JP2011118642A - System and method for analyzing growth potential of service scale, and program for the system and method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a technique for analyzing the growth potential of a service scale, wherein an objective index or a diagram which extracts whether user cluster groups on which user's interest and attention are reflected are variously formed taking into consideration the number of users or only a small number of cluster groups are formed is calculated from data and displayed. <P>SOLUTION: A system for analyzing the growth potential of a service scale includes: a means 206 for successively collecting using states of respective users in the whole users or a partial set of sample-extracted users to evaluate whether service for providing various pieces of information grows by continuous use of general consumers or the service continuously provides diversity sufficient for maintaining a growth scale or not; a means 208 for inputting the collected using states and decomposing the users into cluster groups or sub-cluster groups concerned with liking or interest of the various pieces of information; and a means 209 for calculating or displaying the strength and weakness of deviation to large scale cluster groups or the strength and weakness of deviation to small or medium scale cluster groups as an index of structural features of the decomposed cluster groups or the sub-cluster groups. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention is generally used in services that provide distribution or sharing of contents such as videos and music, or general information such as blogs and bulletin boards, services that provide sales or exchange of various products, or services that aggregate these. Predict or analyze whether users or users can continue to grow on a large scale, or can continue to grow users in the future It relates to a user-scale growth analysis evaluation, marketing, business model evaluation, or a service / system management method.

  Conventional market analysis and market forecasting technologies are as follows. First, there is a macro technique for extrapolating macro market values such as market size from product sales records, survey data such as voluntary sampling surveys, and questionnaire surveys.

  Next, from purchase behavior records, sales records, questionnaire surveys, etc., the purchase motivation can be verbalized and the product attributes are estimated, and the acceptability of the attributes, the selection probability of the purchase behavior options, etc. are estimated, and this is the customer attribute There is a micro technique such as integration for each segment. Furthermore, a hybrid method combining these micro and macro methods is adopted.

  Various error terms, dummy variables, hidden customer attribute variables, etc. have been added to these to explain market trends and improve prediction accuracy. These are the cases where the characteristics of the market can be inferred from similar products or similar products, or when the user can assume product understanding and is easy to estimate from purchase behavior records or questionnaire surveys, or the purchase motivation to be predicted This is a powerful method when it can be verbalized as an attribute, and the acceptability and selectivity of the attribute can be inferred from the purchase records of existing products and the probability of the purchase behavior and selection behavior can be easily estimated.

  On the other hand, services that provide distribution or sharing of general information such as videos and music, or general information such as blogs and bulletin boards, services that sell or exchange various products, or services that aggregate these, etc. These existing market analysis and market forecasting methods cannot be used for questions such as whether future growth potential and scale of growth can be maintained for new services.

  This is because not only individual contents and individual products but also a lot of contents, products, and information are exchanged or the characteristics of the platform are problems. Until now, only basics have been known about the growth of the service scale due to the increase in the number of users and usage frequency of the new service.

  This is because services that tend to be used by individual users after being established and used repeatedly are superior in terms of service scale growth and scale maintenance. (Nonpatent literature 1).

  However, this property of stickiness is almost the same feature as the fact that the service scale growth and scale maintenance characteristics are excellent. In order to provide a convincing answer to the question of whether future growth potential and scale of growth can be maintained for new services, the requirements for a sticky service need to be clarified There is.

  It is considered that the requirements for a sticky service are given as follows. In general, when individual consumers use content that may be divided according to tastes and interests, such as content such as videos and music, or general information such as blogs and bulletin boards, by consuming individual targets Individuals and consumers form complex relationships.

  The formation of this complex relationship is due to changes and adaptations of individual consumers, and is acquired by learning by individual consumers. In many cases, individual consumers are engaged in completely new learning. However, as always observed in learning, it is generally much better to learn additional content based on the content acquired once in the past. It is easy and is likely to occur as an action

  Accordingly, the diversity of tastes and interests is, for many individual consumers, not a characteristic of a single consumer but a characteristic of an aggregate of consumers in which a large number of consumers gather.

  And in order to continue to be used by many consumers and grow on a large scale, not only a variety of contents and information are supplied sequentially, but also individual contents, information, products, etc. For content, it is desirable that the content be continuous or relevant so that it can be based on what the consumer has learned in the past.

  In this way, the diversity and continuity of content, information, products, etc. provided to the user through the service increases the stickiness of the user of the service, maintaining high growth potential and scale of the service. It is thought that it leads to development.

