₁ of the target attribute data and the attribute value of the attribute x ₂₎ May be calculated. In S260, the similarity calculation unit 155 calculates the similarity based on the ratio of attributes whose calculation results are within a predetermined threshold among all the attributes. Specifically, when 99 adjacent sections are formed from N = 100 attributes, and the line segment angle difference in 50 sections is within the threshold of 10 degrees, approximately 0.51 is similar. Calculate as degrees.

  The similarity calculation unit 155 may calculate the similarity based on both the distance between both waveforms and the proximity of the line segment angles of both waveforms. For example, the similarity calculation unit 155 may calculate an average or a weighted average of the distance between the two waveforms and the line segment angle between the two waveforms as the similarity. As described above, in (1), the attribute comparison unit 154 and the similarity calculation unit 155 represent the target attribute data as an averaged waveform, and are similar depending on the degree of approximation between the waveform of the target attribute data and the position and shape of the personal attribute data. Calculate the degree.

  Next, regarding S254 and S260, (2) a similarity calculation method in the case of performing comparison using distribution data of a plurality of attributes of target attribute data will be described.

In this case, in S254, the attribute comparison unit 154 calculates the difference between the two by comparing the waveform of the personal attribute data with the distribution of a plurality of attributes of the target attribute data. That is, the attribute comparison unit 154 acquires the distribution data of a plurality of attributes of the target attribute data, calculates the distribution ratio of the target attribute data in a predetermined section including the attribute value of the personal attribute data for each attribute, The degree of approximation (difference) of each attribute. As an example, the distribution ratio of a section (including 100%) of attribute x ₁ of personal attribute data having an attribute value of 100% (corresponding to individual 1 in FIG. 2) and 90% or more of attribute x ₁ of target attribute data If There is a 10% (corresponding to the distribution shown in FIG. 3), the attribute comparison unit 154, the degree of approximation of attributes _{x 1} may be calculated to 0.1 (10%).

  In step S260, the similarity calculation unit 155 calculates an average or weighted average of the approximation degrees of the plurality of attributes, and sets the similarity between the personal attribute data and the target attribute data. Thereby, the degree of similarity is calculated higher as the distribution ratio of the section corresponding to the attribute value of the personal attribute data is larger in the distribution ratio (frequency value) of each section in the distribution of the target attribute data. As described above, in (2), the similarity calculation unit 155 calculates the similarity based on the distribution ratio of the target attribute data and the attribute value of the personal attribute data using the distribution of the target attribute data.

  Next, regarding S254 and S260, (3) a similarity calculation method in the case of performing comparison using distribution parameters for each of a plurality of attributes in the target attribute data will be described. Here, the distribution parameter is, for example, an average, a variance, a skewness, a kurtosis, a mode, a distribution type, or a combination thereof.

  In this case, in S254, the attribute comparison unit 154 compares the target attribute data of one target with the personal attribute data of one individual using the distribution parameter, and calculates the difference between the two. For example, as in (1), the attribute comparison unit 154 compares the attribute value of each attribute of the personal attribute data with the attribute value (average value) of the corresponding attribute of the target attribute data, and calculates the difference. In S260, the similarity calculation unit 155 calculates the similarity based on the ratio of attributes whose calculation results are within a predetermined threshold among all the attributes. Note that the attribute comparison unit 154 may use the mode value of the attribute value of each attribute of the target attribute data instead of using the attribute value (average value) of each attribute of the target attribute data.

  In S254, the attribute comparison unit 154 generates the original distribution from the distribution parameter for each attribute instead of using the distribution parameter itself for comparison, and calculates the difference (degree of approximation) by the same method as in (2). May be. In step S260, the similarity calculation unit 155 calculates an average or weighted average of the approximation degrees of the plurality of attributes, and sets the similarity between the personal attribute data and the target attribute data. Thereby, the degree of similarity is calculated higher as the distribution ratio of the section corresponding to the attribute value of the personal attribute data is larger in the distribution ratio (frequency value) of each section in the distribution of the target attribute data. Thus, in (3), the attribute comparison unit 154 and the similarity calculation unit 155 calculate the similarity in the same manner as in (1) or (2) by using the distribution parameter.

  Here, the similarity calculation unit 155 may calculate the similarity using, for example, an average or a weighted average for the calculation results of the methods (1) to (3). The similarity calculation unit 155 supplies the calculated similarity to the recommendation processing unit 156.

  Next, the estimation unit 140 extracts target attribute data of a target whose similarity to the user's personal attribute data is equal to or higher than a reference similarity from the target attribute data for a plurality of targets (S270). For example, the recommendation processing unit 156 extracts at least one target attribute data having a high similarity from the target attribute data having a reference similarity or higher. As the reference similarity, a value (for example, an average value, a median value, or a value corresponding to a predetermined percentile rank) calculated using a plurality of similarities calculated for a plurality of objects by the similarity calculating unit 155 is used. It may be used or a fixed value may be used.

  Next, the estimation unit 140 determines the attribute value distribution of the attribute that is missing from the attribute values of the plurality of attributes included in the extracted target attribute data, that is, the attribute value distribution for the data missing attribute from the target attribute database 54 and the like. Obtain (S280).

  Then, the estimation unit 140 estimates the attribute value of the missing data attribute included in the user's personal attribute data based on the acquired attribute value distribution (S290). For example, the estimation unit 140 calculates an average value or mode value of attribute values from the acquired attribute value distribution, and sets the attribute value of the data missing attribute in the personal attribute data of the user as the calculated average value or mode value. presume. In this case, as an example, the estimation unit 140 detects that data is missing when the probability that the data missing attribute is one attribute value (for example, the mode value) is equal to or higher than the reference probability in the acquired attribute value distribution. The attribute value of the missing data attribute in the personal attribute data of the user may be estimated as the one attribute value.

  In order to calculate such a probability, the estimation unit 140 first acquires the distribution of attribute values for the missing data attribute, and calculates the approximate waveform by multiple regression analysis. Next, the estimation unit 140 calculates the area of the entire region surrounded by the approximate waveform and the coordinate axis of the attribute value by integrating the approximate waveform over the entire range that the attribute value can take. In addition, the estimation unit 140 integrates the approximate waveform in a section having a predetermined width including the mode value of the attribute value, that is, the attribute value that can be estimated to be missing, in the center, so that the approximate waveform and the attribute in the section are integrated. The area of the region surrounded by the value coordinate axes is calculated. Then, when the ratio of the latter area to the former area is a reference probability, for example, a fixed value such as 20% or more, the estimation unit 140 estimates the attribute value when the distribution is maximum as the missing attribute value. To do.

