JP2013033363A - Provided information selecting device, method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the selection precision of information unknown and appropriate to a group by reducing the problem of trade-off between information unknown and appropriate to the group and inappropriate information.SOLUTION: Preference information of an individual belonging to a group and data for information provision are retrieved from a memory to calculate preference scores of the individual for the retrieved data for information provision. On the basis of the data for information provision and behavior history data, a degree of selection tendency of the designated group for the data for information provision is calculated. Feature vectors of the data for information provision and information indicative of behavior histories are calculated on the basis of the calculated preference scores of respective individuals and the degree of selection tendency of the group, and similarities between the feature vectors are obtained, and data for information provision corresponding to recommendation scores whose similarities of feature vectors calculated with respect to respective pieces of data for information provision exceed a threshold is selected as data appropriate to the group.

Description

  The present invention relates to a provision information selection apparatus, method, and program for selecting and providing appropriate information to a group including a plurality of members.

  In recent years, various services have been proposed to provide information users with information suitable for the information users using a computer network, one of which is appropriate for a group including a plurality of members with different preferences. Attention has been focused on services that use digital signage to provide useful information. As information to be provided, for example, advertisement information, news information, and restaurant information are promising.

  As a method for realizing this type of service, conventionally, for example, the relationship of the group is derived from the preference information of each member and the action history in the target group, and the information that is unknown and optimal for the target group using the relationship. A method of selecting is proposed. According to this method, it is possible to select information that has not been experienced and is suitable for the group by using the group relationship (see, for example, Non-Patent Document 1).

Shunichi Seko et al., “Video content recommendation for group based on viewing history and viewer preference”, IEEE International Conference on Consumer Electronics (ICCE), pp.359-360, 2011.

  However, although the conventional method is good at selecting information that is unknown and optimal for the target group, it may also select information that is inappropriate for the target group at the same time. In addition, when trying to remove inappropriate information, unknown and optimum information may also be removed. As described above, in the conventional method, there is a trade-off relationship between information that is unknown and optimal for the target group and information that is inappropriate for the target group, and it may be difficult to achieve both.

  The present invention has been made paying attention to the above circumstances, and its purpose is to alleviate the trade-off problem that occurs between information that is unknown and appropriate for the group and inappropriate information. It is an object of the present invention to provide a provision information selection device, method, and program that improve the selection accuracy of information that is unknown and appropriate for the user.

  In order to achieve the above object, one aspect of the present invention is to first provide a plurality of data for providing information to a group, information indicating members of the group, information indicating preferences of the members, Information representing the behavior history of the group is acquired and stored in a storage medium. And when the group which is the provision target of the information providing data is specified, the information stored on the basis of the information indicating the member of the group stored in the storage medium and the information indicating the preference of the member is stored. Preference score information representing the strength of preference of each member of the designated group for each of a plurality of information providing data is calculated. At the same time, based on the information providing data and the information indicating the action history stored in the storage medium, information indicating the degree of selection tendency of the designated group for the information providing data is calculated. Then, based on the calculated preference score information of each member and information indicating the selection tendency degree of the group, feature vectors for the information providing data and information indicating the action history are calculated, respectively, and further providing information The degree of similarity between the feature vector calculated for the information providing data and the feature vector calculated for the information representing the action history is calculated for each data. Then, the similarity of the feature vector calculated for each of the information providing data is compared with a preset threshold value, and the information providing data corresponding to the preference score information whose similarity is larger than the threshold value, The data is selected as data suitable for the group.

  Therefore, a dimension of group selection tendency is added to the preference space, and information providing data that is a candidate for information provided to the group and behavior history data of the group are displayed on the expanded preference space. It is determined whether or not each information providing data is suitable for the target group by measuring the similarity at. In other words, in addition to the combination of preference scores of each member for information, the selection tendency of the group with respect to the meta information (information that characterizes the information such as genre) possessed by the information is considered, and data for providing information suitable for the group is Selected. As a result, it is possible to select information that has the advantage of being able to select unknown and appropriate information due to the relationship of the group, and the advantage of eliminating inappropriate information due to the selection tendency of the group, thereby eliminating the trade-off. Thus, information suitable for the group can be selected.

により算出し、次に記憶媒体に記憶された複数の情報提供用データのうち候補として選んだ情報提供用データＤに含まれるメタ情報の数をｍとしたとき、上記行動履歴情報から算出されたメタ情報Ｇの出現頻度ＧＦ(G) に基づいて、当該候補となる情報提供用データＤに対する上記集団の選択傾向度合いＢＳ(D) を

により算出することも特徴とする。
このようにすると、集団の行動履歴情報に含まれるジャンル情報等の出現頻度を利用して、各情報提供用データに対する集団の選択傾向度合いを適切に計算することが可能となる。 Further, according to one aspect of the present invention, when calculating the information indicating the degree of selection tendency, first, based on the information indicating the action history stored in the storage medium, the total number of the action history information is N, an arbitrary When m is the number of meta information representing features of the action history included in the action history information, the appearance frequency GF (G) of certain meta information G is

Calculated from the action history information, where m is the number of meta information included in the information providing data D selected as candidates among the plurality of information providing data stored in the storage medium. Based on the appearance frequency GF (G) of the meta information G, the selection tendency degree BS (D) of the group with respect to the candidate information providing data D is calculated.

It is also characterized by calculating by
In this way, it is possible to appropriately calculate the group selection tendency with respect to each information providing data by using the appearance frequency of genre information or the like included in the group action history information.

  That is, according to one aspect of the present invention, it is possible to reduce a trade-off problem between information that is unknown and appropriate for a group and information that is inappropriate for the group. It is possible to provide a provision information selection device, method, and program that are improved.

