Information processing device, information processing method, and information processing program

Abstract

This information processing device: acquires log data indicating an access history of a user with respect to an application for at least two days; calculates, on a daily basis, data on the number of days before-next-access indicating the number of days before-next-access until the user next accesses the application on the basis of the log data; calculates a histogram showing a relationship between the number of days before-next-access and the frequency of the number of days before-next-access on the basis of the data on the number of days before-next-access; estimates a threshold of the number of days before-next-access at which the user indicates a dropout sign to the application on the basis of the histogram, and performs user profiling on the basis of the threshold.

Description

Information processing device, information processing method, and information processing program

[0001] The present disclosure relates to user profiling techniques.

[0002] Patent Literature 1 discloses a state estimation device that estimates the state of an object at each time based on input observation values ​​of the object at each time. This state estimation device estimates the state at each time by applying a Gaussian sum sampling filter when noise in the observation values ​​follows a normal distribution, and by applying a Monte Carlo filter when noise in the observation values ​​does not follow a normal distribution.

[0003] However, the technology of Patent Document 1 does not take into account the number of days until the next access by the user to the application, and therefore the technology of Patent Document 1 cannot perform user profiling regarding signs of user abandonment from the application with high accuracy.

[0004] JP 2016-218674 A

[0005] The present disclosure has been made to solve such problems, and aims to provide a technology that can perform user profiling regarding signs of user withdrawal with high accuracy.

[0006] An information processing device in one aspect of the present disclosure includes an acquisition unit that acquires log data indicating a user's access history to an application for at least two days or more; a pre-processing unit that calculates, on a daily basis based on the log data, next access days data that indicates the number of days until the user next accesses the application; an estimation unit that calculates, based on the next access days data, a histogram that indicates the relationship between the number of days to next access and the frequency of the number of days to next access, and estimates, based on the histogram, a threshold value for the number of days to next access at which the user shows signs of abandoning the application; and a profiling unit that performs user profiling based on the threshold value.

[0007] According to the present disclosure, user profiling regarding signs of user withdrawal can be performed with high accuracy.

[0008] 1 is a block diagram showing an example of a configuration of an information processing system according to a first embodiment of the present disclosure; FIG. 2 is a diagram showing an example of data on number of days to next access; FIG. 3 is a histogram showing the relationship between the number of days to next access and the frequency of the number of days to next access for user A; FIG. 4 is a histogram showing the relationship between the number of days to next access and the frequency of the number of days to next access for user B; FIG. 5 is a flowchart showing an example of processing when an information processing device according to the first embodiment estimates a threshold; FIG. 6 is a flowchart showing an example of processing when an information processing device according to the first embodiment determines a withdrawal sign; FIG. 7 is a flowchart showing an example of processing when an information processing device according to the second embodiment estimates a threshold;

[0009] (Findings underlying the present disclosure) Applications with communication functions that allow users to communicate with each other are becoming widespread on mobile devices such as smartphones. Utilizing these applications with communication functions as communication tools for residents of a local community is expected to revitalize the community.

[0010] However, unlike game applications that stimulate the user's gambling instinct, applications that focus on communication are difficult to attract users strongly, and it is important to reduce the dropout rate. Therefore, it is necessary to estimate dropout signs. Here, the frequency of application access varies depending on the user.

[0011] Therefore, if a common criterion is applied to all users, it is not possible to accurately estimate the churn signs for an application for each of multiple users. For example, if a common criterion is set for all users based on users who frequently access an application, notifications to prevent churn will be sent frequently to users who access the application infrequently, causing annoyance to such users. Conversely, if a common criterion is set for all users based on users who access the application infrequently, there is a possibility that churn signs for users who frequently access the application will be overlooked.

[0012] Therefore, the inventors have come up with the idea of ​​the present disclosure based on the knowledge that by calculating next access days data indicating the number of days until the next time a user accesses an application for each of several days from the user's access history to the application, and statistically processing this next access days data, it is possible to accurately estimate signs of withdrawal for each of several users.

[0013] (1) An information processing device according to one aspect of the present disclosure includes an acquisition unit that acquires log data indicating a user's access history to an application for at least two days or more; a pre-processing unit that calculates, on a daily basis based on the log data, next access days data indicating the number of days until the user next accesses the application; an estimation unit that calculates, based on the next access days data, a histogram indicating the relationship between the number of days to next access and the frequency of the number of days to next access, and estimates, based on the histogram, a threshold value for the number of days to next access at which the user shows signs of abandoning the application; and a profiling unit that performs user profiling based on the threshold value.