  Further, the estimation unit 140 stores the estimated attribute value in association with the data missing attribute of the user whose data is missing among the plurality of personal attribute data stored in the personal attribute database 52.

  Here, the estimation unit 140 determines that the distribution of the attribute value distribution for the missing data attribute included in the target attribute data of the target whose similarity to the user's personal attribute data is equal to or higher than the reference similarity is equal to or less than the reference value. As a condition, an attribute value of a data missing attribute may be estimated. As a result, when the distribution of the attribute value distribution is large and it is difficult to estimate one attribute value as the attribute value for the data missing attribute, it is possible to prevent estimation with low reliability. A fixed value may be used as a reference value for dispersion.

  Further, the estimation unit 140 may determine whether to estimate the attribute value of the data missing attribute based on the degree of influence on the presence or absence of the association with the target having the data missing attribute. For example, the estimation unit 140 estimates the attribute value when the degree of influence that the data missing attribute has on the target whose similarity to the user's personal attribute data is greater than or equal to the reference similarity is smaller than the reference influence. You may decide not to.

  As an example, when the data missing attribute is “gender” (statistical probability that the user is a woman) and the similarity to the user whose data is missing is equal to or higher than the reference similarity, “cosmetics”, As described above, since the degree of influence of “sex” on the presence / absence of association with the target “cosmetics” is large, the estimation unit 140 may perform estimation.

  On the other hand, for example, when the data missing attribute is “gender” and the target whose similarity to the user whose data is missing is equal to or higher than the standard similarity is “umbrella”, as described above, “sex” is the target “ Since the degree of influence on the presence / absence of the relationship with the “umbrella” is small, the estimation unit 140 may not perform estimation. Thereby, it is possible to prevent estimation with low reliability.

  As described above, the apparatus 100 according to the present embodiment executes the flow of steps S210 to S290, so that the attributes of a plurality of attributes included in the target attribute data of the target whose similarity to the user's personal attribute data is equal to or higher than the reference similarity. Of the values, the missing attribute value can be estimated based on the distribution of the attribute value with respect to the data missing attribute.

FIG. 8 shows personal attribute data and target attribute data similar to FIG. However, in the personal attribute data of the individual 1 shown in this figure, for convenience of explanation, the attribute value for the attribute _XN-1 is missing.

For example, when the target attribute data of the target 1 is extracted as the target attribute data of the target having the highest similarity to the personal attribute data of the individual 1 in S270, the attribute value of the attribute X _N-1 in the target attribute data The mode value “55%” is estimated as the attribute value of the attribute X _N-1 in the personal attribute data of the individual 1.

FIG. 9 shows an approximate waveform of the distribution of attributes related to the target 1 and the target 3. Here, an approximate waveform of the distribution of the attribute X _N-1 of the target _{1 and} the distribution of the attribute X _N-1 of the target 3 is shown. Further, a section having a default width “55 ± δ” including the mode value “55%” of the attribute value of the object 1 in the center and a default width “including the mode value“ 25% ”of the attribute value of the object 3 in the center. The section of “25 ± δ” is indicated by hatching. As the value of δ, for example, a value within the range of 0.1 to 5 is used.

  The area of the hatched area with respect to the area of the entire area surrounded by the waveform and the horizontal axis is 20% or more for the object 1 and less than 20% for the object 3. In such a case, when the similarity of the target attribute data of the target 1 is the highest with respect to the personal attribute data of the individual 1, the estimation unit 140 estimates the attribute value as 55%. On the other hand, when the similarity of the target attribute data of the target 3 is the highest with respect to the personal attribute data of the individual 1, the estimation unit 140 does not estimate the attribute value.

  FIG. 10 shows a third processing flow of the apparatus 100 according to the present embodiment. The apparatus 100 outputs the object which has object attribute data similar to a user's personal attribute data among several objects by performing the process of S310-S380. Here, the third processing flow is described as being executed when the browser is activated on the user terminal 20, but may be executed at another timing.

  First, the personal attribute acquisition unit 120 acquires personal attribute data for operating the user terminal 20 (S310). For example, the personal attribute acquisition unit 120 may acquire personal attribute data by a process similar to S210.

  Next, the target attribute generation unit 130 acquires target attribute data of a plurality of targets (S340). For example, the target attribute generation unit 130 acquires target attribute data for all targets stored in the target attribute database 54. The target attribute generation unit 130 supplies the acquired target attribute data to the selection unit 153 and the attribute comparison unit 154 of the output processing unit 150.

  Next, the selection unit 153 selects an attribute to be compared in the calculation of the similarity between the target attribute data of one target and the personal attribute data of one individual (S352). That is, the selection unit 153 selects an attribute having an influence level greater than or equal to the reference on the presence / absence of association with the target from among a plurality of attributes in the personal attribute data, and outputs the selection result to the attribute comparison unit 154. For example, the selection unit 153 may select an attribute by the same process as S252.

  Next, the attribute comparison unit 154 compares the personal attribute data of the user acquired in S310 and the target attribute data of each target acquired in S340 for each attribute selected in S352 (S354), and based on the comparison result, The similarity calculation unit 155 calculates the similarity between the user and each target (S360). For example, the attribute comparison unit 154 and the similarity calculation unit 155 may perform the comparison by the same processing as S254 and S260 and calculate the similarity. Then, the similarity calculation unit 155 supplies the calculated similarity to the recommendation processing unit 156.

  Next, the recommendation processing unit 156 selects and prioritizes some of the targets based on the similarity (S370). For example, the recommendation processing unit 156 ranks the objects according to the degree of similarity calculated in S360 for the plurality of objects hit in the search in S330. Then, the recommendation processing unit 156 assigns a higher priority to a higher-order target (that is, a target with a large total value).

  Next, the recommendation processing unit 156 outputs all or some of the plurality of objects according to the priority (S380). For example, the recommendation processing unit 156 generates a search result in which the targets are arranged from the top in the descending order of priority in the user terminal 20, and supplies the upper target determined to be recommended to the user to the user terminal 20. Further, the recommendation processing unit 156 may execute an emphasis process according to priority and output a plurality of targets. For example, the recommendation processing unit 156 may change the display size (font size, image size, etc.) of each target or display modification (font bolding, underlining, color change, etc.) according to the priority. May be changed.

  Thus, the apparatus 100 of this embodiment can output the object which has object attribute data with high similarity with respect to a user's personal attribute data by performing the flow of S310-S380. Therefore, a more appropriate target can be output.