The block diagram which shows the function structure of the provision information selection apparatus concerning one Embodiment of this invention. The block diagram which shows the specific structure of the provision information selection apparatus shown in FIG. The figure which shows an example of the information provision information data memorize | stored in the information provision data storage area provided in the provision information selection apparatus shown in FIG. The figure which shows an example of the group structure information memorize | stored in the group structure information storage area provided in the provision information selection apparatus shown in FIG. The figure which shows an example of the personal preference information memorize | stored in the personal preference information storage area provided in the provision information selection apparatus shown in FIG. The figure which shows an example of the action history information of the group memorize | stored in the action history information storage area of the group provided in the provision information selection apparatus shown in FIG. The figure which shows an example of the action frequency information of the group memorize | stored in the action frequency information storage area of the group provided in the provision information selection apparatus shown in FIG. The flowchart which shows the whole process sequence and the processing content by the provision information selection apparatus shown in FIG. The flowchart which shows the procedure and process content of a group action frequency information calculation process among the whole process procedures shown in FIG. FIG. 9 is a flowchart showing an information selection processing procedure and processing contents in the entire processing procedure shown in FIG. 8. FIG. The flowchart which shows the procedure of the process which determines the information suitable for a group among the procedures of the information selection process shown in FIG. The figure which shows the example which plotted the data for information provision in coordinate space by the conventional method. The figure which shows the example which plotted the data for information provision in coordinate space by the method which concerns on one Embodiment of this invention. The figure which shows an example of the calculation result of the similarity degree of the data for information provision by the provision information selection apparatus shown in FIG. 2, and action history data. The figure which shows an example of the action history data used for the outline | summary description of embodiment. The figure which shows an example of the data for information provision used for the outline | summary description of embodiment. The figure which shows an example of the preference score with respect to the genre used for the outline | summary description of embodiment.

Embodiments according to the present invention will be described below with reference to the drawings.
[Overview]
One embodiment of the present invention takes into account the tendency of information selected by the target group in addition to the relationship of the group, and thereby the trade-off that occurs between information that is unknown and appropriate for the group and information that is inappropriate for the group. Solve a problem.

  In a conventionally proposed method, for example, Non-Patent Document 1, a space (hereinafter referred to as a preference space) around each member's preference score is created, and an action history and information providing data in a group in the space are created. We selected unknown and optimal information based on the similarity. However, this method does not use the meta information (information that characterizes the information such as the genre) possessed by the information providing data and the action history information after creating the space. For this reason, even if the information has completely different meta information, the recommendation score (priority of recommendation) also has the same value if the combination of the preference scores of the members for the information is the same. As described above, since the information providing data and the meta information of the action history information are not used, unknown and optimum information for the target group can be selected, but a large amount of inappropriate information is selected. Problems are likely to occur.

Therefore, in one embodiment of the present invention, by taking into account the group selection tendency with respect to the information providing data and the meta information included in the action history information, it is possible to prevent inappropriate information from being selected while taking advantage of the advantages of the prior art. Is realized.
As a realization method, a dimension of group selection tendency is added to the preference space, and information providing data that is a candidate of information to be provided to the group, for example, “XX mapo tofu shop”, “△△ ramen shop”, etc. A set of restaurant information data and group action history data, for example, visited restaurant information data such as a visit to “□□ Hamburger Shop” are shown on the expanded preference space, and the similarity in that space is shown. A method of determining whether or not each information providing data is suitable for the target group by measuring is used.

  12 and 13 show examples in which information providing data A to C each having different meta information are plotted in a space according to the prior art and a space according to an embodiment of the present invention. 12 and 13, the action history high-density area is an area where the action history of the group is likely to be plotted, that is, an area where the recommendation ranking becomes higher as the information providing data is in and near this area. Is shown. Therefore, the recommendation order in FIG. 12 is A = C> B, and the recommendation order in FIG. 13 is A> B> C. In this way, by taking into account the group selection tendency with respect to the meta information of the data for providing information and behavior history information, it becomes possible to separate those that would have the same recommendation rank in the prior art, and the relationship between the groups It is possible to select information that has the advantage of being able to select unknown and appropriate information due to sex and the advantage of being able to remove inappropriate information due to the selection tendency of the group, which can solve the trade-off problem It becomes.

In one embodiment of the present invention, an example of using the appearance frequency of genre information appearing as an action history is given as a technique for calculating the group selection tendency degree. The frequency of appearance as an action history is calculated for each genre. If the action history data has a plurality of genre information as meta information, the appearance frequency of the genre is added by a value divided by the number of genres possessed. For example, when “XX mapo tofu shop” has two genres of Chinese and tofu, the degree of selection tendency with respect to Chinese and tofu is added by 1/2. That is, when the total number of action history data is N and the number of genres possessed by a certain action history data C is m, the appearance frequency GF (G) of a certain genre G is expressed by the following equation (1).

The group selection tendency degree for each information providing data is calculated using the appearance frequency of the meta information. For example, the group selection tendency degree for each piece of information providing data is the average of the frequency of meta information possessed by each piece of information providing data. For example, when the appearance frequency of “Chinese Chinese” is “2” and the appearance frequency of “Tofu” is “1”, the degree of selection tendency with respect to “XX Mapo Tofuya” having two meta information of Chinese and tofu Is
(Appearance frequency of Chinese food + Appearance frequency of tofu) / (Number of meta information of XX mapo tofu shop)
= (2 + 1) / 2
= 1.5
It becomes.
In addition, when there is no appearance frequency corresponding to the genre possessed, the value is treated as 0.

From the above, when the number of genres possessed by the information providing data D selected as candidates is m and the appearance frequency of a certain genre G is GF (G), the selection tendency degree BS for the information providing data D (D) is represented by the following formula (2).