[0014] According to this configuration, next access days data indicating the number of days until the next access until the user's next access to the application is calculated for each day based on at least two days or more of log data indicating the user's access history to the application. A histogram indicating the relationship between the number of days to next access and the frequency of the number of days to next access is calculated based on this next access days data. A threshold value for the number of days to next access at which the user shows signs of abandoning the application is estimated based on this histogram. User profiling is performed based on this threshold value. As a result, a threshold value appropriate for the user is estimated from the user's access history, and highly accurate user profiling of the user's signs of abandonment is realized.

[0015] (2) In the information processing device described in (1) above, the estimation unit may perform the following: fitting the histogram with a probability distribution that can be regarded as a normal distribution according to the central limit theorem; and estimating, as the threshold value, the number of days to next access when the percentile in the fitted probability distribution becomes a predetermined value.

[0016] According to this configuration, the histogram is fitted with a probability distribution that can be regarded as a normal distribution according to the central limit theorem, and the number of days until the next access at which the percentile in the fitted probability distribution reaches a predetermined value is estimated as the threshold. This provides a statistically supported threshold for the number of days until the next access, enabling highly accurate user profiling.

[0017] (3) In the information processing device described in (2) above, the probability distribution may be a geometric distribution.

[0018] The number of days to next access for each of multiple days consists of two trials: access or no access. The results of each trial are independent of each other, so it can be considered to follow Bernoulli trials. When Bernoulli trials are repeated, the number of trials until the first success follows a geometric distribution. Since the number of days to next access can be considered the number of trials until the first success, the histogram showing the relationship between the number of days to next access and the frequency of the number of days to next access follows a geometric distribution. In this configuration, the histogram showing the relationship between the number of days to next access and the frequency of the number of days to next access is fitted using a geometric distribution. Therefore, the threshold value for the number of days to next access, which indicates signs of churn, can be accurately calculated using limited log data.

[0019] (4) In the information processing device described in (3) above, the fitting may include calculating the average value of the number of days to next access from the histogram, calculating the probability of the geometric distribution from the average value of the number of days to next access, and calculating the expected value and variance of the geometric distribution based on the probability.

[0020] With this configuration, the average number of days to next access is calculated from a histogram showing the relationship between the number of days to next access and the frequency of the number of days to next access, the probability of a geometric distribution is calculated from this average number of days to next access, and the expected value and variance of the geometric distribution are calculated based on the calculated probability. This provides parameters that define the shape of the geometric distribution, allowing accurate fitting of the histogram using the geometric distribution.

[0021] (5) In the information processing device described in (1) above, the estimation unit may identify the number of days to next access at which a percentile in the histogram is equal to or greater than a predetermined value, and estimate the identified number of days to next access as the threshold value.

[0022] According to this configuration, the threshold value for the number of days until the next access is calculated by directly using the histogram, so that the threshold value can be calculated with a small number of processing steps.

[0023] (6) The information processing device according to any one of (1) to (5) above may further include an output unit that outputs the results of the user profiling.

[0024] According to this configuration, the results of user profiling can be presented to the user, the administrator of the application, and the like.

[0025] (7) In the information processing device described in any of (1) to (6) above, the profiling unit may exclude days on which the user did not access the application from the next access days data, and if the latest number of days to next access indicated by the excluded next access days data is greater than the threshold value, determine that the user is showing signs of withdrawal.

[0026] According to this configuration, if the number of days until the next access is greater than the threshold value, it is determined that the user is showing signs of abandonment, so that it is possible to accurately determine whether or not the user is showing signs of abandonment.

[0027] (8) An information processing method in another aspect of the present disclosure is an information processing method in a computer, comprising: acquiring log data indicating a user's access history to an application for at least two days or more; calculating, on a daily basis based on the log data, next access days data indicating the number of days until the next access until the user next accesses the application; calculating, based on the next access days data, a histogram indicating the relationship between the number of days until next access and the frequency of the number of days until next access; estimating, based on the histogram, a threshold value for the number of days until next access at which the user shows signs of abandoning the application; and performing user profiling based on the threshold value.