  Therefore, in order to evaluate the user's ability to grow, maintain the scale, or develop the service, there is a mechanism in which content, information, products, etc. that can be accepted by the user are supplied in a diverse and continuous manner. It will be good if you can evaluate whether or not you are.

  And if the contents, information, or products handled by the service are supplied in a variety and continuously with content that suits the user's tastes and interests, many users can learn from the basic characteristics of learning. Uses the past and present or present and future learning contents in succession. Therefore, the usage behaviors of any two users will take similar usage behaviors if the learning content is shared, and will have very different usage behaviors if not shared.

  It is common knowledge that people today diversify their preferences. Therefore, the user forms various aggregate groups reflecting the overlap and difference of the contents to be learned through the learning contents. Therefore, if it is understood whether the user cluster group reflecting interest and interest is formed in a sufficiently diverse manner in view of the number of users or only a small number of cluster groups are formed, the growth potential of the service, the maintenance of the scale, or The ability to develop will be understood.

  Based on the above, objective indicators and charts that extract whether the user cluster group reflecting interest and interest are formed in a sufficiently diverse manner in view of the number of users or a small number of cluster groups are calculated and displayed from the data Can provide an objective assessment of the service's ability to grow, maintain, or develop, and make decisions that have significant business, economic, and social benefits. It becomes possible and has high industrial utility value.

K. Mohamed, L. Moez, and L. Vanessa, “Online consumer stickiness: a longitudinal study,” J. of Global Information Management, Jul. 2002. Nagaike Iwahori Linear Inequalities and Their Applications Iwanami Lecture Basic Mathematics Linear Algebra vii, Iwanami 1977.

  From the above, the object of the present invention is to provide services such as distribution or sharing of content such as videos and music, or general information such as blogs and bulletin boards, services that provide sales or exchange of various products, or a combination of these. Analyze, evaluate, or predict whether a service can continue to be used and grown by a large number of consumers, or whether large-scale users can be maintained in the future, Or a diverse group of users formed by overlapping and different content that each user learns through a series of consumption, resulting from general and common sense basic characteristics of learning to estimate In other words, regarding user cluster groups reflecting interests and interests, they are formed in various ways in terms of the number of users, or To provide a service scale growth analysis system, a service scale growth analysis method, and a program for the same, which are capable of calculating and displaying objective indicators and charts that extract whether a small number of cluster groups are formed or not. .

  This section describes services that distribute or share content such as videos and music, but services that provide distribution or sharing of general information such as blogs and bulletin boards, services that sell or exchange various products, or these The same applies to a service combining the above, and does not limit the scope of the invention.

  The present invention has the following configuration. In the present invention, a subset of the entire set of users determined by sharing some preference or interest is defined as a “cluster”, and a subset of user clusters determined by sharing or having a greater preference or interest Is defined as “sub-cluster”.

a) The service scale growth analysis system according to the present invention is diverse enough that services providing various types of information grow or maintain the scale of growth through the continuous use of general consumers. Is a service scale growth analysis system for evaluating whether or not a user is continuously provided, and sequentially collects the usage status of each user with respect to the entire user or a sampled subset of users. Means for receiving the collected usage status as input, and means for decomposing the user into a cluster group or sub-cluster group related to the preference or interest of the various information; and an index of structural features of the decomposed cluster group or sub-cluster group As a means of calculating or illustrating the strength of the bias toward the large-scale cluster group or the strength of the bias toward the small to medium-sized cluster group. It is characterized in. Note that large / medium / small of the cluster size is relative and is not limited to this. For example, the scale range is created logarithmically, and is divided into three equal parts. It can be considered to be a small scale area.

b) In the service scale growth analysis system, identifiers i, 1 ≦ i ≦ N are assigned to the users belonging to the entire user or a subset of the sampled users, and identifiers m, The usage status of each user is expressed as the number of times f (i, m) of the use or purchase of the target belonging to the various information to which the identifier m is assigned, for the user to which 1 ≦ m ≦ M is assigned and the identifier i is assigned. Sequentially collected data is input, and the similarity g (i, j) between the user with identifier i and the user with identifier j is g (i, j) = f (i, 1) f (j , 1) + f (i, 2) f (j, 2) +... + F (i, M) f (j, M) and N-th order symmetric matrix G = (g (i, j) 1 ≦ i, j ≦ N) as a preference or interest similarity matrix between users, and a means for decomposing users into clusters related to preferences or interests for the various information.