A calculation example of the group selection tendency degree using the appearance frequency of the genre information described above will be described below.
For example, it is assumed that the behavior history of the target group is eaten at three stores shown in FIG. 15, and each store has the genre shown in FIG. 15 as meta information. In this case, the appearance frequency GF (x) of each genre x is as follows.
GF (Chinese) = 1/2 + 1/3 + 0
= 5/6
GF (tofu) = 1/2 + 0 + 0
= 1/2
GF (noodle) = 0 + 1/3 + 0
= 1/3
GF (Ramen) = 0 + 1/3 + 0
= 1/3
GF (Japanese food) = 0 + 0 + 1/3
= 1/3
GF (Izakaya) = 0 + 0 + 1/3
= 1/3
GF (Sake) = 0 + 0 + 1/3
= 1/3

Further, it is assumed that the information providing data includes, for example, four stores shown in FIG. In this case, the selection tendency degree BS (x) for each information providing data x is as follows.
BS (□□ Chinese restaurant) = 1/2 (5/6 + 0)
= 5/12
BS (◇◇ Udon) = 1/2 (1/3 + 0)
= 1/6
BS (XX steak restaurant) = 1/3 (0 + 0 + 0)
= 0
BS (△△ small dish) = 1/3 (0 + 1/3 + 1/3)
= 2/9

As a method of calculating the similarity between the information providing data and the group action history data, an example using the Euclidean distance is given.
FIG. 14 shows an example in which three group action history data and one information providing data are plotted on a space composed of the preference scores of the user A and the user B and the group selection tendency degree. The group action history data and the information providing data are both represented by vectors having the user A's preference score, the user B's preference score, and the group selection tendency as components. When the number of genres possessed by a certain information providing data D is m and the preference score of the user U for a certain genre G is GP (U, G), the preference score of the user U for a certain information providing data D is CP (U , D) is expressed by the following equation (3). It is assumed that the preference score GP (U, G) of the user U for a certain genre G is acquired in advance through a questionnaire or the like.

For example, it is assumed that the preference scores for the genre of user A and user B are the values shown in FIG. In this case, each user's preference score for XX mapo tofu shop with Chinese and tofu genres is as follows.
CP (User A, Mao Tofu Shop) = 1/2 (5 + 4)
= 4.5
CP (User B, Mao Tofu Shop) = 1 / (3 + 5)
= 4

Using this preference score (value calculated from FIG. 17) and the group selection tendency degree (value calculated from FIG. 15), the feature vector of the action history data of each group and the feature vector of the information providing data , The feature vector V (XX mapo tofu shop) for the group consisting of user A and user B for "XX mapo tofu shop" is as follows.
V (Mao tofu shop) = (4.5, 4, 5/12)

  However, if the value of each feature value is used as it is, the range of values of each dimension (axis) varies, so normalization is performed for each dimension (axis). When normalization is performed, the preference score of each member and the degree of selection tendency of the group fall within the values from 0 to 1, and thus it is possible to suppress variations in the range that each dimension can take.

For normalization, for example, a probability density function is used. Probability Density Function performs normalization using the probability that the value to be normalized or less will occur in the target data set (individual preference score, group selection tendency). Specifically, for the function of the probability distribution, an integrated value from −∞ to x (value to be normalized) is a value after normalization of x. Therefore, when the normalized value of the value x is F (x), the normalization formula can be expressed by the equation (4).

As a function of the probability distribution, for example, a normal distribution is used. That is, the function f (t) in the equation (4) is replaced by the following equation (5).

For example, when there are only six genres shown in FIG. 17 and the preference score for user A's genre is also the value shown in FIG. 17, normalized values of preference scores for user A's Chinese and tofu ( NGP (user A, Chinese) and (NGP (user A, tofu) are as follows.

Therefore, the normalized value NGP (user A, XX mapo tofu shop) of user A with respect to “XX mapo tofu shop” (having Chinese and tofu as genres) is the following value.
NGP (User A, Mao Tofu Shop)
= 1/2 (0.8861 + 0.6185)
≒ 0.7525
Similarly, normalized values NCP (user B, Chinese food), NCP (user B, Chinese tofu), and “XX mapo tofu shop” (Chinese and tofu as genres) The value (NCP (user B, XX mapo tofu shop)) after normalization of the preference score for “having” is as follows.

したがって、「○○麻婆豆腐屋」に対する正規化された特徴ベクトル（ＮＶ(○○麻婆豆腐屋)）は以下のようになる。
（ＮＶ(○○麻婆豆腐屋)）＝（０．７５２３，０．７６４９，０．８００５） Further, when the behavior history of the group consisting of the user A and the user B is as shown in FIG. 15, values after normalization of the degree of selection tendency for Chinese and tofu in this group (NGF (Chinese), NGF (Tofu) ) And “XX mapo tofu shop” (having Chinese and tofu as genres) after normalization (NBS (XX mapo tofu shop)) are as follows.

Therefore, the normalized feature vector (NV (XX mapo tofu shop)) for “XX mapo tofu shop” is as follows.
(NV (XX mapo tofu shop)) = (0.7523, 0.7649, 0.8005)

All the feature vectors of the information providing data / group action history data are normalized in this way. Using these normalized feature vectors, the similarity based on the Euclidean distance between the information providing data and the group action history data is calculated, and the sum of the similarities is used as the recommendation score of the information providing data. Taking FIG. 14 as an example, the information providing data calculates the degree of similarity with each of the three action history data, and the sum thereof becomes the recommended score of the information providing data. Expression (6) shows an example of an expression for calculating the similarity between the feature vector (NV (C)) of the information providing data (C) and the feature vector (NV (H)) of the action history data (H). Is.

As described above, the closer the Euclidean distance between the feature vector of the information providing data and the feature vector of the action history data is, the higher the similarity is. If such characteristics are provided, the calculation formula for the similarity is not limited to this. For example, as in equation (7), the norm may be raised to a power so that a high similarity is not calculated unless the Euclidean distance is closer.