  FIG. 11 shows an example of the output of the device 100. For example, when the browser is activated on the user terminal 20, one target “AB shopping center” similar to the user's personal attribute data is output on the browser.

  FIG. 12 shows a fourth processing flow of the apparatus 100 according to the present embodiment. The apparatus 100 outputs the object which has object attribute data similar to the output object among several objects by performing the process of S410-S480. Here, the fourth processing flow is described as being executed when the browser is operated on the user terminal 20, but may be executed at another timing.

  First, in response to a user's browser operation, the output processing unit 150 outputs one target on the browser (S410). For example, the output processing unit 150 may output the target output in the process of S380 again. Further, the output processing unit 150 may output a target selected by the selection instruction acquired by the instruction acquisition unit 110.

  Next, the target attribute generation unit 130 acquires the target attribute data of one target that is output and the target attribute data of another target that is different from the one target (S440). For example, the target attribute generation unit 130 acquires target attribute data for all targets stored in the target attribute database 54. The target attribute generation unit 130 supplies the acquired target attribute data to the selection unit 153 and the attribute comparison unit 154 of the output processing unit 150.

  Next, the selection unit 153 selects an attribute to be compared in calculating the similarity between the target attribute data of one target and the target attribute data of another target (S452). That is, the selection unit 153 selects an attribute having an influence level greater than or equal to a reference level on the presence / absence of association with one target among a plurality of attributes in the personal attribute data, and outputs the selection result to the attribute comparison unit 154. Further, the selection unit 153 selects an attribute whose influence on presence / absence of the relationship with respect to another target different from the one among the targets for which the target attribute data has been acquired in S440, and compares the selection result with the attribute comparison The data may be output to the unit 154. For example, the selection unit 153 may select an attribute for each of one target and another target by the same process as in S252.

  Next, the attribute comparison unit 154 compares the target attribute data of the one target and the other target acquired in S440 for each attribute (S454), and based on the comparison result, the target attribute of the one target and the other target The similarity calculation unit 155 calculates the data similarity (S460). For example, the attribute comparison unit 154 may compare the target attribute data for each attribute selected in S452.

  Here, regarding S454 and S460, first, (I) a method of calculating the similarity when the comparison is performed using the waveform of the target attribute data will be described. In this case, by using the waveform of the target attribute data of one target and the target waveform data of the other target, the similarity is calculated by the same method as (1) described with respect to S254 and S260. Good.

  Next, with regard to S454 and S460, (II) a method for calculating the degree of similarity when performing comparison using distribution data of a plurality of attributes of target attribute data will be described. In this case, in step S454, the attribute comparison unit 154 calculates a difference between the plurality of attributes by comparing the distribution of attribute values in one target and the distribution of attribute values in another target. That is, the attribute comparison unit 154 acquires distribution data of one attribute in the target attribute data of one target, and calculates an average value or a mode value for the attribute. Next, the attribute comparison unit 154 acquires the distribution data of the attribute in the target attribute data of another target, and the distribution ratio of the target attribute data in a predetermined section including the average value or the mode value calculated for one target And this is used as the degree of approximation (difference) of each attribute. By performing such processing for each attribute, the attribute comparison unit 154 calculates the degree of approximation of each attribute. Note that one object and another object may be used oppositely.

  Then, in S460, the similarity calculation unit 155 calculates the average or weighted average of the approximation degrees of the plurality of attributes by the same method as (2) described with respect to S254 and S260. The similarity of the target attribute data.

  Further, for example, in S454 and S460, the attribute comparison unit 154 and the similarity calculation unit 155 calculate the similarity according to the overlapping degree of the attribute value distribution for each attribute in the target attribute data for one target and the other target. Good. As an example, in S454, the attribute comparison unit 154, for each of a plurality of attributes, an approximate waveform of the distribution of attribute values in the target attribute data of one target and an approximate waveform of the distribution of attribute values in the target attribute data of another target Are calculated by multiple regression analysis or the like. Here, the approximate waveform of the distribution of attribute values is, for example, a waveform that approximates the distribution on a coordinate plane with the attribute value as one coordinate axis and the distribution ratio as the other coordinate axis. After calculating the approximate waveform, the attribute comparison unit 154 is surrounded for each attribute by the approximate waveform of the attribute value for one target and the coordinate axis of the attribute value, and the approximate waveform and attribute of the attribute value for the other target The area of the region surrounded by the coordinate axis of the value is calculated as an approximation. Then, the similarity calculation unit 155 calculates an average of approximations or a weighted average calculated for each attribute as an attribute suitability.

  Next, with regard to S454 and S460, (III) a similarity calculation method in the case of performing comparison using distribution parameters for each of a plurality of attributes in the target attribute data will be described.

  In this case, for example, in S454, the attribute comparison unit 154 compares the target attribute data of one target with the target attribute data of another target by comparing the distribution parameters, and calculates a difference in distribution. . As an example, in S454, the attribute comparison unit 154 calculates the attribute value (average value) of each attribute of one target attribute data and the attribute value (average value) of the corresponding attribute of another target attribute data. The difference is calculated by comparison. In S <b> 460, the similarity calculation unit 155 calculates the similarity based on the ratio of attributes whose calculation results are within a predetermined threshold among all the attributes. Note that the attribute comparison unit 154 may use the mode value of the attribute value of each attribute of the target attribute data instead of using the attribute value (average value) of each attribute of the target attribute data.

  Further, for example, in S454, the attribute comparison unit 154 generates the original distribution from the distribution parameter for each attribute instead of using the distribution parameter itself for comparison, and calculates the difference by the same method as in (II). Good. That is, first, the attribute comparison unit 154 virtually generates the original distribution of the attribute of the target attribute data from data such as the average, variance, and distribution type of the distribution of one attribute of the target attribute data. Next, the attribute comparison unit 154 detects a section having the largest distribution ratio among the distributions virtually generated for one target. Next, the attribute comparison unit 154 calculates the distribution ratio of the detected section among the virtually generated distributions for other objects, and sets this as the degree of approximation (difference) of the attribute. By performing such processing for each attribute, the attribute comparison unit 154 may calculate the degree of approximation of each attribute. Note that one object and another object may be used oppositely. Then, in S460, the similarity calculation unit 155 calculates an average or weighted average of the approximation degrees of a plurality of attributes, and sets the similarity of the target attribute data of one target and another target.