c) Further, in the service scale growth analysis system, means for calculating eigenvectors of an Nth-order symmetric matrix G that is a similarity matrix between users, and for each eigenvector v, vp (i) = Max {0, v (i)}, 1 ≦ i ≦ N and vm (i) = Max (0, −v (i)), 1 ≦ i ≦ N A means for calculating the strength s (v) of the eigenvector v with respect to the eigenvector v with respect to the eigenvector v or vm, and when the s (v) satisfies a certain reference value, the control of the eigenvector v It is characterized in that it is determined that the typical fixed-word composition corresponds to a cluster or sub-cluster, and the proportion of the cluster or sub-cluster in the whole eigenvector representing the correlation between users is extracted.

d) The service scale growth analysis system further includes means for calculating a cluster size distribution in the sense of the number of users of the cluster group or the sub-cluster group, and calculates an index indicating the bias of the cluster size distribution. Based on the above, it is possible to continuously respond to the variety of user preferences or interests in the various information by calculating or illustrating the strength of the bias toward the large-scale cluster group or the strength of the bias toward the small to medium-sized cluster group. It is characterized by evaluating whether or not.

e) In the service scale growth analysis system, the various types of information are content as video or music, information written on a blog / bulletin board, information on various products, or a combination thereof. It is characterized by that.

f) A program according to the present invention is a program characterized by causing a computer to function as each means in the service scale growth analysis system according to any one of claims 1 to 5.

g) The service scale growth analysis method according to the present invention is a method for analyzing contents as moving images or music, information written on a blog / bulletin board by a user, information on various products, or information combining these. Computers to assess whether the services they provide continue to provide enough diversity to continue to grow or maintain the scale of growth through general consumer use A service scale growth analysis method using the method of sequentially collecting the usage status of each user for the entire user or a subset of the sampled users, and using the collected usage status as input, Disassembling the user into clusters or sub-clusters relating to preferences or interests in the various information; Calculating or illustrating the strength of the bias toward the large-scale cluster group or the strength of the bias toward the small to medium-sized cluster group as an index of the structural characteristics of the cluster group or the sub-cluster group. It is a feature. Note that large / medium / small of the cluster size is relative and is not limited to this. For example, the scale range is created logarithmically, and is divided into three equal parts. It can be considered to be a small scale area.

h) Further, in the service scale growth analysis method, identifiers i, 1 ≦ i ≦ N are assigned to each user belonging to the entire user or a subset of sampled users, and identifiers m, The usage status of each user is expressed as the number of times f (i, m) of the use or purchase of the target belonging to the various information to which the identifier m is assigned, for the user to which 1 ≦ m ≦ M is assigned and the identifier i is assigned. Sequentially collected data is input, and the similarity g (i, j) between the user with identifier i and the user with identifier j is g (i, j) = f (i, 1) f (j , 1) + f (i, 2) f (j, 2) +... + F (i, M) f (j, M) and N-th order symmetric matrix G = (g (i, j) 1 ≦ i, j ≦ N) as a preference or interest similarity matrix between users, and a step of decomposing users into clusters related to preferences or interests for the various information.

i) Further, in the service scale growth analysis method, a step of calculating eigenvectors of an Nth-order symmetric matrix G that is a similarity matrix between users, and for each eigenvector v, vp (i) = Max {0, v (i)}, 1 ≦ i ≦ N and vm (i) = Max (0, −v (i)), 1 ≦ i ≦ N A step of calculating a strength s (v) where the eigenvector v is biased toward the fixed composition vp or vm with respect to the eigenvector v, and when the s (v) satisfies a certain reference value, the control of the eigenvector v It is characterized in that it is determined that a typical fixed-word composition corresponds to a cluster or sub-cluster, and the proportion of the cluster or sub-cluster in the entire eigenvector representing the correlation between users is extracted.

j) In the service scale growth analysis method, the method includes a step of calculating a cluster size distribution in the sense of the number of users of a cluster group or a sub-cluster group, and calculating an index indicating a bias of the cluster size distribution, Whether or not the various types of information continuously respond to a variety of user preferences or interests by calculating or illustrating the level of bias toward a large-scale cluster group or the level of bias toward a small to medium-sized cluster group It is characterized by evaluating.

  According to the present invention, a service that provides distribution or sharing of content such as video and music, or general information such as a blog / bulletin board, a service that provides sales or exchange of various products, or a service that aggregates these services A diverse group of users, in other words, interests, formed by the overlap and difference of what each user learns through a series of consumption, resulting from the general and common sense basic characteristics of learning Technology that calculates and displays objective indexes and charts that extract whether user clusters that reflect interests or interests are formed in a variety of ways or only a small number of cluster groups from the scale of users Can be provided.