この推薦スコアを求める処理を全ての情報提供用データに対して行い、推薦スコアが高い情報提供用データほど集団に適した情報とみなし、推薦スコアの高い順に推薦を行う。 This similarity is calculated for all action history data, and the sum is used as the recommendation score of the information providing data. Therefore, when the number of all action history data is N, the recommendation score (RS (C)) of the information providing data (C) can be expressed by the equation (8).

The process for obtaining the recommendation score is performed on all the information providing data, and the information providing data having a higher recommendation score is regarded as information suitable for the group, and the recommendation is performed in descending order of the recommendation score.

  As described above, the preference score of each member and the behavior history of the group are used as input data, and whether or not the information providing data is information suitable for the target group is determined from the preference score of each member and the group selection tendency degree. In such a space, it is possible to select information suitable for the group in which the trade-off that has been a problem in the conventional method is eliminated by performing evaluation using the similarity to the group's behavior history data.

[Functional configuration of the device]
FIG. 1 is a block diagram showing a functional configuration of a provision information selection apparatus according to an embodiment of the present invention.
A provided information selection device according to an embodiment of the present invention includes an input / output interface 1 for inputting / outputting data to / from an input device, a display device, and a storage device (not shown), and a program and data for storing the program and data. A storage unit 2, an information selection unit 3, a group behavior frequency information calculation unit 4, a communication interface unit 5 for transmitting / receiving information to / from other systems via a communication network (not shown), and a start trigger And a timer unit 6 for generating.

  The input / output interface unit 1 or the communication interface unit 5 receives personal preference information, group action history information, group configuration information, and information provision data from an input device, storage device, information provider or information user terminal, Each received input data is stored in the storage unit 2. However, regarding the action history information of the group among the input data, the input data is supplied to the action frequency information calculation unit 4 of the group. Upon receiving the behavior history information of the group, the behavior frequency information calculation unit 4 of the group calculates behavior frequency information for each genre based on the group ID and items included in the behavior history information of the group, and the calculation The result is stored in the storage unit 2.

  The input / output interface unit 1 gives the information selection unit 3 the start trigger input from the operator or the timer unit 6. When the information selection unit 3 receives the start trigger, the information selection unit 3 requests the storage unit 2 for necessary data and receives the corresponding data. Then, based on the received data, information suitable for the specified group is selected from the information providing data, and the information suitable for the selected group is stored in the storage unit 2. If necessary, information suitable for the group is presented from the input / output interface unit 1 to the operator, and further transmitted from the communication interface unit 5 to the terminal of the information user.

  The information selection unit 3 includes a personal information preference score calculation unit 31, a group selection tendency calculation unit 32, and a recommendation score calculation unit 33 as functions necessary for carrying out the home embodiment. Details of these functions will be described later.

[Specific configuration of device]
FIG. 2 is a block diagram showing a specific configuration of the provided information selection apparatus shown in FIG.
As shown in the figure, the provided information selection apparatus includes a central processing unit (CPU) 11, a program memory 13 and a data memory 14 are connected to the CPU 11 via a bus 12, and a communication interface 16 is further connected. The input / output interface 17 and the timer interface 20 are connected.

  The communication interface 16 includes an antenna 15 and communicates with a portable terminal (not shown) possessed by an information user via a communication network such as an IP network. An input device 18 and a display device 19 are connected to the input / output interface 17. The input device 18 includes a keyboard and mouse or a tablet type input device. The display device 19 includes a liquid crystal display (LCD) or an organic EL display. The timer interface 20 receives the start trigger signal generated from the timer 21 and notifies the CPU 11 of it.

  The data memory 14 stores personal preference information storage area 141, group behavior history information storage area 142, and preference cluster information storage as areas for storing information data necessary for carrying out one embodiment of the present invention. An area 143, an information providing data storage area 144, an information storage area 145 suitable for the group, a group configuration information storage area 146, and a group action frequency information storage area 147 are provided.

  The information providing data storage area 144 is used for storing providing data to be provided to a group of information users. FIG. 3 shows an example of stored information providing data. This data introduces a restaurant to an information user, defines a store name as an item name, and associates information indicating the store's genre name, location, and business hours with this store name (meta Information). In this way, when eating out is taken as an example, a set of information such as an item name (store name) and a genre name becomes one data. One data may have a plurality of genres, and even if there are n genres that can be set, it is not always necessary to have n genres. Moreover, if it is an item limited to a place and time like a restaurant, the limitation conditions (location location, opening time, etc.) may be included in one data.

  The group composition information storage area 146 is used to store information representing the composition of the information user group. FIG. 4 shows an example of the stored group composition information, and this information is formed by associating the user ID of the information user with the ID of the group to which the information user belongs. One information user can belong to a plurality of groups.

  The personal preference information storage area 141 is used for storing information representing information user preferences. FIG. 5 shows an example of stored personal preference information. This information consists of information relating to the genre name and the preference score associated with the user ID of the information user. The preference score is a quantification of the information user's preference with respect to the genre, and is calculated, for example, by a method described in Non-Patent Document 1 or by conducting a questionnaire to the information user.

  The group action history information storage area 142 is used to store group action history information. FIG. 6 shows an example of the action history information of the stored group. This information consists of information relating to the group ID of the group and the date and time of action, item name (restaurant name) and expense. For example, when an information user visits a restaurant in a group, the group action history information includes information indicating the date and time of action, the item name (restaurant name), and the cost using the mobile terminal. The provision information selection device receives the input information and associates it with the group ID of the group to which the information user belongs.

  The group action frequency information storage area 147 is used to store information representing the action frequency for each group. FIG. 7 shows an example of the stored group behavior frequency information. The behavior frequency information of the group is formed by associating information indicating the genre name and the behavior frequency with the group ID of the group. The action frequency is obtained, for example, by accumulating the number of visits in association with a genre called Chinese when visiting a Chinese restaurant.

