JP2015005013A - Recommendation condition optimization device and method thereof, recommendation system and computer program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a recommendation condition optimization device or the like capable of efficiently selecting recommendation conditions especially providing recommendation effects from among a plurality of recommendation conditions in the case of optimizing the recommendation conditions, and optimizing the recommendation conditions to be optimized on the basis of the selected recommendation conditions.SOLUTION: A recommendation condition optimization device 1 includes: a similarity calculation part 2 for calculating similarity between a recommendation content associated with a first recommendation content ID for identifying first recommendation conditions and a recommendation content associated with a second recommendation content ID; and a recommendation condition changing part 3 for obtaining the specific second recommendation content ID similar to the recommendation content associated with the first recommendation content ID from among the calculated similarity, and for generating third recommendation conditions on the basis of the first recommendation conditions associated with the first recommendation content ID and the second recommendation conditions associated with the second recommendation content ID.

Description

  The present invention relates to a technical field for optimizing a recommendation condition in a recommendation system for distributing content, for example.

  In a service (hereinafter referred to as “information distribution service”) that distributes various types of information and advertisements (hereinafter referred to as “mobile advertisement”) to mobile terminals such as mobile terminals, users who are interested in the distributed advertisements It is important to select a delivery destination.

  The reason for this is that an advertisement distributed by an information distribution service may be annoying to the user when the advertisement is not of interest to the user, for example, it may be treated as a commonly known spam message. is there.

  Therefore, when an advertisement that is not of interest to the user is continuously distributed, the information distribution service is not used by the user, and there is a possibility that the trust of customers such as companies requesting distribution of information may be impaired.

  On the other hand, as an advertising business providing an information distribution service, it is desired to distribute advertisements to more users in order to increase the advertising effect.

  However, when advertisements are distributed to many users, users who are not interested in the distributed advertisements are not interested in the information distribution service itself as described above. Therefore, the advertising effect of the information distribution service may be reduced in the long term.

  As described above, in the information distribution service, it is important to select a user to be distributed so that the advertising effect is maximized for the user and the advertising business.

  In the following description, for convenience of explanation, a service that provides (recommends) various information such as mobile advertisements in the hope that the user will take some action is referred to as a “recommended service” (hereinafter referred to as each “recommended service”). The same applies to the embodiment).

  The recommendation service can objectively evaluate the advertising effect by the index indicating “success rate” and “success number”.

  Here, the “success rate” is the ratio of the number of users who take the action expected by the recommended service provider to the number of users recommended by the recommended service.

  In the recommendation service, a higher “success rate” indicates that unnecessary information is not provided to the user. In other words, when the “success rate” is high, the information can be said to be information having high value for the user.

  The “success number” is the number of users who take the action expected by the recommended service provider among the users who are recommended by the recommended service.

  In the recommendation service, the greater the “success number”, the greater the advertising effect on the recommended information. That is, when the “success number” is large, it can be said that the information is valuable information for the advertising agency.

  Next, the recommendation effect (hereinafter referred to as “recommendation effect”), which is an index indicating value in the recommendation service, can be expressed by the function of the “success rate” and “success number” described above.

  When the “success rate” is high or the “success number” is large, such a function has a characteristic that the recommendation effect becomes large. However, in the function, “success rate” and “success number” are in a trade-off relationship.

  More specifically, for example, when a coupon for expecting to visit a restaurant (restaurant coupon) is distributed to a user in the vicinity of the restaurant, the user is physically close to the restaurant. You can easily stop by.

  Therefore, the “success rate” is high. On the other hand, since the distribution destination of the coupon is limited to the user in the vicinity of the restaurant, the “success number” decreases.

  For example, when distributing such a coupon to a user away from a restaurant, the user may be unable to stop easily because the user is physically away from the restaurant.

  Therefore, the “success rate” is low. On the other hand, since the coupon is distributed to more users, the “success number” increases.

  As described above, the recommended service is a service having high value for the user and the advertising agency when the “success rate” is high and the “success number” is large.

  However, with the recommendation service, it is difficult to increase both the “success rate” and the “success number”. That is, the recommendation service has a problem that the recommendation effect cannot be improved efficiently.

  Here, for example, Patent Document 1 discloses a technique related to a service provision condition optimization method, a service search apparatus, and a service provision condition optimization system as related techniques existing prior to the application of the present application.

  More specifically, the service provision condition optimization method disclosed in Patent Document 1 is a condition for determining whether or not to provide a service for delivering an advertisement to a user (hereinafter referred to as “service provision condition”). Can be optimized.

  Further, when providing a service, the service provision condition optimization method collects information indicating an effect on the provided service and analyzes the collected information.

  The service provision condition optimization method changes a service provision condition composed of one or more attribute conditions based on the analyzed information so that the effect on the provided service is enhanced.

  Here, in Patent Document 1, information indicating a user's whereabouts (position), time, surrounding conditions, and the like is represented by a set of attributes. FIG. 21 is a diagram specifically illustrating a configuration of information indicating a state in Patent Document 1.

  In FIG. 21, the first column represents an attribute name (attribute name). The second column represents values for attribute names (attribute values). That is, the information indicating the state is information in which an attribute name and an attribute value are associated with each other.

  The attribute condition includes an attribute name, an attribute value, and information indicating a relationship between them. More specifically, as an example, when the attribute condition indicates that “the temperature is 30 ° C. (Celsius) or higher”, the attribute name is “temperature”. The attribute value is “30 ° C.”. The relationship between the attribute name and the attribute value is “≧ (or more)”.

  That is, “change service provision conditions” in Patent Document 1 is to change attribute values described as attribute conditions that constitute service provision conditions.

  As described above, the technique disclosed in Patent Document 1 reduces the trouble of continuously analyzing the service provided by the service provider to which user and updating the service provision conditions. can do.

  Patent Document 2 discloses a technology relating to a history collection device, a recommendation device, a history collection method, and a computer-readable storage medium.

  The history collection method disclosed in Patent Document 2 is based on a recommendation condition that is a basis for determining whether to recommend content such as a document, an image, and a video, and an expectation that defines a user state expected after the content is recommended Link the condition.

  Furthermore, the history collection method acquires state information indicating the state of the user after the content is recommended, and compares the acquired state information with an expected condition.

  The history collection method determines that the recommendation is successful when the state information matches the expected condition as a result of the comparison.

  That is, the history collection method can collect the recommendation effect without explicitly inputting whether or not the recommendation is successful by the user.

  Thus, by combining the above-described Patent Documents 1 and 2, for example, the service provision condition optimization method disclosed in Patent Document 1 is based on the recommendation effect collected in the history collection method disclosed in Patent Document 2. Thus, the recommendation condition can be optimized.

  Next, Patent Document 3 discloses a technique related to a recommendation system, a recommendation method, and a recommendation program.

  The recommendation system disclosed in Patent Literature 3 is interested in the user by using information on how the user has performed the target of content, products, etc. that the user is interested in. A subject can be recommended.

  Patent Document 4 discloses a technique related to an advertisement distribution method and an advertisement distribution server. The advertisement distribution server collects attribute information such as product names, serial numbers, and manufacturer names of various devices such as televisions and digital cameras, and information indicating the usage status of various devices.

  Furthermore, the advertisement distribution server distributes advertisement data related to various devices to the user based on the collected attribute information and information indicating the usage status.

  Even if the advertisement distribution server cannot collect the attribute information and the information indicating the usage situation, the attribute information and the information indicating the usage situation are present when there are more than a predetermined percentage of users to whom the advertisement data has been distributed. The advertisement data can be distributed to users who cannot collect data.

  Patent Document 5 discloses a technique related to an advertisement distribution method based on keywords and conditions. The advertisement distribution method acquires a keyword by analyzing a keyword input for searching at a user terminal used by the user or a website.

  Furthermore, the advertisement distribution method can distribute appropriate advertisement data to the user terminal based on the acquired keyword and a predetermined condition set in advance.

WO2012132947 Publication WO20121275757 JP 2011-227720 A JP 2010-128628 A JP 2008-102174 A

  However, in Patent Document 1 to Patent Document 5 described above, a recommendation is made by changing an attribute value included in a pre-defined recommendation condition based on, for example, a purchase rate or similarity calculated by a predetermined process. It is only described that the conditions are optimized, and the recommendation effect may not be sufficiently enhanced.

  For example, when information indicating a position (latitude / longitude) is defined in advance as the attribute value, Patent Documents 1 to 5 change the values of latitude and longitude (that is, attribute values). Optimize the recommendation conditions.

  However, as a recommendation effect, there is a case where a sufficient effect cannot be obtained only by changing the attribute value.

  More specifically, as an example, when recommending a predetermined restaurant to a user, the recommendation effect may increase when recommending on the condition of “distance between the user and the restaurant” If you recommend that you have arrived at the nearest station in the vicinity, it may be larger.

  However, for example, when the recommended condition is “getting off at the nearest station located around the restaurant”, the method of changing the attribute value included in the recommended condition disclosed in Patent Document 1 to Patent Document 5, Such recommendation conditions cannot be generated.

  For this reason, the techniques disclosed in Patent Documents 1 to 5 cannot sufficiently enhance the recommendation effect.

  As means for solving such a problem, not only the value of the attribute included in the pre-defined recommendation condition is changed, but for example, information indicating the location of the user (latitude / longitude) and information indicating the location of the restaurant Based on (latitude / longitude), a method of generating a recommendation condition of calculating a distance between the user and the restaurant and recommending according to the calculated distance is conceivable.

  However, with such means, it is not possible to specify a combination of an attribute that has a particularly high recommendation effect and a combination of processing that, for example, calculates a distance, for an attribute that exists indefinitely and the attribute. It is very difficult for the parties.