It is a figure which shows the structural example of the service scale growth property analysis system which concerns on the Example of this invention. It is a figure which shows the processing flow of the service scale growth property analysis system which concerns on the Example of this invention. It is a figure showing the example of an output analyzed by the service scale growth analysis system concerning the example of the present invention. It is a figure showing the growth condition of the site seen from the frequency | count of use of the site containing the website analyzed in the example output of FIG.

(Overview)
First, the outline of the main configuration of the present invention will be described.
(1) <Means for reading recording data of usage activities used by users for each content and totaling the usage status of each content of each user>

  For example, the usage record data of all users who have used the service in a certain month or N users sampled are read. The usage record data includes information for identifying N users, information for identifying each content, and information on the number of times each user has used each content during the period.

  Information identifying each user is rewritten, each user is identified by a natural number i satisfying 1 ≦ i ≦ N, and information identifying each content is rewritten, and each content is identified by a natural number m satisfying 1 ≦ m ≦ M. Here, M is the total number of contents used by one of these N users during the month. Also, the usage frequency f (i, m) of the content of the identifier m of the user of the identifier i is totaled, and the usage status of the user of the identifier i is represented by the vector F (i) = (f (i, 1),. Calculate as f (i, m)).

(2) <Means for totalizing measures for measuring the similarity between users from data obtained by counting the usage status of each user>

  For example, by using a vector F (i) obtained by aggregating the usage status of the user of the identifier i from the number of usages f (i, m) of each content, the similarity between the users assigned the identifier i and the identifier j is expressed as F (i ) And F (j), g (i, j) = f (i, 1) f (j, 1) + f (i, 2) f (j, 2) + ... + f (i, M ) Count by f (j, M).

(3) <Regarding the formation of a user cluster group by sharing the content learned by the user from the aggregate data of the scale for measuring the similarity between each user, a large number of clusters or sub-clusters are formed when viewed from the user scale. > A means to aggregate indicators about whether or not the number of people remains

  For example, in the case where the data g (i, j) of the measure for similarity between the users assigned the identifier i and the identifier j is given by the above inner product, the Nth-order square matrix G = When (g (i, j); 1 ≦ i, j ≦ N) is a nonnegative symmetric matrix, eigenvalues and eigenvector sets obtained by spectrally decomposing G are calculated, and for each eigenvector v, vp ( i) = Max {0, v (i)}, 1 ≦ i ≦ N and vm (i) = Max (0, -v (i)), 1 ≦ i ≦ N Calculate vm. Here, v = vp-vm.

Furthermore, for each eigenvector v, the strength s (v) at which the eigenvector v is biased toward the fixed composition vp or vm is
s (v) = Max {| vp |, | vm |} / | v |
And is combined with the eigenvalue λ (v) of the eigenvector v and plotted as (λ (v), s (v)).

Here, for a vector u = (u (1), ..., u (N)) consisting of N real elements
| u | = Max {u (1),-u (1)} + ・ ・ ・ + Max {u (N),-u (N)}
And calculate. According to the so-called Perron-Frobenius theory (for example, Non-Patent Document 2, p145-p157), an exact cluster component without any overlap corresponds to an eigenvector v of s (v) = 1.

  Since clusters and sub-clusters usually allow some overlap, if the value of s (v) is close to 1, the eigenvector v is close to the Perron-Frobenius eigenvector, and the eigenvector v is It can be considered to correspond to a cluster or sub-cluster.

  Therefore, it can be seen that if the number of eigenvectors having a value of s (v) close to 1 is large from the whole eigenvector (that is, the dimension N of the matrix), the number of clusters or sub-clusters tends to be larger than the overall user scale. .

  At this time, the eigenvalue λ (v) gives a scale when the cluster or sub-cluster is viewed from the usage number (traffic). Therefore, the graph in which each eigenvector is plotted with (λ (v), s (v)) distinguishes the scale of the cluster or subcluster when viewed from the number of traffic (traffic), and the cluster or subcluster group is Gives information about whether there is more or less from the user scale.