  The preference cluster information storage area 143 is used to store information indicating the degree of preference of each information user belonging to the group for each item. The preference cluster information is formed by associating a group ID of a group with an item name (store name) and a preference cluster number of each user belonging to the group. The preference cluster number is a number indicating which section the target preference score corresponds to when the theoretical minimum and maximum sections of the preference score are divided into n. For example, when the theoretical minimum value is 1, the theoretical maximum value is 5, and the data is divided into 10 sections, 1.0 to 1.4 are sections 1, and 1.4 to 1.8 are sections 2. .., 4.6 to 5.0 are section numbers such as section 10, and each number is referred to as a preference cluster number.

  The program memory 13 includes an input / output control program 131, an individual preference score calculation program 132, a group selection tendency degree calculation program 133, and a group action frequency as application programs necessary for carrying out the present invention. A calculation program 134, a recommendation score calculation program 135, and an information selection program 136 suitable for the group are stored.

  The input / output control program 131 receives personal preference information, group action history information, group configuration information, and information provision data from the input device 18, external storage device, information provider terminal, or information user terminal. 17 or via the communication interface 16. Each received input data is stored in the personal preference information storage area 141, the group action history information storage area 142, the group configuration information storage area 146, and the information providing data storage area 144 in the data memory 14, respectively. The CPU 11 is caused to execute the storing process.

  The personal preference score calculation program 132 is for realizing the function of the personal preference score calculation unit 31 described in FIG. 1. When a group to be provided with data for providing information is specified, an information selection process is performed. In the process, a plurality of information stored in the information providing data storage area 144 is stored based on the information indicating the members of the group stored in the group configuration information storage area 146 and the information indicating the preferences of the members. The CPU 11 is caused to execute processing for calculating preference score information indicating the strength of preference of each member of the designated group for each piece of information providing data.

  The group selection tendency degree calculation program 133 is for realizing the function of the group selection tendency degree calculation unit 32 shown in FIG. 1. First, the action history information stored in the action history information storage area 142 of the group is shown. Based on the above, the appearance frequency of meta information representing the feature of the action history, for example, genre information is calculated. Next, for each piece of information providing data stored in the information providing data storage area 144, the degree of selection tendency of the group with respect to the information providing data is calculated using the calculated appearance frequency of the genre information. The CPU 11 is caused to execute the processing to be performed.

  When the group behavior frequency calculation program 134 receives new group behavior history information, the past group behavior history to which the same ID as the identification ID (group ID) included in the group behavior history information is given. Information is read from the collective action history information storage area 142. Then, the CPU 11 performs processing for calculating the behavior frequency information of the group based on the read group behavior history information and storing the calculation result in the behavior frequency information storage area of the group in the data memory 14. Let it run.

  The recommendation score calculation program 135 represents the information providing data and the action history based on the calculated preference score information of each member and information indicating the selection tendency degree of the group in the course of the information selection process. Each of the feature vectors for the information is calculated, and for each information providing data, the similarity between the feature vector calculated for the information providing data and the feature vector calculated for the information representing the action history is calculated. . Then, the CPU 11 is caused to execute a process of setting the sum of the similarities as a recommendation score of the information providing data.

  The information selection program 136 suitable for the group includes individual preference information storage area 141, group behavior frequency information storage area 147, group configuration information storage area 146, and information providing data storage area 144 in the data memory 14. Preference information, group behavior frequency information, group composition information, and information providing data are read out. Then, based on the calculated recommendation score, the CPU 11 is caused to execute a process of selecting a predetermined number of pieces of information providing data in descending order of the recommendation score.

[Device operation]
Next, the operation of the provided information selection apparatus configured as described above will be described.
The provided information selection device executes processing in the following procedure. FIG. 8 is a flowchart showing the overall processing procedure and processing contents.

(1) Input Data Storage Processing The provided information selection device monitors data input under the control of the input / output control program 131 in a standby state. In this state, when data is received via the input / output interface 17 or the communication interface 16, it is first determined in step S1 whether the input data is a start trigger. If the result of this determination is not a start trigger, it is subsequently determined in step S3 whether the input data is group action history information or other data. If it is data other than the group action history information, the input data is stored in the corresponding storage area in the data memory 14 in step S7.

  For example, if the input data is information providing data sent from the information provider terminal, the data is stored in the information providing data storage area 144. At this time, if the same item name as the input data is already stored in the information providing data storage area 144, the corresponding data is overwritten. In addition, if the input data is personal preference information sent from an information user's mobile terminal or the like, it is stored in the personal preference information storage area 141. At this time, if the same user ID and genre name as the input data are already stored in the preference information storage area 141, the corresponding data is overwritten, and otherwise added. Similarly, if the input data is group composition information, the input data is stored in the group composition information storage area 146. At this time, if the same user ID as that of the input data is already stored in the group composition information storage area 146, the corresponding data is overwritten, otherwise it is additionally written.

(2) Group Behavior Frequency Calculation Processing On the other hand, it is assumed that the input data is information representing the behavior history of the group. In this case, the provided information selection device stores the input group action history information in the group action history information storage area 142 in the data memory 14 under the control of the input / output control program 131 in step S4. Next, in step S5, the group behavior frequency calculation program 134 is activated, and the process for calculating the group behavior frequency under the control of the behavior frequency calculation program 134 is executed as follows. FIG. 9 is a flowchart showing the processing procedure and processing contents.

  That is, in step S51, the provided information selection device first selects a combination of group ID and item name from the group action history information stored in the group action history information storage area 142. The combinations of group IDs and item names to be selected are all combinations of group IDs and item names included in the behavior history information input this time that have not been subjected to action frequency addition processing. It is a combination. For example, if the group ID of the first record of the action history information of the group input this time is “Group 1” and the item name is “XX Mapo Tofuya”, “Group 1 + OO Mapo Tofuya” It becomes a pair.