  The main object of the present invention is to efficiently select a recommendation condition that can obtain a recommendation effect from a plurality of recommendation conditions when optimizing the recommendation condition, and to optimize the recommendation condition based on the selected recommendation condition. It is an object of the present invention to provide a recommendation condition optimizing device and the like that can optimize the recommendation condition.

  In order to achieve the above object, a recommendation condition optimization apparatus according to the present invention is characterized by comprising the following configuration.

That is, the recommended condition optimization device according to the present invention is:
The similarity between the recommended content associated with the first recommended content ID that can identify the first recommended condition and the recommended content associated with the second recommended content ID different from the first recommended content ID is calculated. A similarity calculation unit,
A specific second recommended content ID similar to the recommended content associated with the first recommended content ID is determined from the calculated similarities, and a recommended condition including one or more recommended conditions and a recommended content ID A first recommendation condition associated with the first recommended content ID and a second recommendation condition associated with the determined second recommended content ID are determined from the information, and the determined first Based on the first and second recommendation conditions, a conjunctive clause set is generated, the success rate, the number of successes, and the effect for each conjunctive clause belonging to the conjunctive clause set are calculated by a predetermined operation. A recommendation condition changing unit for generating a third recommendation condition based on the largest conjunctive clause and the first recommendation condition;
It is characterized by providing.

  Alternatively, this object can also be achieved by a recommendation system including the recommendation condition optimization device described above.

  In order to achieve the object, the recommended condition optimization method according to the present invention is characterized by having the following configuration.

That is, the recommended condition optimization method according to the present invention is:
The similarity between the recommended content associated with the first recommended content ID that can identify the first recommended condition and the recommended content associated with the second recommended content ID different from the first recommended content ID is calculated. And
A specific second recommended content ID similar to the recommended content associated with the first recommended content ID is determined from the calculated similarities, and a recommended condition including one or more recommended conditions and a recommended content ID A first recommendation condition associated with the first recommended content ID and a second recommendation condition associated with the determined second recommended content ID are determined from the information, and the determined first Based on the first and second recommendation conditions, a conjunctive clause set is generated, the success rate, the number of successes, and the effect for each conjunctive clause belonging to the conjunctive clause set are calculated by a predetermined operation. Generating a third recommendation condition based on the largest conjunctive clause and the first recommendation condition;
It is characterized by that.

  The same object is achieved by a computer program for realizing the recommended condition optimizing apparatus and the recommended condition optimizing method having the above-described configurations by a computer, and a readable storage medium storing the computer program. Is also achieved.

  According to the present invention, when optimizing a recommendation condition, a recommendation condition that particularly provides a recommendation effect is efficiently selected from a plurality of recommendation conditions, and an optimization target is selected based on the selected recommendation condition. It is possible to provide a recommendation condition optimization device or the like that can optimize a certain recommendation condition.

It is a block diagram which shows the structure of the recommendation condition optimization apparatus in the 1st Embodiment of this invention. It is a figure which illustrates concretely the composition of recommendation content attribute information in a 1st embodiment of the present invention. It is a flowchart which shows the operation | movement which the 1st similarity calculation part of the recommendation condition optimization apparatus in the 1st Embodiment of this invention performs. It is a figure which illustrates concretely the composition of similarity information in a 1st embodiment of the present invention. It is a flowchart which shows the operation | movement which the 1st recommendation condition change part of the recommendation condition optimization apparatus in the 1st Embodiment of this invention performs. It is a figure which illustrates concretely the composition of recommendation condition information in a 1st embodiment of the present invention. It is a figure which illustrates concretely the composition of recommendation effect information in a 1st embodiment of the present invention. It is a figure which illustrates concretely the information showing the success rate, the success number, and effect which the 1st recommendation condition change part in the 1st Embodiment of this invention calculates. It is a block diagram which shows the structure of the recommendation condition optimization apparatus in the 2nd Embodiment of this invention. It is a flowchart which shows the operation | movement which the 2nd recommendation condition change part of the recommendation condition optimization apparatus in the 2nd Embodiment of this invention performs. It is a figure which illustrates concretely the composition of recommendation condition information in a 2nd embodiment of the present invention. It is a block diagram which shows the structure of the recommendation condition optimization apparatus in the 3rd Embodiment of this invention. It is a flowchart which shows the operation | movement which the 2nd similarity calculation part of the recommendation condition optimization apparatus in the 3rd Embodiment of this invention performs. It is the figure which illustrated concretely the composition of recommendation effect information in a 3rd embodiment of the present invention. It is a figure which illustrates concretely the composition of the information containing the 1st product set (common delivery destination user set) which the 2nd similarity calculation part in a 3rd embodiment of the present invention computes. It is a Venn diagram which illustrates notionally the relationship between the 1st user set in the 3rd Embodiment of this invention, and a 2nd user set. It is a figure which illustrates concretely the composition of similarity information in a 3rd embodiment of the present invention. It is a block diagram which shows the structure of the recommendation system containing the recommendation condition optimization apparatus in the 4th Embodiment of this invention. It is a figure which illustrates concretely the composition of history information in a 4th embodiment of the invention. It is a block diagram explaining illustratively the hardware constitutions of the information processor which can realize each embodiment concerning the present invention. It is a figure which illustrates concretely the composition of the information which shows the state in patent documents 1.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

<First Embodiment>
FIG. 1 is a block diagram showing a configuration of a recommended condition optimization apparatus 1 according to the first embodiment of the present invention.

  In FIG. 1, the recommendation condition optimization device 1 includes a first similarity calculation unit 2 and a first recommendation condition change unit 3.

  More specifically, the first similarity calculation unit 2 includes first ID information including a recommended content identifier (identifier: hereinafter referred to as “ID”) that can uniquely identify a recommendation condition to be optimized. In response to receiving 101, the process is started.

  That is, the first similarity calculation unit 2 starts processing with the reception of the first ID information 101 as a trigger.

  In the following description, the recommended content ID included in the first ID information 101 is referred to as “first recommended content ID” for convenience of description. Further, the recommendation condition associated with the first recommended content ID is referred to as “first recommendation condition”. Further, a recommended content ID different from the first recommended content ID is referred to as a second recommended content ID (hereinafter, the same applies to each embodiment).

  Next, the first similarity calculation unit 2 recommends content associated with the first recommended content ID included in the recommended content attribute information 102 stored in the storage unit 4 (hereinafter referred to as “recommended content”). ) And a recommended content associated with a recommended content ID (second recommended content ID) different from the first recommended content ID.

  More specifically, the first similarity calculation unit 2 includes at least one attribute different from the recommended content attribute associated with the first recommended content ID included in the recommended content attribute information 102 and the first recommended content ID. Based on the attribute of the recommended content associated with the second recommended content ID, a process of calculating the similarity (similarity calculation process) is executed.

  The process in which the first similarity calculation unit 2 calculates the similarity will be described later in this embodiment.

  FIG. 2 is a diagram specifically illustrating the configuration of the recommended content attribute information 102 in the first embodiment of the present invention.

  The recommended content attribute information 102 is information in which the recommended content ID and information representing various attributes of the recommended content are associated with each other using the recommended content ID as a key.

  More specifically, in FIG. 2, the recommended content attribute information 102 includes a recommended content ID, a recommended content type (recommended type) indicating the type of recommended content, and a recommended content provider name indicating a provider that provides the recommended content. Attribute information indicating a recommended content position indicating a position associated with the recommended content and a recommended content expiration date indicating the expiration date of the recommended content.

  For example, in the second row of FIG. 2, the recommended content type associated with the recommended content ID “001” is “restaurant coupon” indicating that the recommended content (service) provides a restaurant coupon. The recommended content provider name is “Restaurant A” representing the name of the restaurant.

  Further, the recommended content position is “{35.574294, 139.6653498}” representing the location information (latitude / longitude) of the restaurant. The recommended content expiration date is “2012/1/1 to 2012/12/31”.

  In the above-described embodiment, for convenience of explanation, as an example, the recommended content attribute information 102 includes the recommended content ID, the recommended content type, the recommended content provider name, the recommended content position, and the recommended content expiration date. The configuration including the above has been described as an example.

  However, the embodiment according to the present invention is not limited to such a configuration, and the recommended content attribute information 102 may adopt a configuration including a recommended content ID, a recommended content type, and a recommended content position.

  That is, the recommended content attribute information 102 may employ a configuration that does not include the recommended content provider name and the recommended content valid period.

  In this case, the recommended content provider name and the recommended content validity period may be arbitrarily selected, for example, by the system administrator as to whether or not to be included in the recommended content attribute information 102.

  FIG. 3 is a flowchart showing an operation performed by the first similarity calculation unit 2 of the recommendation condition optimization apparatus 1 according to the first embodiment of the present invention. The operation procedure of the first similarity calculation unit 2 will be described along the flowchart.

  Here, as an example, the operation when calculating the similarity based on the recommended content associated with the first recommended content ID “002” and the second recommended content ID “001” will be described in detail. To do.

  For convenience of explanation, the above-described configuration will be described as an example. However, the present invention described by taking this embodiment as an example is not limited to the above-described configuration (the same applies to the following embodiments).

Step S001:
The first similarity calculation unit 2 starts processing in response to receiving the first ID information 101.

  In the above-described embodiment, for the sake of convenience of explanation, as an example, the first similarity calculation unit 2 has been described as an example of a configuration in which processing is started using the first ID information 101 as a trigger. However, the embodiment according to the present invention is not limited to such a configuration.

  For example, the first similarity calculation unit 2 may start processing according to a trigger given through a management application prepared by a system administrator.

  Further, the first similarity calculation unit 2 may adopt a configuration in which processing is started in accordance with a trigger given by a prepared program that gives a trigger at every predetermined time period (hereinafter, each embodiment) The same applies to the above).