(4) <Calculation of cluster size distribution based on the number of users in a cluster or sub-cluster regarding the formation of a user cluster group by sharing the content learned by users from the aggregated data of a measure for measuring the similarity between each user Means to do>

  For example, in the case where the data g (i, j) of the measure for similarity between the users assigned the identifier i and the identifier j is given by the above inner product, the Nth-order square matrix G = When (g (i, j); 1 ≦ i, j ≦ N) is a non-negative symmetric matrix, the strength that v is biased to the fixed-composition vp or vm for each eigenvector v obtained by spectrally decomposing G An eigenvector v having a value of s (v) greater than or equal to a positive constant s0 is considered to form a cluster or a sub-cluster. Here, s0 is a number close to 1 and smaller than 1.

At this time, for example, when the eigenvector v is biased toward the positive value component vp, the cluster or sub-cluster is set to a small positive number e close to 0.
{1 ≦ i ≦ N | vp (i) / | vp |> e}
Therefore, the scale according to the number of users of the cluster or sub-cluster is calculated by the number of elements of this set.

  In this way, the cluster size distribution based on the number of users in the cluster or sub-cluster group can be calculated by executing the determination by s0 and the calculation of the cluster size for each eigenvector. There are many clustering methods, and the method is not necessarily limited to the method described here, and any known method may be used.

  This section describes services that distribute or share content such as videos and music, but services that provide distribution or sharing of general information such as blogs and bulletin boards, services that sell or exchange various products, or these The same applies to a service that aggregates and does not limit the scope of the invention.

  For services that distribute or share content such as videos and music, a means for reading recorded usage activity data used by each user for each content, and summing up each user's usage status of each content, and each user Users by sharing the content learned by users from the data that aggregates the usage status of data, the means of aggregating measures that measure the similarity between each user, and the aggregated data of the measures that measure the similarity between each user With regard to the formation of cluster groups, there are various user cluster groups that reflect interests and interests by means of aggregating an index on whether a large number of clusters or sub-clusters are formed from the viewpoint of the user or a small number of clusters. Objective indicators and charts that extract whether they are formed in a small number of clusters or remain in a small number of cluster groups. To determine whether the content handled by the service is provided in a variety and continuously with content that suits user preferences and interests. You can know the basic information to do.

  From this, it becomes possible to evaluate and estimate the present and future stickiness of the user of the service, and it is possible to objectively evaluate the growth, maintenance or development of the user scale of the service.

  In addition to the above means, regarding the formation of a user cluster group by sharing the contents learned by the user from the aggregate data of the scale for measuring the similarity between the users, the cluster according to the number of users in the cluster or sub-cluster. By means of calculating the size distribution, it becomes possible to obtain more accurate information about whether the user cluster group reflecting interest and interest is formed in a variety of ways, or whether the number of cluster groups remains small, It is possible to make a more accurate determination as to the growth, maintenance or development of the user scale of the service.

(Example)
The following describes services that distribute or share content such as videos and music, but services that provide distribution or sharing of general information such as blogs and bulletin boards, services that sell or exchange various products, or The same applies to the aggregated services, and does not limit the scope of the invention.

  The service-scale growth analysis / evaluation system for a service according to the present invention reads recording data of usage activity used by a user for each content to a service that distributes or shares content such as video or music, A means for counting the usage status of each user's content, a means for counting the similarity between each user based on the data totaling the usage status of each user, and a scale for measuring the similarity between each user. Regarding the formation of user cluster groups by sharing the contents learned by users from the aggregated data, an index as to whether clusters or sub-clusters are formed in large numbers or only a small number when viewed from the user scale Shares what the user has learned from the means of aggregation and the aggregate data of the scale that measures the similarity between each user With respect to the formation of user clusters by Rukoto, and a means for calculating the distribution of the cluster size according to the number of users of a cluster or subcluster.

  FIG. 1 is a diagram showing a configuration example of a user-scale growth analysis / evaluation system for a service according to the present invention. FIG. 2 is a diagram showing a main processing flow of the growth analysis / evaluation system.

  A service-scale growth analysis / evaluation system 200 for a service according to the present invention includes a user interface processing unit 201, a display control unit 202, a central control unit 203, a main memory 204, a default setting unit 205, and usage status aggregation. A unit 206, an inter-user similarity totaling unit 207, a spectrum decomposing unit 208, a cluster structure formation state analyzing unit 209, and a cluster size distribution creating unit 210.

  Further, the cluster structure formation status analysis unit 209 has a standard composition calculation unit 2091, a bias index calculation unit 2092, and a plot data creation unit 2093.

  The growth analysis analysis system 200 is connected to a user interface 100 such as a display, a keyboard, and a mouse, a service use activity data storage unit 300, and an output storage unit 400.

  The operation of the service-scale growth analysis / evaluation system 200 for a service according to the present invention will be described in detail below using the configuration example of FIG. 1 and the processing flow of FIG.