  Here, when the distribution of attribute values is included in the target attribute data, if the number of attribute value samples is small, the reliability of similarity may be lowered depending on the method used. For example, when the number of persons related to a certain object is small, there may be a case where the sample number of attribute values of each attribute is small in the target attribute data of the target. For example, if there is a missing attribute value in the personal attribute data, the number of samples may be small for the attribute value of the attribute in the target attribute data. In such a case, even if the degree of approximation is calculated using the distribution data as described in (II), the original distribution is virtually generated from the distribution parameters as described in (III) and approximated. Even if the degree is calculated, the reliability of the degree of approximation and thus the degree of similarity is lowered. Therefore, for an attribute whose number of attribute values included in the distribution of at least one of the target attribute data of one target and the other target is equal to or less than the reference, the attribute comparison unit 154 performs the target of the one target and the other target. The distribution difference may be calculated by comparing the distribution parameters of the attribute data, for example, the average value or the mode value of each attribute. A fixed value may be used as a reference for the number of samples.

  Here, the similarity calculation unit 155 may calculate the similarity using, for example, an average or a weighted average for the calculation results of the methods (I) to (III). The similarity calculation unit 155 supplies the calculated similarity to the recommendation processing unit 156.

  Next, the recommendation processing unit 156 selects and prioritizes some of the targets based on the similarity (S470). For example, the recommendation processing unit 156 ranks the targets according to the degree of similarity calculated in S460 for a plurality of targets hit in the search in S430. Then, the recommendation processing unit 156 assigns a higher priority to a higher-order target (that is, a target with a large total value).

  Next, the recommendation processing unit 156 outputs all or some of the plurality of objects according to the priority (S480). For example, the recommendation processing unit 156 generates a search result in which the targets are arranged from the top in the descending order of priority in the user terminal 20, and supplies the upper target determined to be recommended to the user to the user terminal 20. Further, the recommendation processing unit 156 may execute an emphasis process according to priority and output a plurality of targets. For example, the recommendation processing unit 156 may change the display size (font size, image size, etc.) of each target or display modification (font bolding, underlining, color change, etc.) according to the priority. May be changed.

  As described above, the apparatus 100 according to the present embodiment outputs a target having target attribute data having high similarity to the target attribute data of one target that is output by executing the flow of S410 to S480. Can do. Therefore, a more appropriate target can be output.

FIG. 13 shows an approximate waveform of the distribution of attributes related to the target 1 and the target 2. Here, an approximate waveform of the distribution of the attribute x ₁ distribution and target 2 attributes x ₁ of the subject 1. As indicated by hatching in the figure, the approximate waveforms of the distributions of the target 1 and the target 2 partially overlap. As described in (II) regarding the processing of S454 and S460, when calculating the degree of approximation of each attribute according to the degree of overlapping of the attribute value distribution for each attribute in the target attribute data, and thus the similarity between the targets, The attribute comparison unit 154 calculates the overlapping area of the distribution as the degree of approximation of the attribute x ₁ of the target 1 and the target 2.

FIG. 14 shows an example of the similarity between the object 1 and the object 2 calculated by (II). As shown in the figure, the attribute comparison unit 154 calculates the closeness of the target 1 and the target 2 for N attributes of the attribute x ₁ , the attribute x ₂ ,..., The attribute x _N. For example, the attribute comparison unit 154 may calculate the degree of approximation of the attribute x ₁ as 25 from the area of the distribution overlap in FIG. The similarity calculation unit 155 may calculate the final similarity of the target 1 and the target 2 by averaging or weighting the approximations of a plurality of attributes.

  FIG. 15 shows an example of the output of the apparatus 100. For example, when the user applies for a regular subscription of the target “ABC fashion magazine” at the user terminal 20, an application completion screen for the subscription of “ABC fashion magazine” is displayed on the browser in S410. Then, another target “XY cosmetics” similar to the one target “ABC fashion magazine” is output on the browser.

  FIG. 16 shows a fifth processing flow of the apparatus 100 according to the present embodiment. By executing the processing of S510 to S580, the apparatus 100 performs output in which the target is prioritized so that the target attribute data matches the user's personal attribute data in accordance with the selection instruction input by the user. Here, a case where the user inputs a search key as a selection instruction will be described.

  First, the personal attribute acquisition unit 120 acquires personal attribute data for operating the user terminal 20 (S510). For example, the personal attribute acquisition unit 120 may acquire personal attribute data by a process similar to S210.

  Next, the instruction acquisition unit 110 acquires a search key as a selection instruction from the user (S520). The search key is used to select only a part of the objects related to the search key from the countless objects. For example, the instruction acquisition unit 110 supplies a search screen to the user terminal 20 and causes the user terminal 20 to acquire keywords such as general names or abbreviations of products / services related to the target that the user wants to select. The instruction acquisition unit 110 supplies the acquired search key to the target fitness calculation unit 152 of the output processing unit 150.

  Next, the target fitness calculation unit 152 executes a search using the search key, and acquires a target that hits the search key (S530). For example, the target fitness calculation unit 152 requests the search server 30 to execute search using the search key. In response to this, the search server 30 performs a search by using a synonym, a synonym, or a related word of a target that matches a part or all of the search key, a part or all of the search key, and is identical to the search key or Data related to a target that is more than a predetermined degree is extracted from the search database 32. For example, the search server 30 may extract a web page that includes a keyword corresponding to the search key in the title and / or content. As an example, when the search key is a specific product, the search server extracts a plurality of web pages related to the product as information related to the target. The search server 30 supplies one or a plurality of targets hit in the search to the target attribute generation unit 130 and the target fitness calculation unit 152.

  Next, the target attribute generation unit 130 acquires target attribute data of each target hit in the search of S530 (S540). For example, the target attribute generation unit 130 searches the target attribute database 54 for each target hit in the search (for example, a product / service name included in the web page), and acquires target attribute data corresponding to each target. Thereby, the target attribute generation unit 130 acquires the target attribute data of each target generated based on the personal attribute data of a plurality of individuals related to each target in S150. The target attribute generation unit 130 supplies the acquired target attribute data of each target to the selection unit 153 and the attribute comparison unit 154 of the output processing unit 150.

  Next, the target fitness level calculation unit 152 calculates a target fitness level that is the fitness level of each target with respect to the search key (S550). For example, the target fitness calculation unit 152 matches the character string of the search key, the name of each web page hit by the search key, the text, the number of occurrences of the same or related character string as the search key in the meta tag, and the character string match The target fitness is calculated based on the ratio, the arrangement position, and / or the usage mode. As an example, the target suitability calculation unit 152 has a higher degree of matching between a search key and a character string in a web page corresponding to each target, a higher usage frequency, and / or a more forward use. You may calculate the object fitness of each object highly. The target suitability calculation unit 152 may calculate the target suitability using various keyword matching methods that have already been used. Here, it is assumed that the target fitness calculation unit 152 calculates the target fitness of each target on a scale of 0 to 1. Instead of the calculation by the target fitness calculation unit 152, the search server 30 may calculate the target fitness when searching for the target, and supply the calculated target fitness to the target fitness calculation unit 152. The target suitability calculation unit 152 supplies the target suitability of each target to the recommendation processing unit 156.