Step S002:
Based on the first recommended content ID included in the first ID information 101 and the second recommended content ID included in the recommended content attribute information 102, the first similarity calculation unit 2 uses the similarity information ( (Similarity table).

  In the following description, for convenience of description, “first recommended content ID included in first ID information 101” is simply referred to as “first recommended content ID” (hereinafter, the same applies to each embodiment). ).

  FIG. 4 is a diagram specifically illustrating the configuration of similarity information in the first embodiment of the present invention.

  More specifically, in FIG. 4, the similarity information represents a first recommended content ID in the first column. The second column represents the second recommended content ID included in the recommended content attribute information 102. The third column represents the similarity of the recommended content associated with the calculated first recommended content ID and the second recommended content ID by executing the similarity calculation process.

  In order to calculate the similarity, the first similarity calculation unit 2 repeats the processing of each step from step S002 to step S008 for the number of records included in the recommended content attribute information 102.

  In the present embodiment described above, for convenience of explanation, as an example, the first similarity calculation unit 2 uses the first recommended content ID and the second recommended content ID included in the recommended content attribute information 102 as an example. Based on this, the configuration for generating similarity information has been described as an example.

  However, the embodiment according to the present invention is not limited to such a configuration, and the first similarity calculation unit 2 uses the first recommended content ID and the second recommended content ID included in the recommendation condition information 103. Based on this, a configuration for generating similarity information may be employed (hereinafter, the same applies to each embodiment).

Step S003:
The first similarity calculation unit 2 obtains the second recommended content ID from the similarity information and initializes the similarity associated with the obtained second recommended content ID (that is, “0”). Set).

  Here, the first similarity calculation unit 2 calculates the second recommended content ID “001” from the similarity information.

Step S004:
The first similarity calculation unit 2 refers to the recommended content attribute information 102 based on the first recommended content ID and the second recommended content ID.

  Further, the first similarity calculation unit 2 selects a specific recommended content type (specific recommended content type) associated with the first recommended content ID and the second recommended content ID from the recommended content attribute information 102. ) And a recommended content position (specific recommended content position).

  Here, the first similarity calculation unit 2 selects the recommended content type “restaurant coupon” associated with the first recommended content ID “002” from the recommended content attribute information 102 and the recommended content position “{35. 574188, 139.6654959} ”.

  Further, the first similarity calculation unit 2 obtains the recommended content type “restaurant coupon” associated with the second recommended content ID “001” and the recommended content position “{35.574294, 139.6653498}”. We will ask for it.

  The first similarity calculation unit 2 determines whether or not the recommended content types associated with the obtained two recommended content IDs match (same).

“YES” in step S005:
The first similarity calculation unit 2 determines whether or not the recommended content types match as a result of determining whether or not the recommended content types associated with the two recommended content IDs match (identical). The process proceeds to S006.

“NO” in step S005:
The first similarity calculation unit 2 determines whether or not the recommended content types associated with the two recommended content IDs match (identical), and determines that the recommended content types do not match. Proceed to step S008.

Step S006:
The first similarity calculation unit 2 uses the recommended content position (position information) associated with the first recommended content ID obtained in step S004 and the recommended content position (position information) associated with the second recommended content ID. ) To calculate the distance between the positions represented by the two recommended content positions (position information).

  In the following description, for convenience of description, “distance between positions represented by two recommended content positions” is simply referred to as “distance” (the same applies to the following embodiments).

  Here, the first similarity calculation unit 2 determines the distance “0... Based on the recommended content position associated with the first recommended content ID and the recommended content position associated with the second recommended content ID. 133 km "is calculated.

Step S007:
The first similarity calculator 2 adds the reciprocal of the calculated distance to the similarity initialized in step S003, and holds the similarity obtained by adding the reciprocal of the distance in the similarity information.

  Here, the first similarity calculation unit 2 adds the reciprocal “7.52” of the calculated distance “0.133 km” to the initialized similarity. As a result, the similarity shown in the second row in FIG. 4 is “7.52”.

  On the other hand, the similarity shown in the fourth row in FIG. 4 is “0”. The reason is that the first similarity calculation unit 2 determines that the recommended content types associated with the first recommended content ID “002” and the second recommended content ID “004” are not the same. .

  More specifically, the first similarity calculation unit 2 indicates that the recommended content type associated with the first recommended content ID “002” represents “restaurant coupon”, whereas the second recommended content ID “ Since the recommended content type associated with “004” represents “convenience store coupon”, it is determined that the two recommended content types are not the same.

Step S008:
The first similarity calculation unit 2 repeats the processes in steps S002 to S008 for the number of records included in the recommended content attribute information 102.

  Next, the first recommendation condition changing unit 3 receives the result of calculating the similarity from the first similarity calculating unit 2 and starts processing.

  The first recommendation condition changing unit 3 selects a specific second recommendation similar to the first recommendation condition associated with the first recommended content ID from the similarities calculated by the first similarity calculating unit 2. Find the content ID.

  Next, the first recommendation condition changing unit 3 is based on the first recommendation condition associated with the first recommended content ID and the recommendation condition associated with the obtained specific second recommended content ID. Then, a process of generating a third recommendation condition (recommendation condition generation process) is executed.

  In the following description, for convenience of explanation, the recommendation condition associated with the second recommended content ID having the highest similarity is referred to as “second recommendation condition” (hereinafter, the same applies to each embodiment).

  FIG. 5 is a flowchart showing an operation performed by the first recommended condition changing unit 3 of the recommended condition optimizing apparatus 1 according to the first embodiment of the present invention. The operation procedure of the first recommendation condition changing unit 3 will be described along the flowchart.

  Here, as an example, the operation in the case of generating the third recommended condition based on the recommended condition associated with the first recommended content ID “002” and the second recommended content ID “001” will be described. This will be described in detail.

  For convenience of explanation, the above-described configuration will be described as an example. However, the present invention described by taking this embodiment as an example is not limited to the above-described configuration (the same applies to the following embodiments).

Step S101:
The first recommended condition changing unit 3 refers to the recommended condition information 103 based on the first recommended content ID, and also recommends the recommended condition (the first recommended condition associated with the first recommended content ID from the recommended condition information 103). 1 recommendation condition).

  FIG. 6 is a diagram specifically illustrating the configuration of the recommendation condition information 103 in the first embodiment of the present invention.

  The recommendation condition information 103 is information in which a recommended content ID and a recommendation condition are associated with each other. In FIG. 6, the recommendation condition information 103 represents the recommended content ID in the first column. The second column represents recommendation conditions associated with the recommended content ID.

  More specifically, as an example, in the third line in FIG. 6, the recommendation condition associated with the recommended content ID “002” is “Distance (event.position, recommended content attribute.position) ≦ 200 m”. That is, the recommended condition represents that the recommended condition is that the distance between the position where the user is located and the recommended content position is 200 m or less, for example.

  Here, the event is, for example, various data (hereinafter also referred to as “status information”) representing the situation of the user or the like. The event is acquired from, for example, a mobile terminal device held by the user.

  Also, as an example, it is assumed that the recommendation condition information 103 is registered (held) in advance by a system administrator (recommendation condition registrant).

Step S102:
The first recommendation condition changing unit 3 selects a specific recommended content ID (second recommended content ID) associated with the highest similarity from the similarity information generated by the first similarity calculation unit 2. Ask.

  Further, the first recommended condition changing unit 3 refers to the recommended condition information 103 based on the obtained specific second recommended content ID, and also specifies a specific second recommended content ID from the recommended condition information 103. The recommendation condition (second recommendation condition) associated with is obtained.

  Here, the first recommendation condition changing unit 3 obtains the second recommended content ID “001” and the second recommendation content ID “001” associated with the second recommended content ID “001” from the recommendation condition information 103. We will ask for conditions.

Step S103:
The first recommendation condition changing unit 3 converts the obtained first recommendation condition and the second recommendation condition into an additive standard form.

  Note that since the technology itself for converting the recommended condition into the additive standard form can be a general technology at present, a detailed description in this embodiment will be omitted (hereinafter, the same applies to each embodiment).

Step S104:
The first recommendation condition changing unit 3 obtains a conjunctive clause included in the first recommendation condition and the second recommendation condition after conversion into the additive standard form.

  For example, when A, B, C, D, E, and F are atomic logical expressions (logical expressions having no partial logical expression), “(A∧B) ∨ (C∧D∧E) ∨F” is , Additive normal form (a kind of standardization (normalization) of logical expressions in Boolean logic, which expresses logical expressions by disjunctive disjunction).

  In the above-mentioned “(A「 B) ∨ (C∧D∧E) ∨F ”, the conjunctive clauses are“ (A∧B) ”,“ (C∧D∧E) ”, and“ F ”. .

  Here, “∧” is also expressed as “AND”, “joint”, and “logical product”. “∨” is also expressed as “OR”, “disjunction”, and “logical sum”.

Step S105:
The first recommendation condition changing unit 3 includes a conjunction clause included in the second recommendation condition after conversion to the additive standard form and a conjunction clause included in the first recommendation condition after conversion to the additive standard form. Based on this, a set of consecutive clauses (joint clause set) is generated.

  In the following description, for convenience of explanation, the consecutive phrase set is referred to as a “second consecutive phrase set” (the same applies to the following embodiments).

  More specifically, the first recommendation condition changing unit 3 includes one or more of the conjunction clauses included in the second recommendation condition except for the conjunction clauses of the same type as the conjunction clause included in the first recommendation condition. Based on the collocation clause, a second collocation clause set is generated.

  Here, the same kind is a condition that can be the same condition by adjusting a parameter (attribute value).

  For example, “Distance (event.position, recommended content attribute.position) ≦ 300 m” as a conjunctive clause is compared with “Distance (event.position, recommended content attribute.position) ≦ 200 m” as a conjunctive clause. In some cases, the two conjunctive clauses both have the same condition by adjusting the parameter distance (ie, 200 m to 300 m).