  The user interface processing unit 201 receives, via the user interface 100, a command for the user of this system to control the system and execute target analysis, and transfers the command to the central control unit 203.

  The default setting unit 205 provides a function for setting a default for various parameters and designation requests of each unit, and performs a function of displaying, inputting, and saving a default setting request table (corresponding to step S10 in FIG. 2). .

  The service use activity data storage unit 300 reads and stores data on continuous recording of use of a target service or the like from the outside, and transfers data in response to a call.

  Specifically, the usage record data of all users who have used the service or N users sampled for the period determined by the default setting unit 205 is read. The usage record data includes information for identifying N users, information for identifying each content, and information on the number of times each user has used each content during the period.

  The usage status totaling unit 206 reads the recording data of the usage activity used by the user for each content, and executes a function of totalizing the usage status of each content of each user (corresponding to step S20 in FIG. 2).

  Specifically, the information for identifying each user is rewritten, each user is identified by a natural number i of 1 ≦ i ≦ N, the information for identifying each content is rewritten, and each of the natural numbers m by 1 ≦ m ≦ M Identify content. Here, M is the total number of contents used by one of these N users during the month.

  Next, the usage frequency f (i, m) of the content of the identifier m of the user of the identifier i is totaled, and the usage status of the user of the identifier i is represented by the vector F (i) = (f (i, 1),. , f (i, m)).

  The inter-user similarity totaling unit 207 executes a function of totaling a scale for measuring the similarity between each user from data obtained by totaling the usage status of each user. Specifically, the similarity between users assigned identifier i and identifier j is calculated using a vector F (i) obtained by counting the usage status of the user of identifier i from the number of times f (i, m) of each content is used. Inner product of F (i) and F (j), g (i, j) = f (i, 1) f (j, 1) + f (i, 2) f (j, 2) + ... + f ( i, M) F (j, M) is added (corresponding to step S30 in FIG. 2).

  The spectral decomposition unit 208 executes a function of performing spectral decomposition on an N-order square matrix G = (g (i, j); 1 ≦ i, j ≦ N) formed by similarity data and calculating an eigenvalue and an eigenvector set. (Corresponding to step S40 in FIG. 2).

  The cluster structure formation status analysis unit 209 has a cluster or sub-cluster viewed from the user scale regarding the formation of the user cluster group by sharing the content learned by the user from the aggregate data of the scale for measuring the similarity between the users. A function of counting the indicators as to whether a large number is formed or a small number is formed is performed (corresponding to step S50 in FIG. 2).

  Specifically, for each eigenvector v calculated by the spectrum decomposition unit 208, the constant composition component calculation unit 2091 calculates two constant code components vp and vm (corresponding to step S51 in FIG. 2), and further Then, the bias index calculation unit 2092 calculates the strength s (v) at which the eigenvector v is biased toward the fixed number composition for each eigenvector v (corresponding to step S52 in FIG. 2). Further, the plot data creation unit 2093 creates two-dimensional plot information as (λ (v), s (v)) together with the eigenvalue λ (v) of the eigenvector v calculated by the spectrum decomposition unit 208 (FIG. 2 corresponds to step S53).

  The cluster size distribution creation unit 210 relates to the formation of a user cluster group by sharing the content learned by the user from the aggregate data of the scale for measuring the similarity between the users, and the cluster size based on the number of users in the cluster or sub-cluster. The function of calculating the distribution of the above is executed (corresponding to step S60 in FIG. 2).

  Specifically, for each eigenvector v calculated by the spectrum decomposition unit 208, the value of the strength s (v) that is biased toward the composition of the fixed mark calculated by the cluster structure formation state analysis unit 209 is called, and the default unit The set eigenvector v of positive constant s0 which is smaller than 1 and close to 1 is determined to form a cluster or a sub-cluster.

  For the eigenvector v determined to form a cluster or sub-cluster, the magnitudes of the constant sign components | vp | and | vm | are determined. If | vp | ≧ | vm |, u = vp, | vp | In the case of <| vm |, u = vm and the number of elements of the set {1 ≦ i ≦ N | u (i) / | u |> e} is set to the small positive number e set in the default setting unit 205. Tally. The above totaling is executed for all or specified eigenvectors in the specified eigenvalue range, and the cluster size distribution is totalized, for example, as a cumulative frequency distribution. The total results are stored in the output storage unit 400.

  Upon receiving the request, the display control unit 202 displays a list of drawings / tables that can be displayed, receives the designation, and displays the list.