  Next, the selection unit 153 selects an attribute to be compared in calculating the similarity between the target attribute data of one target and the personal attribute data of one individual (S552). That is, the selection unit 153 selects an attribute having an influence level greater than or equal to the reference on the presence / absence of association with the target from among a plurality of attributes in the personal attribute data, and outputs the selection result to the attribute comparison unit 154. For example, the selection unit 153 may select an attribute by the same process as S252.

  Next, the attribute comparison unit 154 compares the personal attribute data of the user acquired in S510 and the target attribute data of each target acquired in S540 for each attribute (S554). Based on the comparison result, the user and each target are compared. Are calculated by the similarity calculation unit 155 (S560). For example, the attribute comparison unit 154 and the similarity calculation unit 155 may perform the comparison by the same processing as S254 and S260 and calculate the similarity. Then, the similarity calculation unit 155 supplies the calculated similarity to the recommendation processing unit 156.

  Next, the recommendation processing unit 156 selects and prioritizes some of the targets based on the target suitability and the similarity (S570). For example, the recommendation processing unit 156 may prioritize targets by the same processing as S370.

  Next, the recommendation processing unit 156 outputs all or some of the plurality of objects according to the priority (S580). For example, the recommendation processing unit 156 may prioritize targets by the same process as S380.

  As described above, the apparatus 100 according to the present embodiment searches the target using the search key input by the user by executing the flow of S510 to S580, and the target fitness of the target selected by the search key and the user. Based on both the similarity of the objects, the retrieved objects can be provided to the user in an appropriate manner.

  FIG. 17 shows an example of the selection instruction from the user and the output of the apparatus 100. For example, the apparatus 100 receives “vanilla ice” as a search key from the user via the user terminal 20 in S520. Then, the apparatus 100 selects a target “ABC vanilla bar”, a target “vanilla ice soft”, a target “ice vanilla super”, and the like related to “vanilla ice”. Furthermore, the device 100 selects the target “ABC vanilla bar”, the target “vanilla ice software”, and the target “ABC vanilla bar” according to the target matching degree of each target selected with the search key “vanilla ice”, and the similarity between the user and each target. The target “ice vanilla super” or the like is prioritized and output together with the URL information of each target. Here, the target “ABC vanilla bar” is output with the highest priority, and then the target “vanilla ice soft” is output with the highest priority.

  FIG. 18 shows an example of target prioritization by the recommendation processing unit 156 in FIG. Here, the apparatus 100 calculates the target fitness and similarity for each of the targets (such as “ABC vanilla bar”) hit in S530. As shown in this example, the recommendation processing unit 156 may further set an output priority for each target. For example, the apparatus 100 may set the output priority based on the amount of money presented from each target manufacturer or the like. As illustrated, the recommendation processing unit 156 may calculate the total of the target fitness, similarity, and output priority as a total score. As illustrated in FIG. 17, the recommendation processing unit 156 displays each target in the search result in the order of the size of the total score.

  FIG. 19 shows a sixth processing flow of the apparatus 100 according to the present embodiment. The apparatus 100 performs the processes of S610 to S690, and outputs a target that matches the user's personal attribute data and the target attribute data selected by the user according to the selection instruction input by the user. Here, a case where the user inputs an instruction to select a target from two or more selection targets included in the screen presented to the user will be described as a selection instruction.

  First, the personal attribute acquisition unit 120 acquires personal attribute data for operating the user terminal 20 (S610). The personal attribute acquisition unit 120 may acquire personal attribute data by a process similar to S210.

  Next, the instruction acquisition unit 110 supplies a screen including a plurality of selection targets to the user terminal 20 (S620). For example, the instruction acquisition unit 110 includes a plurality of selection targets that can be selected by the user terminal 20 via the network 22 (for example, a target to which a hyperlink or a button that jumps to a web page of the next hierarchy is given). Provide a web page screen.

  Next, the instruction acquisition unit 110 acquires a selection instruction indicating a selection target selected by the user from two or more selection targets included in the screen presented to the user (S630). For example, the instruction acquisition unit 110 acquires, from the user terminal 20, information corresponding to one target selected by the user from a plurality of selection targets (for example, clicking on a target hyperlink). The instruction acquisition unit 110 supplies the acquired selection instruction to the target attribute generation unit 130.

  Next, the target attribute generation unit 130 acquires target attribute data corresponding to each of a plurality of selection targets included in the screen presented to the user (S640). That is, the target attribute generation unit 130 acquires the target attribute data generated based on the personal attribute data of a plurality of individuals related to each target in S150 for a plurality of selection targets presented to the user. Here, the target attribute generation unit 130 acquires a plurality of selection target attribute data included in all or some of the screens that may be displayed after the screen displayed in S620 by the user's selection in S630. It's okay.

  For example, the target attribute generation unit 130 may acquire target attribute data of a plurality of selection targets to be displayed on the next screen after selecting a target related to the selection instruction on the screen displayed in S620. Further, the target attribute generation unit 130 is a screen that is displayed after the selection of each of a plurality of selection targets included in the screens after the next screen (for example, the next screen, the next next screen, etc.). You may acquire the object attribute data of all the selection targets to display. The target attribute generation unit 130 may acquire target attribute data of all the selection targets that may be displayed on the screen subsequent to the target related to the selection instruction (also referred to as a lower screen), or the lowest level. All selection target attribute data that may be displayed on a lower screen other than the above screen (a screen that does not include a selectable selection target) may be acquired. As described above, the target attribute generation unit 130 acquires a plurality of selection target attribute data included in the lower hierarchy of the selection target related to the screen being displayed.

  In addition, the target attribute generation unit 130 may acquire a plurality of selection target attribute data included in a screen that may be displayed after the screen displayed in S620 before the user's selection in S630. The target attribute generation unit 130 supplies the acquired target attribute data of each target to the selection unit 153 and the attribute comparison unit 154 of the output processing unit 150.

  Next, the selection unit 153 selects an attribute to be compared in the calculation of the similarity between the target attribute data of one target and the personal attribute data of one individual (S652). In other words, the selection unit 153 selects an attribute having an influence level greater than or equal to a reference level on the presence / absence of association with the target from among a plurality of attributes in the personal attribute data, and outputs the selection result to the attribute comparison unit 154 (S652). For example, the selection unit 153 may select an attribute by the same process as S252.