  Further, the second conjunctive clause set generated based on the recommendation condition associated with the first recommended content ID “002” and the second recommended content ID “001” is {“Station (event.position)”. = “Musashi Kosugi” ”,“ Station (event.position) = “Motosumiyoshi” ”}.

  Here, Musashi Kosugi and Motosumiyoshi are place names in Japan.

Step S106:
Based on the recommended effect information 104 held in the storage unit 4, the first recommendation condition changing unit 3 includes each collocation clause (for each collusion clause) included in (belongs to) the generated second collocation clause set. ) Is calculated by a predetermined calculation.

  Here, the recommended number is the number of users that meet the conditions included in the conjunctive clause. The number of successes is the number of users who take an expected action with respect to the recommendation from among users who meet the conditions.

  The success rate is calculated by dividing the number of successes by the number of recommendations. The effect is calculated by multiplying the calculated success rate by the number of successes. However, the present invention described using this embodiment as an example is not limited to the calculation method described above (hereinafter, the same applies to each embodiment).

  FIG. 7 is a diagram specifically illustrating the configuration of the recommendation effect information 104 according to the first embodiment of this invention.

  In FIG. 7, the recommended effect information 104 includes a user ID that can identify a user to whom a recommended content associated with the recommended content ID has been distributed, position information (latitude, longitude) that indicates a location when the recommended content is received. , The time when the recommended content was received, the recommended effect indicating whether or not the user performed an expected action on the recommended content, and the state of the user when the recommended content was received Information including a success factor indicating whether the partial condition is met.

  That is, the recommended effect information 104 is information in which the recommended content ID, the user ID, the position information, the time, the recommended effect, and the success factor are associated using the recommended content ID as a key.

  Here, as a recommendation effect, when the recommended content represents a coupon for a restaurant, “success” indicates that the coupon is used in the restaurant.

  In this case, for example, the recommendation effect may be determined such that the effect is determined when the user presents the coupon when accounting at a restaurant. However, the present invention described using this embodiment as an example is not limited to the configuration described above (hereinafter, the same applies to each embodiment).

  More specifically, in FIG. 7, when the conjunctive clause represents “Station (event.position) =“ Musashi Kosugi ””, the first recommendation condition changing unit 3 selects the position information from the recommended effect information 104. Based on the above, the number of users (recommended number) matching the conjunctive clause “Station (event.position) =“ Musashi Kosugi ”” is obtained. Here, the number of recommendations is ten.

  Further, the first recommendation condition changing unit 3 obtains the number of users (success counts) that match the related clause from the recommended effect information 104 and whose recommendation effect indicates “success”. Here, the number of successes is eight.

  As a result, the success rate is “0.8” calculated by the success number “8” / recommendation number “10”. The effect is “6.4” calculated by the success rate “0.8” × the number of successes “8”.

  FIG. 8 is a diagram specifically illustrating information indicating the success rate, the number of successes, and the effect calculated by the first recommendation condition changing unit 3 according to the first embodiment of the present invention.

  More specifically, FIG. 8 shows the conjunctive clause in the first column, the success rate in the second column, the number of successes in the third column, and the effect in the fourth column. FIG. 8 shows information in which a conjunctive clause, a success rate, the number of successes, and an effect are associated using the conjunctive clause as a key.

Step S107:
The first recommendation condition changing unit 3 obtains the combined clause having the greatest effect from the calculated effects, and sets the third recommended condition based on the first recommended condition and the obtained combined clause. Generate.

  More specifically, as an example, the first recommendation condition changing unit 3 generates a third recommendation condition by connecting the conjunction clauses obtained for the first recommendation condition by disjunction (OR).

  Here, in FIG. 8, the conjunctive clause associated with the highest effect is “Station (event.position) =“ Musashi Kosugi ””. The first recommendation condition is “Distance (event.position, recommended content attribute.position) ≦ 200 m”.

  The first recommendation condition changing unit 3 determines the third recommendation condition “Distance (event.position, recommended content attribute.position) ≦ 200 m OR Station (event) based on the connected clause and the first recommendation condition. . Position) = “Musashi Kosugi” is generated.

  The first recommendation condition changing unit 3 stores the generated third recommendation condition in the recommendation condition information 103 in association with the first recommended content ID.

  For example, the first recommendation condition changing unit 3 may adopt a configuration in which the first recommendation condition included in the recommendation condition information 103 is retained by overwriting the third recommendation condition, or a system administrator ( A configuration may be employed in which the third recommendation condition approved by the recommendation condition registrant) is held in the recommendation condition information 103 (the same applies to the following embodiments).

  In this case, for example, the first recommendation condition changing unit 3 may adopt a configuration in which the third recommendation condition is presented to the recommended condition registrant using a display unit (not shown).

  Next, the storage unit 4 holds recommended content attribute information 102, recommendation condition information 103, and recommendation effect information 104.

  Here, the storage unit 4 is a nonvolatile storage device that can read and write data by a computer. More specifically, as an example, the storage unit 4 may employ a nonvolatile storage device such as a hard disk drive (Hard_disk_drive: hereinafter referred to as “HDD”) mounted in an electronic device such as a server device.

  As an example, the storage unit 4 may employ a storage device (not shown) connected to a communication network (not shown). However, the present invention described using this embodiment as an example is not limited to the above-described configuration (the same applies to the following embodiments).

  As described above, according to the recommendation condition optimizing device 1 according to the present embodiment, when optimizing the recommendation condition, the recommendation condition that particularly provides the recommendation effect is efficiently selected from the plurality of recommendation conditions, Based on the selected recommendation condition, the recommendation condition to be optimized can be optimized. The reason is as described below.

  That is, the first similarity calculation unit 2 of the recommended condition optimization device 1 includes the recommended content associated with the first recommended content ID that can identify the first recommended condition to be optimized, and the first recommended content ID. The similarity between the recommended contents is calculated by executing the similarity calculation process based on the recommended content ID different from the recommended content ID associated with the second recommended content ID.

  Further, the first recommendation condition changing unit 3 of the recommendation condition optimizing device 1 obtains the second recommended content ID having the highest similarity calculated by the first similarity calculation unit 2. The first recommendation condition changing unit 3 removes the second recommendation condition associated with the obtained second recommended content ID from the second recommendation condition and the same kind of conjunction clause of the first recommendation condition. Generate a set of conjunction clauses. The first recommendation condition changing unit 3 obtains the most effective joint clause from the second joint clause set. Next, the first recommendation condition changing unit 3 can generate a third recommendation condition based on the first recommendation condition and the obtained joint clause.

  In other words, the recommended condition optimizing apparatus 1 not only changes the attribute value included in the first recommended condition to be optimized, but also the second recommended condition where the recommended content is similar and the recommendation effect is the largest. Select the conjunctive clauses included in. Furthermore, the recommended condition optimization apparatus 1 can optimize the first recommended condition based on the first recommended condition and the selected conjunction clause.

  Therefore, the recommendation condition optimizing apparatus 1 can generate not only a recommendation condition having the largest recommendation effect, but also a recommendation condition having a large recommendation effect that cannot be predicted by the system administrator, for example.

  More specifically, the recommended condition optimizing apparatus 1 not only increases the distance as a parameter (attribute value) as a relaxation means of the recommended condition “Distance (event.position, recommended content attribute.position) ≦ 200 m”. The third recommendation condition is generated by connecting “Station (event.position) =“ Musashi Kosugi ”” to the first recommendation condition by OR.

  This is because the number of successes increases when only the distance that is a parameter is optimized (for example, the distance is increased). On the other hand, the success rate is low. Therefore, the recommendation condition that optimizes the attribute value may not contribute to the recommendation effect.

  The recommended condition optimizing apparatus 1 according to the first embodiment of the present invention optimizes not only the distance as a parameter, but also, for example, a condition targeting a person who is “at the nearest station from the restaurant location”. The recommendation condition can be optimized by providing more options for increasing the recommendation effect calculated from the number of successes and the success rate.

<Second Embodiment>
Next, a second embodiment based on the recommended condition optimization apparatus 1 according to the first embodiment of the present invention described above will be described. In the following description, the characteristic part according to the present embodiment will be mainly described. At this time, the same reference numerals are assigned to the same configurations as those in the above-described embodiments, and duplicate descriptions are omitted.

  A recommended condition optimizing apparatus 10 according to the second embodiment of the present invention will be described with reference to FIG.

  FIG. 9 is a block diagram showing a configuration of the recommended condition optimization device 10 according to the second exemplary embodiment of the present invention.

  In FIG. 9, the recommended condition optimization device 10 includes a first similarity calculation unit 2 and a second recommended condition change unit 11.

  The recommended condition optimizing device 10 is different from the operation of the first recommended condition changing unit 3 described in the first embodiment in the operation of the second recommended condition changing unit 11.

  In the following description, the operation of the second recommendation condition changing unit 11 in the present embodiment will be described more specifically.

  FIG. 10 is a flowchart showing an operation performed by the second recommended condition changing unit 11 of the recommended condition optimizing apparatus 10 according to the second embodiment of the present invention. The operation procedure of the second recommendation condition changing unit 11 will be described along the flowchart.

  Here, as an example, the operation in the case of generating the third recommended condition based on the recommended condition associated with the first recommended content ID “002” and the second recommended content ID “001” will be described. This will be described in detail.

  For convenience of explanation, the above-described configuration will be described as an example. However, the present invention described by taking this embodiment as an example is not limited to the above-described configuration (the same applies to the following embodiments).

  In FIG. 10, the operations shown in steps S201 to S204 performed by the second recommended condition changing unit 11 are the same as those performed by the first recommended condition changing unit 3 described with reference to FIG. 5 in the first embodiment. Thru | or the operation | movement similar to step S104. Therefore, the overlapping description is omitted.