  In step S52, the provided information selection device searches the information providing data storage area 144 for a genre name corresponding to the selected item name. For example, if the item name is “XX Mapo Tofuya” and “Chinese” and “Tofu” are registered as the genre names corresponding to the “XX Mapo Tofuya” in the information providing data, Search for “Chinese” and “Tofu” as genre names.

  Subsequently, in step S53, the provided information selection device performs an action frequency addition process corresponding to the set of the selected group ID and genre name as follows. That is, first, the behavior frequency information of the group corresponding to the set of the selected group ID and genre name is read from the behavior frequency information storage area 147 of the group. And the action frequency of the read action frequency information of the group is incremented (+1 addition). For example, if the selected group ID is “Group 1” and the genre names are “Chinese Chinese” and “Tofu”, the group's action frequency information is “Group 1 / Chinese” and “Group 1 / Tofu”. Increment action frequency of some data. Then, the action frequency information corresponding to the same group ID and genre name stored so far is updated to the action frequency after the increment. If no action frequency information of the group corresponding to the selected group ID / genre name exists in the storage area 147, the action frequency value corresponding to the pair of the selected group ID and genre name is set to “ 1 ”, which is used as new stored information.

  Finally, in step S54, the provided information selection device determines whether there is a combination that has not yet undergone addition processing among the combinations of the group ID and the genre name included in the group action history information input this time. Determine whether. As a result of the determination, if there remains a combination that has not been subjected to the addition process, the process returns to step S51 to execute the action frequency addition process for the combination. On the other hand, if there is no combination that has not been subjected to addition processing, all the behavior frequency information after the addition processing is stored in the storage area 147 in step S6, and then the group behavior frequency information is stored. The calculation process is terminated.

(3) Information Providing Data Selection Process Now, assume that a start trigger is generated from the timer 21 in this state, or an operator inputs a start trigger at the input device 18. The start trigger always includes a start flag for executing an information selection process suitable for the group and a group ID as an information selection target, and also includes other information as necessary. The group ID is used for the purpose of extracting information matching the group ID from the storage unit 2. The other information is, for example, information such as a place and a time zone, and is not necessarily included. The location and time zone information may be the location and time zone where the start trigger sender is present, or the location and time zone where the start trigger sender is going to go. Note that the user ID, group ID, genre name, and item name are unified in all information data, and the same ID and name indicate the same thing.

  When the start trigger is input, the provision information selection device shifts from step S1 to step S2, where the information provision data selection process is executed as follows. FIG. 10 is a flowchart showing the control procedure and control contents.

(3-1) Retrieval of Stored Data That is, the provided information selection device first searches the group behavior frequency information, the group composition information, and the group behavior history information corresponding to the group ID included in the received start trigger in step S21. Data is read from the storage areas 147, 146, 142 in the data memory 14. Subsequently, in step S22, the corresponding personal preference information is read from the storage area 141 based on the user ID included in the read group composition information.

In step S23, the read personal preference information and group behavior frequency information are normalized. This normalization is performed for each data group, for example, the preference information of the user 1, the preference information of the user 2, and the behavior frequency information of the group. For normalization, for example, a probability density function using a normal distribution is used. When the preference information for the data to be normalized, for example, user A's Chinese, is x, the data set to which the data x belongs, for example, the average value of the set of preference information of the user A is μ, and the standard deviation value is σ, the data x The normalized value F (x) is obtained by the following equation.

  In step S24, the information providing data is read from the information providing data storage area 144. At this time, if the start trigger includes information indicating a location or time zone, only data that matches the location or time zone is read out. For example, if “Yokohama” is included in the location information of the start trigger and “18:00:00 to 19:00” is included in the time zone information, the location information of the information providing data matches “Yokohama”, and the business hours Only data whose information is in a range including “18: 0 to 19:00” is selectively read out. Note that the information providing data to be read does not need to be limited to one, and if there are a plurality of corresponding data, a plurality of data is read.

(3-2) Creation of Feature Vectors for Information Providing Data and Behavior History Information Now, when the reading of each information is completed, the provision information selection device subsequently creates feature vectors for the information provision data and behavior history information. The processing to be executed is executed as follows.
That is, first, in step S25, it is determined whether there is information providing data and action history information for which a feature vector has not been calculated yet. If there is data for providing information and action history information for which a feature vector has not been calculated, the process proceeds to steps S26 to S28, and if not, the process proceeds to step S29.

なお、この際に利用する嗜好スコアは、ステップＳ２３にて正規化された値を利用する。このように、選択された情報提供用データもしくは行動履歴情報に対する個人の嗜好スコアを、必要なユーザの分だけ算出する。 (3-2-1) Calculation of Preference Score In step S26, the provided information selection device selects from the data memory 14 in order to calculate a personal preference score for the selected information providing data or action history information. The genre name included in the data for information provision or the action history information and the personal preference information corresponding to the genre name and the target user are read out. Then, if the number of genres possessed by the selected information providing data or action history information data D is m, and the preference score of the user U with respect to a certain genre G is GP (U, G), the selected information providing data The user U's preference score CP (U, D) for the data D is calculated by the following equation.

Note that the preference score used at this time uses the value normalized in step S23. Thus, the personal preference score for the selected information providing data or action history information is calculated for the necessary users.

により算出する。 (3-2-2) Calculation of group selection tendency for information providing data In step S27, the providing information selection device calculates a group selection degree for the selected information providing data. The genre name included in the selected action history information and the action frequency information corresponding to the genre name and the target group are read from the data memory 14. Now, assuming that the number of all action history information is N and the number of genres possessed by a certain action history information C is m, the appearance frequency GF (G) of a certain genre G is

Calculated by

により算出する。 Subsequently, using the calculated appearance frequency GF (G) of a certain genre G, a selection tendency degree BS (D) for a certain information providing data D is calculated.