FIG. 3 is a diagram showing an output example analyzed by the system implementing the present invention.
3 (a) to 3 (c) show the usage record information in the Internet audience rating survey data for an actual Internet video distribution and video sharing website well known among Internet users in Japan. This is an analysis of the cluster formation status. Analysis for sample users.

  In FIG. 3, with respect to the eigenvalue λ of the top 50%, the circles correspond to the points of the rank (horizontal axis) of the eigenvalue λ and the deviation s (λ) (vertical axis) toward the fixed-word composition. FIG. 3A shows that the value of the bias s (λ) is generally small, and the number of clusters is smaller than the number of users. On the other hand, FIGS. 3B and 3C show that the value of the bias s (λ) is not small and the number of clusters is larger than the number of users.

  The point that the present invention can objectively analyze or evaluate the growth of the user scale and the maintenance of the scale will be described.

  FIG. 4 is a sample of usage records regarding the growth status of the site as seen from the number of times of use (traffic meaning of page view) of the site including the websites of FIGS. 3 (a), 3 (b), and 3 (c). Created from information.

  In FIG. 4, Site1 is the website of FIG. 3 (a), and the analysis result of the cluster formation status of Site2 is almost the same as FIG. 3 (a), so it is omitted in FIG. Further, Site3 is the website shown in FIG. 3B, and Site4 is the website shown in FIG.

  The results in FIG. 4 show that the success or failure of scale related to site usage corresponds to the number of clusters or sub-clusters formed. Therefore, it can be considered that the implementation of the present invention can be used for future analysis by objectively analyzing or evaluating user-scale growth and maintenance of scale.

  Therefore, the present invention distributes contents such as videos and music, or general information such as blogs / bulletin boards, etc., by decision makers such as corporations, public organizations, government agencies, etc., who make decisions on research management, development investment, policies, etc. Objectively analyze, evaluate, or predict the future of user-scale growth and scale maintenance for services that provide sharing, services that sell or exchange various products, or a combination of these This makes it possible to make decisions that will bring significant management, economic, and social benefits in the formulation of strategies such as improvement / improvement of current services, related research, development, investment, policies, etc. Is expensive.

  The processing and functions performed by each processing means (each unit and each unit) of the growth analysis analysis system shown in FIG. 2 are the CPU, memory, various registers, etc. built in the computer constituting the growth analysis evaluation system. This hardware resource is used to execute a program corresponding to a process executed by each processing means (each unit or each unit). Further, these programs can be widely distributed in the market via computer-readable recording media such as FD, CD-ROM, and DVD, and networks such as the Internet.

100: User interface (display, keyboard, mouse, etc.)
200: Growth analysis analysis system 201: User interface processing unit 202: Display control unit 203: Central control unit 204: Main memory 205: Default setting unit 206: Usage status totaling unit 207: Inter-user similarity totaling unit 208: Spectral decomposition Unit 209: cluster structure formation state analysis unit 2091: standard composition calculation unit 2092: bias index calculation unit 2093: plot data creation unit 210: cluster size distribution creation unit 300: service use activity data storage unit 400: output storage unit

Claims (10)