Step S205:
The second recommendation condition changing unit 11 changes the conjunction clause (first conjunction clause) to be changed from the conjunction clauses included in the first recommendation condition after being converted into the additive standard form in step S204. Is selected.

  Here, as a method for selecting the first conjunction clause, the success rate, the number of successes and the effect for each conjunction clause included in the first recommendation condition are calculated, and the conjunction clause having the smallest calculated effect is used. A configuration may be adopted in which is selected as the first consecutive clause. However, the present invention described using this embodiment as an example is not limited to the configuration described above (hereinafter, the same applies to each embodiment).

  In addition, the process (step S205, step S207) which calculates a success rate, the number of successes, and an effect for every conjunctive clause which the 2nd recommendation condition change part 11 performs is the 1st recommendation condition demonstrated in 1st Embodiment. This is the same as the processing for calculating the success rate, the number of successes, and the effect for each consecutive clause performed by the changing unit 3. Therefore, the overlapping description is omitted.

  Here, in FIG. 11, the second recommendation condition changing unit 11 selects the first consecutive clause “Distance (event.position, recommended content attribute.position) ≦ 200 m”.

  FIG. 11 is a diagram specifically illustrating the configuration of the recommendation condition information 105 in the second embodiment of the present invention.

  In FIG. 11, recommended condition information 105 represents a recommended content ID in the first column. The second column represents recommendation conditions associated with the recommended content ID.

Step S206:
The second recommendation condition changing unit 11 is the same as the first conjunction clause among the conjunction clauses included in the second recommendation condition converted into the additive standard form in step S204, and the first recommendation clause change section 11 Find a conjunction clause that contains one or more atomic logical expressions that are not the atomic logical expressions that make up the conjunction clause, and generate a set of conjunction clauses (second conjunction clause set) based on the obtained conjunction clause To do.

  Here, for example, when A, B, C, and D are atomic logical expressions, the first conjunction clause is represented by “A∧B∧C”. The second recommendation condition is represented by “(A∧B) ∨ (B∧D)”.

  In this case, the atomic logical expressions A and B constituting the conjunctive clause “(A∧B)” included in the second recommendation condition are the ones constituting the first conjunctive clause “A∧B∧C”. This is the atomic formula. For this reason, the conjunctive clause “(A∧B)” included in the second recommendation condition is not included in the second conjunctive clause set.

  On the other hand, among the atomic logical expressions B and D constituting the conjunction clause “(B∧D)” included in the second recommendation condition, D constitutes the first conjunction clause “A∧B∧C”. Is not any atomic formula. Therefore, the conjunctive clause “(B∧D)” included in the second recommendation condition is included in the second conjunctive clause set.

  Here, “Distance (event.position, recommended content attribute.position) ≦ 300 m” and “20 ≦ event.age <30”, which are atomic logical expressions constituting the conjunctive clause included in the second recommendation condition. Is any atomic logical expression constituting the first conjunction clause.

  More specifically, “Distance (event.position, recommended content attribute.position) ≦ 300 m”, which is an atomic logical expression constituting the conjunction clause included in the second recommendation condition, constitutes the first conjunction clause. It is the same kind as “Distance (event.position, recommended content attribute.position) ≦ 200 m”, which is an atomic logical formula.

  Furthermore, “20 ≦ event.age <30”, which is an atomic logical expression that constitutes a conjunctive clause included in the second recommendation condition, represents an age range in the atomic logical expression that constitutes the first conjunctive clause, It becomes the same kind by changing from the minimum value that can be taken as the age to the maximum value.

  As a result, the second consecutive clause set is the consecutive clause {“Station (event.position) =“ Musashi Kosugi ”AND 20 ≦ event.age <30”} included in the second recommendation condition.

Step S207:
Based on the recommended effect information 104 held in the storage unit 4 and the respective joint clauses included in (belonging to) the second joint clause set generated, the second recommendation condition changing unit 11 The success rate, the number of successes, and the effect of the conjunction clause (for each conjunction clause) are calculated by a predetermined calculation.

  Here, the conjunctive clause having the greatest effect is “20 ≦ event.age <30”.

Step S208:
The second recommendation condition changing unit 11 generates a third conjunction clause based on the conjunction clause having the largest effect among the calculated effects and the first conjunction clause.

  More specifically, the second recommendation condition changing unit 11 connects the conjunction clause having the greatest effect from the effects calculated in the first conjunction clause by connecting the conjunction clause (AND). Generates a collocation clause.

Step S209:
The second recommendation condition changing unit 11 generates the third recommendation condition by replacing the first consecutive phrase with the generated third consecutive phrase.

  The second recommendation condition changing unit 11 stores the generated third recommendation condition in the recommendation condition information 105 in association with the first recommended content ID.

  Here, the second recommendation condition changing unit 11 replaces the first consecutive phrase with the generated third consecutive phrase, thereby generating the third recommendation condition “Distance (event.position, recommended content attribute. (Position) ≦ 200 m AND 20 ≦ event.age <30 ”.

  As described above, according to the recommended condition optimizing apparatus 10 according to the present embodiment, it is possible to enjoy the effects described in the first embodiment, and even when the recommended conditions are strict (regulated), a plurality of recommended conditions are obtained. Among the conditions, it is possible to efficiently select a recommendation condition that provides a recommendation effect, and to optimize a recommendation condition to be optimized based on the selected recommendation condition.

  The reason is that the second recommendation condition changing unit 11 of the recommendation condition optimizing device 10 selects the first conjunction clause to be changed from the conjunction clauses included in the first recommendation condition. Further, the second recommendation condition changing unit 11 is the same kind as the first conjunction clause among the conjunction clauses included in the second recommendation condition, and the atomic logic constituting the first conjunction clause. A conjunctive clause including one or more atomic logical expressions that are not expressions is obtained, and a second conjunctive clause set is generated based on the obtained one or more conjunctive clauses. The second recommendation condition changing unit 11 calculates the success rate, the number of successes, and the effect for each consecutive phrase included in the second consecutive phrase set. The second recommendation condition changing unit 11 generates a third conjunction clause based on the conjunction clause having the largest effect among the calculated effects and the first conjunction clause. This is because the second recommendation condition changing unit 11 can generate the third recommendation condition by replacing the first consecutive phrase with the generated third consecutive phrase.

  More specifically, the recommended condition optimizing apparatus 10 not only reduces the distance as a parameter (attribute value) as a restricting means of the recommended condition “Distance (event.position, recommended content attribute.position) ≦ 200 m”. The third recommendation condition is generated by connecting “20 ≦ event.age <30” to the first recommendation condition by AND.

  This is because the success rate is high when only the parameter distance is optimized (for example, the distance is small). On the other hand, the number of successes decreases. Therefore, the recommendation condition that optimizes the attribute value may not contribute to the recommendation effect.

  The recommendation condition optimization device 10 according to the second exemplary embodiment of the present invention not only optimizes the distance as a parameter, but also recommends the recommendation calculated from the number of successes and the success rate, for example, a condition related to age is added. The recommendation conditions can be optimized by providing more options for increasing the effect.

<Third Embodiment>
Next, a third embodiment based on the recommended condition optimization apparatus 1 according to the first embodiment of the present invention described above will be described. In the following description, the characteristic part according to the present embodiment will be mainly described. At this time, the same reference numerals are assigned to the same configurations as those in the above-described embodiments, and duplicate descriptions are omitted.

  A recommended condition optimizing device 20 according to a third embodiment of the present invention will be described with reference to FIG.

  FIG. 12 is a block diagram showing a configuration of the recommended condition optimization device 20 in the third exemplary embodiment of the present invention.

  In FIG. 12, the recommendation condition optimization device 20 includes a second similarity calculation unit 21 and a first recommendation condition change unit 3.

  In the recommended condition optimizing device 20, the operation in the second similarity calculation unit 21 is different from the operation of the first similarity calculation unit 2 described in the first embodiment.

  The recommended condition optimization device 20 calculates the similarity based on the reaction of the user who has received the recommended content.

  In the following description, the operation of the second similarity calculation unit 21 in the present embodiment will be described more specifically.

  FIG. 13 is a flowchart illustrating an operation performed by the second similarity calculation unit 21 of the recommendation condition optimization device 20 according to the third exemplary embodiment of the present invention. The operation procedure of the second similarity calculation unit 21 will be described along the flowchart.

  Here, as an example, the similarity is calculated based on the recommendation result of the user who has distributed the recommended content associated with the first recommended content ID “C001” and the second recommended content ID “C002”. The operation in this case will be described in detail.

  For convenience of explanation, the above-described configuration will be described as an example. However, the present invention described using this embodiment as an example is not limited to the above-described configuration (the same applies to the following embodiments).

  In FIG. 13, the operations shown in steps S301 and S308 performed by the second similarity calculation unit 21 are the same as steps S001 and S308 performed by the first similarity calculation unit 2 described in FIG. 3 in the first embodiment. The operation is the same as that shown in step S008. Therefore, the overlapping description is omitted.

Step S302:
Based on the first recommended content ID and one or more second recommended content IDs different from the first recommended content ID included in the recommendation condition information 103, the second similarity calculation unit 21 Information (similarity table) is generated.

  In order to calculate the similarity, the second similarity calculation unit 21 repeats the processing of each step from step S302 to step S308 by the number of records included in the recommendation condition information 103.

Step S303:
The second similarity calculation unit 21 obtains the second recommended content ID from the similarity information, and initializes the similarity associated with the obtained second recommended content ID.

Step S304:
The second similarity calculation unit 21 refers to the recommended effect information 106 held in the storage unit 4 based on the first recommended content ID and the second recommended content ID.