  Next, the attribute comparison unit 154 compares the personal attribute data of the user acquired in S610 and the target attribute data of each target acquired in S640 for each attribute (S654). Based on the comparison result, the user and each target are compared. Are calculated by the similarity calculation unit 155 (S660). For example, the attribute comparison unit 154 and the similarity calculation unit 155 may perform the comparison by the same processing as S254 and S260 and calculate the similarity. Then, the similarity calculation unit 155 supplies the calculated similarity to the recommendation processing unit 156.

  Next, the recommendation processing unit 156 prioritizes each selection target to be included in the next screen to be presented to the user according to the selection target selected by the user based on the comparison result comparing the personal attribute data and the target attribute data. (S670). For example, the recommendation processing unit 156 assigns a high priority to a plurality of selection targets to be displayed on and after the next screen calculated in S660 in descending order of similarity.

  Next, the recommendation processing unit 156 outputs the next selection target according to the priority (S680). For example, the recommendation processing unit 156 may determine that the selection target with the highest priority (the highest priority selection target) is recommended to the user, and supply a screen related to this target to the user terminal 20 as a display screen. The recommendation processing unit 156 may display a plurality of selection targets (highest priority next hierarchy selection targets) of the next hierarchy that is the highest priority selection target in the highest priority selection target screen. Here, the recommendation processing unit 156 may output a plurality of highest-priority next-layer selection targets in order of highest priority of the highest-priority next-layer selection target, or according to the priority of the highest-priority next-layer selection target. The display size or display modification of each object may be changed.

  Next, the instruction acquisition unit 110 determines whether or not the current display screen includes a next selectable selection target (S690). If the current display screen includes a selectable selection target, the instruction acquisition unit 110 returns the process to S630 and accepts the next selection instruction from the user terminal 20 again. If not, the instruction acquisition unit 110 may end the process.

  As described above, the apparatus 100 according to the present embodiment executes the flow of steps S610 to S690, so that the user can select the next screen based on the similarity from the selection targets subsequent to the selection target selected from the plurality of selection targets. The selection target to be displayed on the screen is determined. Accordingly, the apparatus 100 can support, for example, the user following a link and aiming at a target page using the user's personal attribute data as a hint.

  FIG. 20 shows an example of a hierarchical structure to be selected. The apparatus 100 may be based on a plurality of selection targets having a hierarchical structure according to FIG. 20 when executing the sixth processing flow. In response to the user selecting one of a plurality of selection targets on the screen of the user terminal 20 by clicking or tapping, the device 100 is included in any one of the layers below the selected selection target. A screen including a selection target is supplied to the user terminal 20.

  As shown in the figure, the selection target “weekday afternoon” exists in the first hierarchy, and the selection targets “rest”, “search for nearby stations”, and “taxi arrangement” exist in the second hierarchy after “weekday afternoon”. There are selection targets “Convenience Store Search”, “Coffee Shop Search” and “News Search” in the third hierarchy next to “Holiday Break”, and “Café Store A” in the fourth hierarchy next to “Coffee Shop Search”. , “Café shop B” and “Coffee shop C” exist. In FIG. 20, there may be selection targets other than those illustrated. For example, although not shown, there may be further selection targets in one or a plurality of lower hierarchies such as “search for nearby stations” and “arrange a taxi”.

  FIG. 21 shows an example of a display screen in the first hierarchy. For example, the instruction acquisition unit 110 provides the user terminal 20 with a display screen related to “weekday afternoon” in the first layer in the first process of S620. Here, on the display screen, three selection targets such as “break” relating to the second hierarchy are displayed as selectable. Here, for example, when the user clicks on the “Holiday” link and inputs the “Holiday” to the user terminal 20 as a selection instruction, conventionally, “Convenience store search” on the next layer of “Holiday”, etc. The selection target is displayed as a selectable selection target.

  FIG. 22 shows an example of a display screen after inputting a selection instruction. According to the apparatus 100 of the present embodiment, for example, when the user clicks on the “break” link in FIG. 21 and inputs the “break” to the user terminal 20 as a selection instruction in S630, the apparatus 100 displays “break” in S660. The degree of similarity between all selection targets in the lower hierarchy (such as “convenience store search”, “coffee shop search”, “news search”, “cafe store A”, “cafe store B”, and “coffee specialty store C”) In step S670, “coffee shop search” having the highest similarity is determined as the highest priority selection target. In step S680, the apparatus 100 outputs a display screen related to “search for a coffee shop” illustrated in FIG. Here, the apparatus 100 includes a plurality of selection targets (such as “cafe shop A”) in the next level in a selectable state on the display screen related to “coffee shop search”.

  Normally, the user cannot display the display screen related to the “coffee shop search” unless “coffee shop search” is selected from a plurality of selection targets (“convenience store search” or the like) in the third hierarchy. According to the apparatus 100, the selection can be omitted and a “coffee shop search” screen can be provided to the user as soon as possible. As a result, the apparatus 100 also puts forward a web page suitable for the user according to the degree of similarity between the user and each link item (target) even when the user follows a predetermined link hierarchy on the web page. Can be displayed.

  FIG. 23 shows an example of the configuration of a computer 1900 according to this embodiment. A computer 1900 according to this embodiment functions as the apparatus 100. In addition, the computer connected to the network 22 as the search server 30 and / or the user terminal 20 and the computer connected to the network 42 as the store terminal 40 and / or the history management server 46 have the same configuration as the computer 1900. You may take it.

  A computer 1900 according to this embodiment is connected to a CPU peripheral unit having a CPU 2000, a RAM 2020, a graphic controller 2075, and a display device 2080 that are connected to each other by a host controller 2082, and to the host controller 2082 by an input / output controller 2084. An input / output unit having a communication interface 2030, a hard disk drive 2040, and a DVD drive 2060, a ROM 2010 connected to an input / output controller 2084, a flash memory drive 2050, and a legacy input / output unit having an input / output chip 2070. .

  The host controller 2082 connects the RAM 2020 to the CPU 2000 and the graphic controller 2075 that access the RAM 2020 at a high transfer rate. The CPU 2000 operates based on programs stored in the ROM 2010 and the RAM 2020 and controls each unit. The graphic controller 2075 acquires image data generated by the CPU 2000 or the like on a frame buffer provided in the RAM 2020 and displays it on the display device 2080. Instead of this, the graphic controller 2075 may include a frame buffer for storing image data generated by the CPU 2000 or the like.