To evaluate whether services that provide various types of information continue to provide enough diversity to continue to grow or maintain the scale of growth through the continued use of general consumers Service scale growth analysis system,
Means for sequentially collecting each user's usage for a whole user or a subset of sampled users;
Means for taking the collected usage status as input and decomposing the user into a cluster group or sub-cluster group relating to the preference or interest of the various information;
As an index of the structural characteristics of the decomposed cluster group or sub-cluster group, there is provided means for calculating or illustrating the strength of the bias toward the large-scale cluster group or the strength of the bias toward the small to medium-sized cluster group. Service scale growth analysis system characterized by In the service scale growth analysis system according to claim 1,
An identifier i, 1 ≦ i ≦ N is assigned to each user belonging to the whole user or a subset of sampled users, an identifier m, 1 ≦ m ≦ M is assigned to the various information, and an identifier i is assigned. As the number of usages or purchases f (i, m) of the objects belonging to the various information to which the identifier m is assigned, data obtained by sequentially collecting the usage status of each user is input and the identifier i is assigned. G (i, j) = g (i, j) = f (i, 1) f (j, 1) + f (i, 2) f (j , 2) + ・ ・ ・ + f (i, M) f (j, M) and N-th order symmetric matrix G = (g (i, j); 1 ≦ i, j ≦ N) between users A service scale growth analysis system comprising means for decomposing a user into a cluster group relating to preference or interest in the various information as a preference or interest similarity matrix. In the service scale growth analysis system according to claim 2,
Means for calculating an eigenvector of an Nth-order symmetric matrix G that is a similarity matrix between users;
For each eigenvector v, vp (i) = Max {0, v (i)}, 1 ≦ i ≦ N and vm (i) = Max (0, -v (i)), 1 ≦ i ≦ N Means to calculate the fixed-word composition vp and vm of the fixed v,
For each eigenvector v, there is provided a means for calculating the strength s (v) at which the eigenvector v is biased toward the fixed composition vp or vm,
When the s (v) satisfies a certain reference value, it is determined that the dominant fixed-word composition of the eigenvector v corresponds to a cluster or sub-cluster, and the cluster occupies the whole eigenvector representing the correlation between users Alternatively, a service scale growth analysis system characterized by extracting a ratio of sub-clusters. In the service scale growth analysis system according to any one of claims 1 to 3,
Means for calculating a cluster size distribution in terms of the number of users in a cluster group or sub-cluster group,
By calculating an index indicating the bias of the cluster size distribution, calculating or illustrating the strength of the bias to the large-scale cluster group or the strength of the bias to the small to medium-scale cluster group, the user for the various information A service scale growth analysis system that evaluates whether or not it continuously responds to a variety of preferences or interests. In the service scale growth analysis system according to any one of claims 1 to 4,
The service scale growth analysis system characterized in that the various types of information are content as moving images or music, information written on a blog / bulletin board, information about various products, or a combination thereof.   A program for causing a computer to function as each means in the service scale growth analysis system according to any one of claims 1 to 5. Services that provide various types of information such as video or music content, information written on blogs or bulletin boards by users, information on various products, or a combination of these, are continuously used by general consumers A computer-based service scale growth analysis method for assessing whether or not it is continuously providing enough diversity to grow or maintain a scale of growth, comprising:
Collecting each user's usage sequentially for the entire user or a sampled subset of users;
Using the collected usage status as an input, and decomposing the user into a cluster group or sub-cluster group related to the preference or interest in the various information;
As an index of the structural features of the decomposed cluster group or sub-cluster group, a step of calculating or illustrating the strength of the bias toward the large-scale cluster group or the strength of the bias toward the small to medium-scale cluster group is provided. A service scale growth analysis method characterized by this. The service scale growth analysis method according to claim 7,
An identifier i, 1 ≦ i ≦ N is assigned to each user belonging to the whole user or a subset of sampled users, an identifier m, 1 ≦ m ≦ M is assigned to the various information, and an identifier i is assigned. As the number of usages or purchases f (i, m) of the objects belonging to the various information to which the identifier m is assigned, data obtained by sequentially collecting the usage status of each user is input and the identifier i is assigned. G (i, j) = g (i, j) = f (i, 1) f (j, 1) + f (i, 2) f (j , 2) + ・ ・ ・ + f (i, M) f (j, M) and N-th order symmetric matrix G = (g (i, j); 1 ≦ i, j ≦ N) between users A service scale growth analysis method comprising a step of decomposing a user into clusters related to preference or interest with respect to the various information as a preference or interest similarity matrix. The service scale growth analysis method according to claim 8,
Calculating an eigenvector of an Nth-order symmetric matrix G, which is an inter-user similarity matrix;
For each eigenvector v, vp (i) = Max {0, v (i)}, 1 ≦ i ≦ N and vm (i) = Max (0, -v (i)), 1 ≦ i ≦ N A step of calculating vp and vm for the fixed-word composition of fixed v;
For each eigenvector v, the step of calculating the strength s (v) at which the eigenvector v is biased toward the fixed-mark composition vp or vm comprises:
When the s (v) satisfies a certain reference value, it is determined that the dominant fixed-word composition of the eigenvector v corresponds to a cluster or sub-cluster, and the cluster occupies the whole eigenvector representing the correlation between users Alternatively, a service scale growth analysis method characterized by extracting a ratio of sub-clusters. The service scale growth analysis method according to any one of claims 7 to 9,
A step of calculating the distribution of the cluster size in the sense of the number of users of the cluster group or the sub-cluster group, and calculating the index indicating the bias of the cluster size distribution, thereby increasing or decreasing the bias to the large-scale cluster group, or A service scale characterized by evaluating whether or not the various types of information continuously respond to a variety of user preferences or interests by calculating or illustrating the strength of bias toward small to medium-sized clusters. Growth analysis method.

Images