  The second similarity calculation unit 21 is configured to perform a first operation between a user to whom recommended content associated with the first recommended content ID is distributed and a user to which recommended content associated with the second recommended content ID is distributed. Product set (common delivery destination user set).

  More specifically, the second similarity calculation unit 21 refers to the recommended effect information 106, the user ID associated with the first recommended content ID from the referenced recommended effect information 106, and the second The user ID associated with the recommended content ID is obtained.

  Further, the second similarity calculation unit 21 determines the common distribution destination user set based on the user ID associated with the obtained first recommended content ID and the user ID associated with the second recommended content ID. Is calculated.

  That is, the second similarity calculation unit 21 calculates a similarity for a user to whom the recommended content associated with the first recommended content ID and the recommended content associated with the second recommended content ID are distributed. Extract as

  Thereby, the second similarity calculation unit 21 can collect responses to two recommended contents from the same user and compare the collected responses.

  FIG. 14 is a diagram specifically illustrating the configuration of the recommendation effect information 106 according to the third embodiment of the present invention.

  More specifically, in FIG. 14, the recommended effect information 106 is information including a recommended content ID, a user ID, position information, time, a recommended effect, and a success factor.

  That is, the recommended effect information 106 is information in which the recommended content ID, the user ID, the position information, the time, the recommended effect, and the success factor are associated with the recommended content ID as a key.

  Here, in FIG. 14, the user ID associated with the first recommended content ID “C001” is {56, 651, 185, 493, 13, 631, 698}.

  In addition, the user ID associated with the second recommended content ID “002” is {56, 651, 185, 493, 13, 631, 597, 517}.

  As a result, the first product set (common delivery destination user set) of user IDs is {56, 651, 185, 493, 13, 631}.

  FIG. 15 is a diagram specifically illustrating a configuration of information including the first product set (common distribution destination user set) calculated by the second similarity calculation unit 21 according to the third embodiment of the present invention. .

  More specifically, in FIG. 15, the first column represents the first recommended content ID. The second column represents the second recommended content ID. The third column represents the calculated first product set (common delivery destination user set).

Step S305:
The second similarity calculation unit 21 calculates a subset of the common distribution destination user set.

  More specifically, the second similarity calculation unit 21 refers to the recommended effect information 106 based on the common distribution destination user set, and is included in the common distribution destination user set from the referenced recommended effect information 106. A user ID whose recommendation effect associated with the user ID and the first recommended content ID indicates “success” is obtained.

  The second similarity calculation unit 21 calculates a first user set based on the obtained user ID.

  Furthermore, the second similarity calculation unit 21 determines that the recommendation effect associated with the user ID included in the common distribution destination user set and the second recommended content ID from the referred recommendation effect information 106 is “success”. And a second user set is calculated based on the obtained user ID.

  Here, the first user set is {56, 651, 493, 631}. The second user set is {56, 493, 13, 631}.

Step S306:
The second similarity calculation unit 21 calculates a union and a product set based on the calculated first user set and the second user set. Furthermore, the second similarity calculation unit 21 calculates the ratio of the product set to the union.

  More specifically, the second similarity calculation unit 21 calculates the ratio of the product set to the union by dividing the product set by the union.

  Here, the union is {56, 651, 493, 631, 13}. The product set is {56, 493, 631}.

  That is, the ratio of the product set “3” to the sum set “5” is “0.6” calculated by the product set “3” / the sum set “5”.

  FIG. 16 is a Venn diagram conceptually illustrating the relationship between the first user set and the second user set in the third embodiment of the present invention.

  In FIG. 16, the first user set is represented by a solid closed curve. The second set of users is represented by a dashed closed curve. The common distribution destination user set is represented by a rectangle.

  And the union calculated | required based on the 1st user set and the 2nd user set is represented by the closed curve of a continuous line and a broken line. The product set is represented by a region (a region indicated by diagonal lines) where a solid closed curve and a dashed closed curve overlap.

Step S307:
The second similarity calculation unit 21 adds the ratio of the product set to the calculated union to the similarity initialized in step S303, and holds the added similarity in the similarity information.

  That is, the degree of similarity between two recommended contents means that the more users who show the same reaction, the higher the value.

  FIG. 17 is a diagram specifically illustrating the configuration of similarity information according to the third embodiment of the present invention.

  In FIG. 17, the similarity information represents the first recommended content ID in the first column. The second column represents the second recommended content ID. The third column represents the similarity calculated in step S307. That is, the similarity information is information in which the first recommended content ID, the second recommended content ID, and the similarity are associated with each other.

  In the above-described embodiment, for convenience of explanation, as an example, the second similarity calculation unit 21 has been described using a configuration in which the similarity is calculated based on a set having user IDs as elements.

  However, the embodiment according to the present invention is not limited to such a configuration, and the second similarity calculation unit 21 defines and defines a user type that uniquely identifies an attribute representing a user's gender, age, and the like. A configuration may be employed in which the similarity is calculated based on a set having user types as elements.

  Moreover, the recommended condition optimization apparatus 20 according to the present embodiment described above can also be applied to the second embodiment.

  As described above, according to the recommended condition optimizing device 20 according to the present embodiment, the effects described in the embodiments can be enjoyed, and further, even if the attribute of the recommended content is not known, the similarity between the recommended contents The degree can be calculated.

  The reason is that the second similarity calculation unit 21 of the recommendation condition optimization device 20 can calculate the similarity between recommended contents based on the recommendation effect information 106.

<Fourth Embodiment>
Next, a fourth embodiment based on the recommended condition optimizing apparatus 1 according to the first embodiment of the present invention described above will be described. In the following description, the characteristic part according to the present embodiment will be mainly described. At this time, the same reference numerals are assigned to the same configurations as those in the above-described embodiments, and duplicate descriptions are omitted.

  A recommended condition optimizing device 30 according to the fourth embodiment of the present invention will be described with reference to FIG.

  FIG. 18 is a block diagram showing a configuration of a recommendation system 50 including a recommendation condition optimization device 30 according to the fourth embodiment of the present invention.

  In FIG. 18, the recommendation system 50 includes a recommendation condition optimization device 30, a storage unit 4, a recommended content distribution unit 51, and a recommendation necessity determination unit 52.

  More specifically, the recommendation condition optimizing device 30 includes a first similarity calculation unit 2, a first recommendation condition change unit 3, and a recommendation effect measurement unit 31.

  The recommended condition optimizing device 30 is further different from the recommended condition optimizing device 1 described in the first embodiment in that a recommended effect measuring unit 31 is provided.

  In the following description, the operation of the recommended effect measurement unit 31 in the present embodiment will be described more specifically.

  Based on the effect information (recommended effect information 106) obtained over a predetermined period, the recommended effect measuring unit 31 calculates the success rate, the number of successes, and the effect for the first recommended condition and the third recommended condition. It is calculated by a predetermined calculation.

  More specifically, as an example, when the predetermined period is set to one week, the recommended condition optimizing device 30 sets the first recommended condition and the third recommended condition to one week in the recommended condition. Provide cross-border service.

  That is, the recommended condition optimizing device 30 provides a service for a predetermined period of time in the first recommended condition and the third recommended condition so that the conditions are the same when the recommended effect measuring unit 31 calculates the effect. provide.

  The process of calculating the success rate, the number of successes, and the effect by the recommended effect measuring unit 31 is the process of calculating the success rate, the number of successes, and the effect performed by the first recommendation condition changing unit 3 described in the first embodiment. It is the same. Therefore, the overlapping description is omitted.

  Next, the recommended effect measuring unit 31 compares the effect with respect to the calculated first recommended condition and the third recommended condition, and the first recommended condition or the third according to the result of comparing the effects. One of the recommended conditions is selected, and the selected recommended condition is stored in the recommended condition information 103 in association with the first recommended content ID.

  More specifically, the recommended effect measuring unit 31 selects a recommended condition having a larger effect as a result of comparing the effects on the first recommended condition and the third recommended condition.

  Next, the recommendation necessity determination unit 52 searches the recommended condition information 103 based on the state information 107 in response to the state information 107 being notified from the user at an arbitrary timing, for example.

  Here, the state information 107 includes information representing a state such as a user ID that can identify the user, a user's whereabouts (position), time, and surrounding conditions.

  The recommendation necessity determination unit 52 searches the recommended condition information 103 based on the state information 107, and if there is a specific recommended condition that matches the state information 107, the recommended condition that matches the recommended condition information 103. In addition to obtaining the recommended content ID associated with, the recommended content ID obtained is provided to the recommended content distribution unit 51.

  The recommended content distribution unit 51 refers to the recommended content attribute information 102 based on the recommended content ID, and also recommends the recommended content based on various attribute information associated with the recommended content ID from the recommended content attribute information 102. Deliver against.

  In the present embodiment described above, for the convenience of explanation, as an example, the recommendation necessity determination unit 52 determines whether there is a recommendation condition that matches the state information 107 based on the state information 107 and the recommendation condition information 103. The configuration for determining whether or not is described as an example.

  However, the embodiment according to the present invention is not limited to such a configuration, and the recommendation necessity determination unit 52, for example, the status information 107, the recommendation condition information 103, and the history of actions taken by the user who is the distribution target. A configuration may be adopted in which it is determined whether or not there is a recommendation condition that matches the state information 107 based on information indicating the above (hereinafter referred to as “history information”).

  In that case, the storage unit 4 holds recommended content attribute information 102, recommendation condition information 103, recommended effect information 106, and history information 108.

  For example, when the status information 107 is notified from the user at an arbitrary timing, the recommendation necessity determination unit 52 searches the history information 108 based on the user ID included in the status information 107 and As a result of searching the information 108, the history information associated with the user ID is obtained from the history information 108.

  Further, the recommendation necessity determination unit 52 searches the recommended condition information 103 based on the history information obtained from the history information 108 and the state information 107, and also searches the recommended condition information 103. As a result, the state information When there is a recommendation condition that matches 107, the recommended content ID associated with the recommended recommendation condition is obtained from the recommendation condition information 103.