  The input / output controller 2084 connects the host controller 2082 to the communication interface 2030, the hard disk drive 2040, and the DVD drive 2060, which are relatively high-speed input / output devices. The communication interface 2030 communicates with other devices via a network by wire or wireless. The communication interface functions as hardware that performs communication. The hard disk drive 2040 stores programs and data used by the CPU 2000 in the computer 1900. The DVD drive 2060 reads a program or data from the DVD 2095 and provides it to the hard disk drive 2040 via the RAM 2020.

  The input / output controller 2084 is connected to the ROM 2010, the flash memory drive 2050, and the relatively low-speed input / output device of the input / output chip 2070. The ROM 2010 stores a boot program that the computer 1900 executes at startup and / or a program that depends on the hardware of the computer 1900. The flash memory drive 2050 reads a program or data from the flash memory 2090 and provides it to the hard disk drive 2040 via the RAM 2020. The input / output chip 2070 connects the flash memory drive 2050 to the input / output controller 2084 and inputs / outputs various input / output devices via, for example, a parallel port, a serial port, a keyboard port, a mouse port, and the like. Connect to controller 2084.

  A program provided to the hard disk drive 2040 via the RAM 2020 is stored in a recording medium such as a flash memory 2090, a DVD 2095, or an IC card and provided by a user. The program is read from the recording medium, installed in the hard disk drive 2040 in the computer 1900 via the RAM 2020, and executed by the CPU 2000.

  A program installed in the computer 1900 and causing the computer 1900 to function as at least a part of the apparatus 100 includes an instruction acquisition module, a personal attribute acquisition module, a target attribute acquisition module, an output processing module, a target fitness calculation module, a similarity calculation module, And at least one of the target output modules. These programs or modules may work on the CPU 2000 or the like to cause the computer 1900 to function as the instruction acquisition unit 110, the personal attribute acquisition unit 120, the target attribute generation unit 130, the estimation unit 140, and the output processing unit 150, respectively.

  When the information processing described in these programs is read into the computer 1900, it works on the CPU 2000, which is a specific means in which the software and the various hardware resources described above cooperate, and instructs the computer 1900 to It functions as the acquisition unit 110, the personal attribute acquisition unit 120, the target attribute generation unit 130, the estimation unit 140, and the output processing unit 150. And the specific apparatus 100 according to the intended use is constructed | assembled by implement | achieving the calculation or processing of the information according to the intended use of the computer 1900 in this embodiment by these specific means.

  As an example, when communication is performed between the computer 1900 and an external device or the like, the CPU 2000 executes a communication program loaded on the RAM 2020 and executes a communication interface based on the processing content described in the communication program. A communication process is instructed to 2030. Under the control of the CPU 2000, the communication interface 2030 reads transmission data stored in a transmission buffer area or the like provided on a storage device such as the RAM 2020, the hard disk drive 2040, the flash memory 2090, or the DVD 2095, and transmits it to the network. Alternatively, the reception data received from the network is written into a reception buffer area or the like provided on the storage device. As described above, the communication interface 2030 may transfer transmission / reception data to / from the storage device by a DMA (direct memory access) method. Instead, the CPU 2000 transfers the storage device or the communication interface 2030 as a transfer source. The transmission / reception data may be transferred by reading the data from the data and writing the data to the communication interface 2030 or the storage device of the transfer destination.

  Further, the CPU 2000 DMAs all or necessary portions of files or databases stored in an external storage device such as the hard disk drive 2040, the DVD drive 2060 (DVD 2095), the flash memory drive 2050 (flash memory 2090), or the like. The data is read into the RAM 2020 by transfer or the like, and various processes are performed on the data on the RAM 2020. Then, CPU 2000 writes the processed data back to the external storage device by DMA transfer or the like. In such processing, since the RAM 2020 can be regarded as temporarily holding the contents of the external storage device, in the present embodiment, the RAM 2020 and the external storage device are collectively referred to as a memory, a storage unit, or a storage device.

  For example, the storage unit of the device 100 includes data received from and / or provided to the instruction acquisition unit 110, the personal attribute acquisition unit 120, the target attribute generation unit 130, the estimation unit 140, and the output processing unit 150, and personal attributes. The database 52 and the database in the target attribute database 54 may be stored as appropriate. For example, the storage unit may receive and store personal attribute data that the personal attribute acquisition unit 120 supplies to the output processing unit 150. The storage unit may store the similarity degree calculated by the similarity degree calculation unit 155 and the like.

  Various types of information such as various programs, data, tables, and databases in the present embodiment are stored on such a storage device and are subjected to information processing. Note that the CPU 2000 can also store a part of the RAM 2020 in the cache memory and perform reading and writing on the cache memory. Even in such a form, the cache memory bears a part of the function of the RAM 2020. Therefore, in the present embodiment, the cache memory is also included in the RAM 2020, the memory, and / or the storage device unless otherwise indicated. To do.

  In addition, the CPU 2000 performs various operations, such as various operations, information processing, condition determination, information search / replacement, etc., described in the present embodiment, specified for the data read from the RAM 2020 by the instruction sequence of the program. Is written back to the RAM 2020. For example, when performing the condition determination, the CPU 2000 determines whether or not the various variables shown in the present embodiment satisfy the conditions such as large, small, above, below, equal, etc., compared to other variables or constants. If the condition is satisfied (or not satisfied), the program branches to a different instruction sequence or calls a subroutine.

  Further, the CPU 2000 can search for information stored in a file or database in the storage device. For example, in the case where a plurality of entries in which the attribute value of the second attribute is associated with the attribute value of the first attribute are stored in the storage device, the CPU 2000 displays the plurality of entries stored in the storage device. The entry that matches the condition in which the attribute value of the first attribute is specified is retrieved, and the attribute value of the second attribute that is stored in the entry is read, thereby associating with the first attribute that satisfies the predetermined condition The attribute value of the specified second attribute can be obtained.

  In addition, when a plurality of elements are listed in the description of the embodiment, elements other than the listed elements may be used. For example, when “X executes Y using A, B, and C”, X may execute Y using D in addition to A, B, and C.

  As mentioned above, although this invention was demonstrated using embodiment, the technical scope of this invention is not limited to the range as described in the said embodiment. It will be apparent to those skilled in the art that various modifications or improvements can be added to the above-described embodiment. It is apparent from the scope of the claims that the embodiments added with such changes or improvements can be included in the technical scope of the present invention.