Calculated by

(3-2-3) Calculation of Feature Vector Subsequently, in step S28, the provided information selection apparatus calculates the user preference score CP (U, D) calculated in steps S26 and S27 and the group selection tendency degree BS ( Based on D), the feature vector for the selected information providing data and action history information is calculated as follows.

That is, the number of members of the group is N, the preference score of the user U with respect to the selected information providing data D is CP (U, D), and the degree of selection tendency with respect to the selected information providing data D is BS (D ), The feature vector (V (D)) of the selected information providing data D is
V (D) = {CP (U1, D), CP (U2, D), ..., CP (UN, D), BS (D)} (11)
Calculated by Then, the calculated feature vector (V (D)) of the information providing data D is stored in the data memory 14. The feature vector for the action history information is similarly calculated and stored in the data memory 14.

  When the processes in steps S26 to S28 described above are completed, the provision information selection device returns to step S25 and determines whether or not information provision data and action history information for which a feature vector has not yet been calculated exist. . Then, as long as the information providing data for which the feature vector has not been calculated and the action history information remain, a series of processes in steps S26 to S28 are repeated.

(3-3) Selection of information providing data (3-3-1) Calculation of similarity using feature vector When the calculation of feature vectors for all information providing data and action history information is completed, the information is provided. The information selection apparatus moves to step S29, and in this step S29, using the calculated information providing data and the feature vector of the action history information, it is determined whether the information providing data is suitable for the target group. To do. Here, the determination method using the Euclidean distance will be described as an example, and the process will be described. FIG. 11 is a flowchart showing the processing procedure and processing contents.

  That is, first, in step S31, it is determined whether there is information providing data for which a recommendation score has not yet been calculated. If there is uncalculated information provision data, a recommendation score calculation process is executed in steps S32 to S36, and if not, all calculated recommendation score information is stored in the data memory 14 and stored. The process proceeds to step S30 of FIG. That is, the recommendation score calculation process in steps S32 to S36 is repeatedly executed until the recommendation score calculation process for all the information provision data read from the information provision data storage area 144 in step S24 of FIG. 10 is completed. When calculation of the recommendation score for all information providing data is completed, the process proceeds to step S30 in FIG.

  In step S32, one of the information provision data for which the recommendation score has not been calculated is selected, and in step S33, the group behavior read from the group behavior history information storage area 142 in step S21 of FIG. In the history data, it is determined whether or not there is action history data for which similarity calculation has not yet been performed with the selected information providing data. As a result of the determination, if it exists, in step S34, one is selected from the action history data for which the similarity with the selected information providing data has not yet been calculated, and the process proceeds to step S35. To do.

により算出する。そして、この算出された類似度をデータメモリ１３に保存する。以上のステップＳ３４及びステップＳ３５による類似度の計算処理は、類似度を計算していない行動履歴データがなくなったことが上記ステップＳ３３で検出されるまで繰り返し実行される。 In step S35, the similarity between the selected information providing data and the action history data of the selected group is calculated. The similarity is calculated by using the Euclidean distance of the feature vector for each data. Note that the feature vector used is that calculated in step S28 of FIG. That is, the similarity between the feature vector (V (C)) of the selected information providing data (C) and the feature vector (V (H)) of the selected action history data (H) is:

Calculated by Then, the calculated similarity is stored in the data memory 13. The similarity calculation process in steps S34 and S35 is repeatedly executed until it is detected in step S33 that there is no action history data for which the similarity is not calculated.

により算出する。この算出された値を、上記選択された情報提供用データの推薦スコアとし、図１０のステップＳ３０に移行する。
かくして、全ての情報提供データについて推薦スコアが算出される。 (3-3-2) Calculation of recommendation score When the calculation of the similarity between the selected information providing data and all the action history data is completed, the process proceeds to step S36. In step S36, the recommended score of the selected information providing data is determined using the calculated similarity. The recommendation score of the information providing data is the sum of the similarities obtained in steps SS33 to S35. Therefore, if the number of all action history data is N, the recommendation score (RS (C)) of the information providing data (C) is

Calculated by The calculated value is used as the recommended score of the selected information providing data, and the process proceeds to step S30 in FIG.
Thus, a recommendation score is calculated for all information provision data.

(3-3-3) Selection and Output of Information Providing Data When the recommendation score calculation process in step S29 is completed, the provided information selecting apparatus proceeds to step S30. Then, based on the recommendation score calculated in step S29, information provision data suitable for the target group is selected. Specifically, all the information provision data that are candidates for provision are arranged in descending order of the recommendation score calculated in step S29, and the top n pieces of information provision data are selected. Note that n is set in advance or included in the start trigger information.

  The information provision data selected as the provision target is stored in the information storage area 145 suitable for the group in the data memory 14 as information provision data suitable for the target group. Then, the stored information provision data is displayed on a display device for digital signage installed on a street or a meeting place where the group exists, via the communication interface unit 5, thereby presenting the information to the members of the group. To do. Further, it is displayed on a display unit attached to the apparatus as necessary so that the system operator can confirm it, or transmitted to another system via the communication interface unit 5.

[Effects of Embodiment]
As described above in detail, in this embodiment, an information providing data set that is a set of information provided to the group, information indicating the structure of the group, and information indicating the preference of each individual constituting the group are acquired. The data is stored in the storage area of the data memory 14, and collective action history information is acquired. Based on this information, the collective action frequency is calculated and stored in the storage area. And when a start trigger occurs, for providing information corresponding to the preference information of individuals belonging to the group corresponding to the group ID specified by the start trigger, and the location and business hours specified by the start trigger Each of the data is searched, and a personal preference score for the searched information providing data is calculated. At the same time, based on the information providing data and the action history data, information indicating the degree of selection tendency of the designated group for the information providing data is calculated. Then, based on the calculated preference score information of each member and information indicating the selection tendency degree of the group, the feature vectors for the information providing data and the information indicating the action history are respectively calculated, and these characteristics are calculated. Preference score information in which the similarity is greater than the threshold by calculating the similarity between the vectors and comparing the similarity of the feature vector calculated for each of the information providing data with a preset threshold. Is selected as data suitable for the group.