  More specifically, in the following description, for convenience of explanation, as an example, the history information 108 is information representing the purchase history of the user.

  FIG. 19 is a diagram specifically illustrating the configuration of the history information 108 in the fourth embodiment of the invention.

  In FIG. 19, the history information 108 represents the user ID in the first column. The second column represents the date and time of purchase. The third column represents the place of purchase. The fourth column represents purchased products (hereinafter referred to as “purchase targets”). The fifth column represents the purchased amount.

  That is, the history information 108 is information that associates a user ID, a date, a place, a product (purchasing target), and an amount with the user ID as a key.

  By holding the history information 108, the system administrator can, for example, “a user who is located within 300 m from the position associated with the recommended content and has passed one week or more after eating at a restaurant in the past”. On the other hand, it is possible to set a recommendation condition such as distribution of recommended content.

  From these, for example, status information 107 including the user ID “254”, the date and time “2012/12/3”, and position information indicating that the user is within 300 m from the position associated with the recommended content is notified. If it is determined, the recommendation necessity determination unit 52 determines that there is no recommendation condition that matches the state information 107.

  More specifically, in the third line shown in FIG. 18, the recommendation necessity determination unit 52 is included in the status information 107 and the date and time “2012/12/2 13:52” associated with the user ID “254”. Based on the date and time, it is determined that one week or more has not elapsed.

  That is, since the user associated with the user ID “254” has not eaten for more than one week since eating at the restaurant, the recommended content is not distributed.

  On the other hand, for example, status information 107 including user ID “254”, date and time “2012/12/20”, and position information indicating that the user is within 300 m from the position associated with the recommended content is notified. If it is determined, the recommendation necessity determination unit 52 determines that there is a recommendation condition that matches the state information 107.

  More specifically, in the third line shown in FIG. 18, the recommendation necessity determination unit 52 is included in the status information 107 and the date and time “2012/12/2 13:52” associated with the user ID “254”. It is determined that one week or more has passed based on the date and time.

  That is, since the user associated with the user ID “254” has been eating for one week or more after eating at the restaurant, the recommended content is distributed.

  The recommended condition optimization device 30 according to the present embodiment described above can also be applied to the first to fourth embodiments.

  As described above, according to the recommended condition optimizing device 30 according to the present embodiment, the effects described in the embodiments can be enjoyed, and the recommended conditions can be further optimized.

  The reason is that the recommended condition optimizing device 30 is based on the effect information obtained over a predetermined period of time, the effect on the first recommended condition that is the recommended condition before the recommended condition is optimized, and the recommended condition. This is because, as a result of comparing the effect on the third recommended condition, which is the recommended condition after optimizing the value, it is possible to select the recommended condition that is most effective.

  In addition, the recommendation condition optimizing apparatus 30 can provide more conditions (options) for increasing the recommendation effect calculated from the number of successes and the success rate by holding the history information 108.

In addition, a part or all of each embodiment mentioned above and its Example may be described also as the following additional remarks. However, the present invention described by way of example using the above-described embodiment and its examples is not limited to the following. That is,
(Appendix 1)
The similarity between the recommended content associated with the first recommended content ID that can identify the first recommended condition and the recommended content associated with the second recommended content ID different from the first recommended content ID is calculated. A similarity calculation unit,
A specific second recommended content ID similar to the recommended content associated with the first recommended content ID is determined from the calculated similarities, and a recommended condition including one or more recommended conditions and a recommended content ID A first recommendation condition associated with the first recommended content ID and a second recommendation condition associated with the determined second recommended content ID are determined from the information, and the determined first Based on the first and second recommendation conditions, a conjunctive clause set is generated, the success rate, the number of successes, and the effect for each conjunctive clause belonging to the conjunctive clause set are calculated by a predetermined operation. A recommendation condition changing unit for generating a third recommendation condition based on the largest conjunctive clause and the first recommendation condition;
A recommendation condition optimizing device comprising:
(Appendix 2)
The recommendation condition changing unit generates the third recommendation condition,
The first and second recommendation conditions are converted into an additive standard form, and a conjunctive clause is obtained from the converted first and second recommendation conditions, and included in the obtained second recommendation condition. By generating a conjunctive clause set by excluding a conjunctive clause of the same kind as the conjunctive clause included in the obtained first recommendation condition from a clause;
The recommendation condition optimizing device according to appendix 1, wherein the third recommendation condition is generated by connecting the joint clause having the greatest effect to the first recommendation condition by disjunction.
(Appendix 3)
The recommendation condition changing unit generates the third recommendation condition,
The first recommendation condition and the second recommendation condition are converted into an additive standard form, the first conjunction clause to be changed is selected from the converted first recommendation condition, and the converted second phrase is selected. Based on a conjunctive clause including one or more atomic logical expressions that are the same type as the first conjunctive clause included in the recommended condition and that are not atomic logical expressions constituting the first conjunctive clause. , Generate the conjunctive clause set,
A conjunctive clause having the greatest effect is connected to the first conjunctive clause by a conjunctive to generate a third conjunctive clause, and the first conjunctive clause is changed to the third conjunctive clause. The recommendation condition optimization device according to appendix 1, wherein the third recommendation condition is generated by replacing with a clause.
(Appendix 4)
The similarity calculation unit calculates the similarity as a process:
The first recommended content ID and the second recommendation are selected from recommended content attribute information including a recommended type indicating a recommended content type, position information indicating a position associated with the recommended content, and a recommended content ID. When the specific recommendation type associated with each of the content IDs and the specific position information are obtained, it is determined whether or not the obtained two recommended types match, and the obtained two recommended types match. When determined, based on the obtained two position information, a distance between the positions represented by the two position information is calculated, and the reciprocal of the calculated distance is added to the initialized similarity. The recommendation condition optimizing device according to any one of appendix 1 to appendix 3.
(Appendix 5)
The similarity calculation unit calculates the similarity as a process:
Based on the first recommended content ID and the second recommended content ID, a recommended content ID, a user ID that can identify a user to whom the recommended content has been distributed, and an expected behavior for the recommended content Based on the recommended effect information including the recommended effect indicating whether or not the user has performed, based on the user ID associated with the first recommended content ID and the user ID associated with the second recommended content ID The first user is calculated based on the user ID included in the first product set and the user ID representing the success of the recommendation effect associated with the first recommended content ID. A set is calculated, a second user set is calculated based on a user ID indicating that the recommendation effect associated with the second recommended content ID indicates success, and the calculated Calculating a union and a product set based on the first and second user sets, calculating a ratio of the product set to the calculated union, and adding the calculated ratio to the initialized similarity. 4. The recommended condition optimizing device according to any one of appendix 1 to appendix 3, which is a feature.
(Appendix 6)
Based on the recommendation effect information obtained over a predetermined period, the success rate, the number of successes and the effect for the first recommendation condition and the third recommendation condition are calculated by a predetermined calculation, and the calculated The effect on the first recommendation condition is compared with the effect on the second recommendation condition, and the recommendation condition of either the first recommendation condition or the third recommendation condition is determined according to the comparison result. Select the recommended effect measurement section
The recommendation condition optimizing device according to any one of appendix 1 to appendix 5, further comprising:
(Appendix 7)
The similarity calculation unit includes:
Recommended content associated with the first recommended content ID by the number of records included in either the recommended condition information or the recommended content attribute information, and recommended content associated with the second recommended content ID The recommendation condition optimizing device according to any one of appendix 1 to appendix 5, wherein the similarity is calculated.
(Appendix 8)
The recommended condition information is searched based on state information including a user ID that can identify the user and information indicating the user state, and as a result of the search, it is determined that the state information matches a specific recommended condition. A recommendation necessity determination unit for obtaining a recommended content ID associated with the recommendation condition,
Recommended content distribution that refers to the recommended content attribute information based on the recommended content ID and distributes recommended content to the target user based on attribute information associated with the recommended content ID from among the above And
The recommended condition optimizing device according to any one of appendix 1 to appendix 7,
A recommendation system comprising:
(Appendix 9)
The recommendation necessity determination unit includes:
Whether the specific recommendation condition is met by searching history information including the user ID and information representing a history of actions taken by the user and the recommendation condition information based on the state information. The recommendation system according to appendix 8, characterized by determining whether or not.
(Appendix 10)
The similarity between the recommended content associated with the first recommended content ID that can identify the first recommended condition and the recommended content associated with the second recommended content ID different from the first recommended content ID is calculated. And
A specific second recommended content ID similar to the recommended content associated with the first recommended content ID is determined from the calculated similarities, and a recommended condition including one or more recommended conditions and a recommended content ID A first recommendation condition associated with the first recommended content ID and a second recommendation condition associated with the determined second recommended content ID are determined from the information, and the determined first Based on the first and second recommendation conditions, a conjunctive clause set is generated, the success rate, the number of successes, and the effect for each conjunctive clause belonging to the conjunctive clause set are calculated by a predetermined operation. Generating a third recommendation condition based on the largest conjunctive clause and the first recommendation condition;
A method for optimizing recommended conditions.
(Appendix 11)
A computer program for controlling the operation of the recommended condition optimizing device, the computer program,
The similarity between the recommended content associated with the first recommended content ID that can identify the first recommended condition and the recommended content associated with the second recommended content ID different from the first recommended content ID is calculated. Function to
A specific second recommended content ID similar to the recommended content associated with the first recommended content ID is determined from the calculated similarities, and a recommended condition including one or more recommended conditions and a recommended content ID A first recommendation condition associated with the first recommended content ID and a second recommendation condition associated with the determined second recommended content ID are determined from the information, and the determined first Based on the first and second recommendation conditions, a conjunctive clause set is generated, the success rate, the number of successes, and the effect for each conjunctive clause belonging to the conjunctive clause set are calculated by a predetermined operation. A function for generating a third recommendation condition based on the largest conjunctive clause and the first recommendation condition;
A computer program characterized in that a computer is realized.