  The order of execution of each process such as operations, procedures, steps, and stages in the apparatus, system, program, and method shown in the claims, the description, and the drawings is particularly “before” or “prior to”. It should be noted that the output can be realized in any order unless the output of the previous process is used in the subsequent process. Regarding the operation flow in the claims, the description, and the drawings, even if it is described using “first”, “next”, etc. for convenience, it means that it is essential to carry out in this order. It is not a thing.

DESCRIPTION OF SYMBOLS 10 System 20 User terminal 22 Network 30 Search server 32 Search database 40 Store terminal 42 Network 44 Purchasing history database 46 History management server 52 Personal attribute database 54 Target attribute database 100 Device 110 Instruction acquisition part 120 Personal attribute acquisition part 130 Target attribute generation Unit 140 estimation unit 150 output processing unit 152 target fitness calculation unit 153 selection unit 154 attribute comparison unit 155 similarity calculation unit 156 recommended processing unit 1900 computer 2000 CPU
2010 ROM
2020 RAM
2030 Communication interface 2040 Hard disk drive 2050 Flash memory drive 2060 DVD drive 2070 Input / output chip 2075 Graphic controller 2080 Display device 2082 Host controller 2084 Input / output controller 2090 Flash memory 2095 DVD

Claims (13)

A personal attribute acquisition unit that acquires personal attribute data including a plurality of attribute values possessed by an individual with respect to a plurality of attributes;
Each attribute data of the plurality of attributes in the target attribute data of the target based on a distribution of attribute values for each of the plurality of attributes in the plurality of personal attribute data for a plurality of individuals related to a certain target A target attribute generation unit for generating
An attribute comparison unit that compares the target attribute data of one target with at least one of the personal attribute data of one individual and the target attribute data of another target for each of the plurality of attributes;
Based on the comparison result by the attribute comparison unit, a similarity calculation unit that calculates the similarity of at least one of the one object and the one individual, and the one object and the other object;
A device comprising:   The attribute comparison unit uses the distribution parameter for each of the plurality of attributes in the target attribute data of the one target, and the personal attribute data of the one individual and the target attribute data of the other target The apparatus according to claim 1, wherein the comparison is performed.   In the comparison between the target attribute data of the one target and the target attribute data of the other target, the attribute comparison unit, for each of the plurality of attributes, the distribution of attribute values in the one target and the The apparatus according to claim 1, wherein a difference in distribution of attribute values among other objects is calculated.   The attribute comparison unit is configured to determine whether the number of attribute value samples included in at least one distribution of the target attribute data of the one target and the target attribute data of the other target is equal to or less than a reference. The apparatus according to claim 3, wherein distribution parameters of the target attribute data of a target and the target attribute data of the other target are compared with each other.   When the attribute value for one attribute in the personal attribute data of the one individual is missing, the one attribute included in the target attribute data of the target whose similarity to the one individual is equal to or higher than a reference similarity The apparatus according to any one of claims 1 to 4, further comprising: an estimation unit that estimates an attribute value of the one attribute in the personal attribute data of the one individual based on a distribution of attribute values with respect to the attribute.   In the distribution of attribute values for the one attribute included in the target attribute data of a target whose similarity to the one individual is equal to or higher than a reference similarity, the estimation unit is configured to determine the personal attribute data in the personal attribute data of the one individual. The apparatus according to claim 5, wherein the one attribute in the personal attribute data of the one individual is estimated as the one attribute value when a probability that the one attribute becomes one attribute value is equal to or higher than a reference probability. .   The estimation unit is configured on the condition that a distribution of attribute value distributions for the one attribute included in the target attribute data of the target whose similarity to the one individual is equal to or higher than a reference similarity is equal to or less than a reference value. The apparatus according to claim 5, wherein an attribute value of the one attribute in the personal attribute data of the one individual is estimated based on the distribution.   In the case where an attribute value for the one attribute in the personal attribute data of the one individual is missing, the estimation unit determines whether there is a relationship with a target whose similarity to the one individual is equal to or higher than a reference similarity. 6. The apparatus according to claim 5, wherein whether or not to estimate an attribute value of the one attribute is determined based on an influence level of the one attribute. Among the plurality of attributes, further comprising a selection unit that selects an attribute having an influence level or more with respect to the presence or absence of the association with the one target,
The attribute comparison unit, for each attribute selected by the selection unit among the plurality of attributes, the target attribute data of the one target, the personal attribute data of the one individual, and the target of the other target The apparatus according to claim 1, wherein the apparatus compares at least one of the attribute data. The selection unit further selects an attribute whose influence on the presence or absence of association with the other target is greater than or equal to a reference,
The attribute comparison unit, for each of the attribute having an influence on the presence or absence of the relationship with respect to the one target is higher than the standard and the attribute having an influence on the presence or absence of the relationship with respect to the other target is higher than the reference The apparatus according to claim 9, wherein the target attribute data of one target is compared with the target attribute data of the other target. The target attribute generation unit includes the plurality of pieces of the target attribute data for the product or service as the target in the plurality of personal attribute data for the plurality of individuals that have purchased or collected information about the product or service. Based on the distribution of attribute values for each attribute,
The apparatus determines whether or not to recommend the one product or service to the one individual based on the target attribute data for the one product or service and the personal attribute data of the one individual. The apparatus according to claim 1, further comprising a recommended processing unit. A method performed by a computer,
A personal attribute acquisition stage for acquiring personal attribute data including a plurality of attribute values that an individual has for a plurality of attributes;
Each attribute data of the plurality of attributes in the target attribute data of the target based on a distribution of attribute values for each of the plurality of attributes in the plurality of personal attribute data for a plurality of individuals related to a certain target The target attribute generation stage for generating
An attribute comparison step of comparing the target attribute data of one target with at least one of the personal attribute data of one individual and the target attribute data of another target for each of the plurality of attributes;
A similarity calculation step of calculating the similarity of at least one of the one object and the one person, and the one object and the other object, based on the comparison result of the attribute comparison step;
A method comprising: When executed, the computer
A personal attribute acquisition unit that acquires personal attribute data including a plurality of attribute values possessed by an individual with respect to a plurality of attributes;
Each attribute data of the plurality of attributes in the target attribute data of the target based on a distribution of attribute values for each of the plurality of attributes in the plurality of personal attribute data for a plurality of individuals related to a certain target A target attribute generation unit for generating
An attribute comparison unit that compares the target attribute data of one target with at least one of the personal attribute data of one individual and the target attribute data of another target for each of the plurality of attributes;
Based on the comparison result by the attribute comparison unit, a similarity calculation unit that calculates the similarity of at least one of the one object and the one individual, and the one object and the other object;
Program to make it work.

Images