  Therefore, when selecting information provision data suitable for the group, in addition to the combination of each member's preference score for the information provision data, the group selection tendency with respect to the genre of the information is taken into consideration, and information provision suitable for the group is provided. Data is selected. As a result, it is possible to select information that has the advantage of being able to select unknown and appropriate information due to the relationship of the group, and the advantage of eliminating inappropriate information due to the selection tendency of the group, thereby eliminating the trade-off. Thus, information suitable for the group can be selected.

  The present invention is not limited to the above embodiment. For example, the group may be two people or three or more people. Moreover, although the said embodiment demonstrated as an example the case where the information regarding a restaurant was handled as information provision data, it is also possible to handle the information regarding events, such as other goods and services, news, an advertisement, a movie, and a concert.

In addition, the installation location and configuration of the provided information selection device, the processing procedure and information for selecting information, and the like can be variously modified without departing from the scope of the present invention.
In short, the present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the constituent elements without departing from the scope of the invention in the implementation stage. Further, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, you may combine suitably the component covering different embodiment.

  DESCRIPTION OF SYMBOLS 1 ... Input / output interface part, 2 ... Memory | storage part, 3 ... Information selection part, 4 ... Collective action frequency information calculation part, 5 ... Communication interface part, 6 ... Timer part, 11 ... Central processing unit (CPU), 12 ... Bus, 13 ... Program memory, 14 ... Data memory, 15 ... Antenna, 16 ... Communication interface, 17 ... Input / output interface, 18 ... Input device, 19 ... Display device, 20 ... Timer interface, 21 ... Timer, 31 ... Personal Preference score calculation unit, 32 ... group selection tendency calculation unit, 33 ... recommendation score calculation unit, 131 ... input / output control program, 132 ... individual preference score calculation program, 133 ... group selection tendency calculation program, 134 ... Group behavior frequency calculation program, 135 ... Recommendation score calculation program, 136 ... Information suitable for group Selection program, 141 ... personal preference information storage area, 142 ... group action history information storage area, 143 ... preference cluster information storage area, 144 ... information providing data storage area, 145 ... information storage area suitable for group, 146 ... group composition information storage area, 147 ... group action frequency information storage area.

Claims (5)

A plurality of information providing data to be provided to the group, information representing the members of the group, information representing the preferences of the members, and information representing the behavior history of the group are obtained and stored in a storage medium. Means for storing;
When a group to be provided with the information providing data is designated, a plurality of stored data are stored based on information representing a member of the group stored in the storage medium and information representing a preference of the member. Means for calculating preference score information representing the strength of preference of each member of the designated group for each of the information providing data;
Based on the information providing data and the information indicating the action history stored in the storage medium, information indicating the degree of selection tendency of the designated group for the information providing data is calculated for each information providing data. Means,
Means for respectively calculating feature vectors for the information providing data and the information indicating the action history based on the calculated preference score information of each member and the information indicating the calculated selection tendency degree of the group; ,
Means for calculating the degree of similarity between the feature vector calculated for the information providing data and the feature vector calculated for the information representing the action history, for each of the information providing data;
The similarity of the feature vector calculated for each information providing data is compared with a preset threshold value, and the information providing data corresponding to the preference score information having a similarity greater than the threshold Providing information selecting apparatus, comprising: means for selecting data suitable for the device. The means for calculating the information indicating the degree of selection tendency is:
Based on information representing the action history stored in the storage medium, the total number of the action history information is N, and the number of meta information representing the feature of the action history included in any action history information is m And the appearance frequency GF (G) of certain meta information G

Means for calculating by
Appearance of meta information G calculated from the action history information, where m is the number of meta information included in the information providing data D selected as a candidate among the plurality of information providing data stored in the storage medium Based on the frequency GF (G), the selection tendency degree BS (D) of the group with respect to the candidate information providing data D is calculated.

The provision information selection device according to claim 1, further comprising: A plurality of information providing data to be provided to the group, information representing the members of the group, information representing the preferences of the members, and information representing the behavior history of the group are obtained and stored in a storage medium. The process of remembering,
When a group to be provided with the information providing data is designated, a plurality of stored data are stored based on information representing a member of the group stored in the storage medium and information representing a preference of the member. Calculating preference score information representing the strength of preference of each member of the designated group for each of the information providing data;
Based on the information providing data and the information indicating the action history stored in the storage medium, information indicating the degree of selection tendency of the designated group for the information providing data is calculated for each information providing data. Process,
A process of calculating feature vectors for the information providing data and the information indicating the action history, based on the calculated preference score information of each member and information indicating the calculated group selection tendency, ,
For each of the information providing data, a process of calculating the similarity between the feature vector calculated for the information providing data and the feature vector calculated for the information representing the action history,
The similarity of the feature vector calculated for each information providing data is compared with a preset threshold value, and the information providing data corresponding to the preference score information having a similarity greater than the threshold Providing information selection method comprising: selecting as data suitable for The process of calculating the information indicating the degree of selection tendency is as follows:
Based on information representing the action history stored in the storage medium, the total number of the action history information is N, and the number of meta information representing the feature of the action history included in any action history information is m And the appearance frequency GF (G) of certain meta information G

The process of calculating by
Appearance of meta information G calculated from the action history information, where m is the number of meta information included in the information providing data D selected as a candidate among the plurality of information providing data stored in the storage medium Based on the frequency GF (G), the selection tendency degree BS (D) of the group with respect to the candidate information providing data D is calculated.

The provided information selecting method according to claim 3, further comprising:   The program which makes a computer perform the process corresponding to the process with which the provision information selection method of Claim 3 or 4 is provided.

Images