(Hardware configuration example)
Each unit shown in the drawings in the embodiment described above can be regarded as a function (processing) unit (software module) of a software program. Each of these software modules may be realized by dedicated hardware. However, the division of each part shown in these drawings is a configuration for convenience of explanation, and various configurations can be assumed for mounting. An example of the hardware environment in this case will be described with reference to FIG.

  FIG. 20 is a diagram illustrating an exemplary configuration of an information processing apparatus 300 (computer) that can execute the recommended condition optimization apparatus according to the exemplary embodiment of the present invention. That is, FIG. 20 shows the recommended condition optimization apparatus 1 shown in FIG. 1, the recommended condition optimization apparatus 10 shown in FIG. 9, the recommended condition optimization apparatus 20 shown in FIG. 12, and the recommendation shown in FIG. A hardware environment capable of realizing each function in the above-described embodiment, which is a configuration of a computer (information processing apparatus) such as a server that can implement the entire or a part of the recommended condition optimizing apparatus 30 Represents.

  An information processing apparatus 300 illustrated in FIG. 20 includes a CPU (Central_Processing_Unit) 301, a ROM (Read_Only_Memory) 302, a RAM (Random_Access_Memory) 303, a hard disk 304 (storage device), and a communication interface (Interface: hereinafter “I”). / F ”) 305, and a reader / writer 308 capable of reading and writing data stored in a storage medium 307 such as a CD-ROM (Compact_Disc_Read_Only_Memory), and these components are connected via a bus 306 (communication line). It is a general computer.

  The present invention described by taking the above-described embodiment as an example is the block configuration diagram (FIG. 1, FIG. 9, FIG. 12, FIG. 18) referred to in the description of the information processing apparatus 300 shown in FIG. This is achieved by supplying a computer program capable of realizing the functions of the flowcharts (FIGS. 3, 5, 10, and 13) and then reading the computer program to the CPU 301 of the hardware and executing it. Further, the computer program supplied to the apparatus may be stored in a nonvolatile storage device such as a readable / writable temporary storage memory (RAM 303) or a hard disk 304.

  In the above case, the computer program can be supplied to the hardware by a method of installing the computer program via various storage media 307 such as a CD-ROM or a communication line such as the Internet. Currently, a general procedure can be adopted, such as a method of downloading from the outside. In such a case, the present invention can be regarded as being configured by a code constituting the computer program or a storage medium storing the code.

DESCRIPTION OF SYMBOLS 1 Recommendation condition optimization apparatus 2 1st similarity calculation part 3 1st recommendation condition change part 4 Memory | storage part 10 Recommendation condition optimization apparatus 11 2nd recommendation condition change part 20 Recommendation condition optimization apparatus 21 2nd similarity Degree calculation unit 30 Recommended condition optimization device 31 Recommended effect measurement unit 50 Recommendation system 51 Recommended content distribution unit 52 Recommendation necessity determination unit 101 First ID information 102 Recommended content attribute information 103 Recommendation condition information 104 Recommendation effect information 105 Recommendation condition Information 106 Recommended effect information 107 Status information 108 History information 300 Information processing device 301 CPU
302 ROM
303 RAM
304 Hard Disk 305 Communication Interface 306 Bus 307 Storage Medium 308 Reader / Writer

Claims (10)

The similarity between the recommended content associated with the first recommended content ID that can identify the first recommended condition and the recommended content associated with the second recommended content ID different from the first recommended content ID is calculated. A similarity calculation unit,
A specific second recommended content ID similar to the recommended content associated with the first recommended content ID is determined from the calculated similarities, and a recommended condition including one or more recommended conditions and a recommended content ID A first recommendation condition associated with the first recommended content ID and a second recommendation condition associated with the determined second recommended content ID are determined from the information, and the determined first Based on the first and second recommendation conditions, a conjunctive clause set is generated, the success rate, the number of successes, and the effect for each conjunctive clause belonging to the conjunctive clause set are calculated by a predetermined operation. A recommendation condition changing unit for generating a third recommendation condition based on the largest conjunctive clause and the first recommendation condition;
A recommendation condition optimizing device comprising: The recommendation condition changing unit generates the third recommendation condition,
The first and second recommendation conditions are converted into an additive standard form, and a conjunctive clause is obtained from the converted first and second recommendation conditions, and included in the obtained second recommendation condition. Generating the set of joint clauses by removing the joint clause of the same kind as the joint clause included in the first recommendation condition obtained from the clause;
The recommendation condition optimizing device according to claim 1, wherein the third recommendation condition is generated by connecting the joint clause having the greatest effect to the first recommendation condition by disjunction. . The recommendation condition changing unit generates the third recommendation condition,
The first recommendation condition and the second recommendation condition are converted into an additive standard form, the first conjunction clause to be changed is selected from the converted first recommendation condition, and the converted second phrase is selected. Based on a conjunctive clause including one or more atomic logical expressions that are the same type as the first conjunctive clause included in the recommended condition and that are not atomic logical expressions constituting the first conjunctive clause. , Generate the conjunctive clause set,
A conjunctive clause having the greatest effect is connected to the first conjunctive clause by a conjunctive to generate a third conjunctive clause, and the first conjunctive clause is changed to the third conjunctive clause. The recommendation condition optimization apparatus according to claim 1, wherein the third recommendation condition is generated by replacing with a clause. The similarity calculation unit calculates the similarity as a process:
The first recommended content ID and the second recommendation are selected from recommended content attribute information including a recommended type indicating a recommended content type, position information indicating a position associated with the recommended content, and a recommended content ID. When the specific recommendation type associated with each of the content IDs and the specific position information are obtained, it is determined whether or not the obtained two recommended types match, and the obtained two recommended types match. When determined, based on the obtained two position information, a distance between the positions represented by the two position information is calculated, and the reciprocal of the calculated distance is added to the initialized similarity. The recommendation condition optimizing device according to any one of claims 1 to 3. The similarity calculation unit calculates the similarity as a process:
Based on the first recommended content ID and the second recommended content ID, a recommended content ID, a user ID that can identify a user to whom the recommended content has been distributed, and an expected behavior for the recommended content Based on the recommended effect information including the recommended effect indicating whether or not the user has performed, based on the user ID associated with the first recommended content ID and the user ID associated with the second recommended content ID The first user is calculated based on the user ID included in the first product set and the user ID representing the success of the recommendation effect associated with the first recommended content ID. A set is calculated, a second user set is calculated based on a user ID indicating that the recommendation effect associated with the second recommended content ID indicates success, and the calculated Calculating a union and a product set based on the first and second user sets, calculating a ratio of the product set to the calculated union, and adding the calculated ratio to the initialized similarity. The recommendation condition optimizing device according to any one of claims 1 to 3, wherein the recommendation condition optimizing device is provided. Based on the recommendation effect information obtained over a predetermined period, the success rate, the number of successes and the effect for the first recommendation condition and the third recommendation condition are calculated by a predetermined calculation, and the calculated The effect on the first recommendation condition is compared with the effect on the second recommendation condition, and the recommendation condition of either the first recommendation condition or the third recommendation condition is determined according to the comparison result. Select the recommended effect measurement section
The recommendation condition optimization apparatus according to claim 1, further comprising: The recommended condition information is searched based on state information including a user ID that can identify the user and information indicating the user state, and as a result of the search, it is determined that the state information matches a specific recommended condition. A recommendation necessity determination unit for obtaining a recommended content ID associated with the recommendation condition,
Recommended content distribution that refers to the recommended content attribute information based on the recommended content ID and distributes recommended content to the target user based on attribute information associated with the recommended content ID from among the above And
A recommendation condition optimization device according to any one of claims 1 to 6,
A recommendation system comprising: The recommendation necessity determination unit includes:
Whether the specific recommendation condition is met by searching history information including the user ID and information representing a history of actions taken by the user and the recommendation condition information based on the state information. The recommendation system according to claim 7, wherein a determination is made. The similarity between the recommended content associated with the first recommended content ID that can identify the first recommended condition and the recommended content associated with the second recommended content ID different from the first recommended content ID is calculated. And
A specific second recommended content ID similar to the recommended content associated with the first recommended content ID is determined from the calculated similarities, and a recommended condition including one or more recommended conditions and a recommended content ID A first recommendation condition associated with the first recommended content ID and a second recommendation condition associated with the determined second recommended content ID are determined from the information, and the determined first Based on the first and second recommendation conditions, a conjunctive clause set is generated, the success rate, the number of successes, and the effect for each conjunctive clause belonging to the conjunctive clause set are calculated by a predetermined operation. Generating a third recommendation condition based on the largest conjunctive clause and the first recommendation condition;
A method for optimizing recommended conditions. A computer program for controlling the operation of the recommended condition optimizing device, the computer program,
The similarity between the recommended content associated with the first recommended content ID that can identify the first recommended condition and the recommended content associated with the second recommended content ID different from the first recommended content ID is calculated. Function to
A specific second recommended content ID similar to the recommended content associated with the first recommended content ID is determined from the calculated similarities, and a recommended condition including one or more recommended conditions and a recommended content ID A first recommendation condition associated with the first recommended content ID and a second recommendation condition associated with the determined second recommended content ID are determined from the information, and the determined first Based on the first and second recommendation conditions, a conjunctive clause set is generated, the success rate, the number of successes, and the effect for each conjunctive clause belonging to the conjunctive clause set are calculated by a predetermined operation. A function for generating a third recommendation condition based on the largest conjunctive clause and the first recommendation condition;
A computer program characterized in that a computer is realized.

Images