[0028] According to this configuration, an information processing method is provided that can realize highly accurate user profiling regarding the user's withdrawal symptoms.

[0029] (9) In another aspect of the present disclosure, an information processing program causes a computer to perform the following operations: acquire log data indicating a user's access history to an application for at least two days or more; calculate, for each day based on the log data, next access days data indicating the number of days until the next access until the user next accesses the application; calculate, based on the next access days data, a histogram indicating the relationship between the number of days until next access and the frequency of the number of days until next access; estimate, based on the histogram, a threshold value for the number of days until next access at which the user shows signs of abandoning the application; and perform user profiling based on the threshold value.

[0030] According to this configuration, an information processing program is provided that can realize highly accurate user profiling regarding the user's withdrawal symptoms.

[0031] The present disclosure can also be realized as an information processing system operated by such an information processing program. Needless to say, such a computer program can be distributed on a non-transitory computer-readable recording medium such as a CD-ROM or via a communication network such as the Internet.

[0032] Note that each of the embodiments described below represents a specific example of the present disclosure. The numerical values, shapes, components, steps, and step orders shown in the following embodiments are merely examples and are not intended to limit the present disclosure. Furthermore, among the components in the following embodiments, components that are not described in the independent claims that represent the highest concept are described as optional components. Furthermore, in all of the embodiments, the respective contents can be combined.

[0033] 1 is a block diagram showing an example of a configuration of an information processing system 100 according to a first embodiment of the present disclosure. The information processing system 100 includes an information processing device 1 and a terminal 4.

[0034] The terminal 4 is a device that acquires information about a user who is the subject of information processing. The terminal 4 is, for example, a smartphone, a personal computer, a wearable terminal, a physical measurement device, etc. The wearable terminal is, for example, smart glasses, a smart watch, etc. The physical measurement device is, for example, a pedometer. Note that the form of the terminal 4 is not limited to these examples. There may be one or more terminals 4. The terminal 4 inputs the acquired log data to the information processing device 1 at a predetermined timing.

[0035] The log data indicates a history of user accesses to an application. For example, when the terminal 4 detects a user's input of an instruction to start an application, the terminal 4 may generate log data and transmit the log data to the information processing device 1.

[0036] The log data includes, for example, a timestamp indicating the date and time when the user started the application, and the user ID of the user who started the application.

[0037] The application is a specific application that is the target of monitoring for user withdrawal signs. Hereinafter, this application will be referred to as the "target application." The target application may be, for example, an application for revitalizing a local community. The target application accepts posts from users about recommended visiting spots and posts the visiting spots on a bulletin board that can be viewed by all users. The target application accepts comments from other users about the visiting spots and posts the comments on the bulletin board in association with the visiting spots. This allows the target application to motivate users to visit the visiting spots and take a walk, thereby promoting the user's health. Furthermore, the user can get to know other users through the comments and build a personal network among local residents. Furthermore, the target application may have a step count recording function that records the user's step count to promote the user's health.

[0038] The information processing device 1 is a device for performing user profiling using log data, and is configured with a computer including a processor 2, a memory 3, and an interface circuit (not shown).

[0039] The processor 2 is hardware for performing user profiling of a user based on a user log, and is, for example, a central processing unit. The processor 2 includes an acquisition unit 21, a preprocessing unit 22, an estimation unit 23, a profiling unit 24, and an output unit 25. The acquisition unit 21 to the output unit 25 may be realized by the central processing unit executing an information processing program, or may be configured by dedicated hardware circuits such as an ASIC. The acquisition unit 21 to the output unit 25 may be distributed among multiple computers. Some of the acquisition unit 21 to the output unit 25 may be included in the terminal 4.

[0040] The memory 3 is configured by a rewritable nonvolatile storage device such as a solid state drive (SSD) or a hard disk drive (HDD). The memory 3 includes a log storage unit 31. The log storage unit 31 stores the log data acquired by the acquisition unit 21 from the terminal 4.

[0041] The acquisition unit 21 acquires log data for at least two days from the terminal 4 and stores the acquired log data in the log storage unit 31. The log storage unit 31 stores the log data input from the acquisition unit 21. The log storage unit 31 stores the next access date data calculated by the preprocessing unit 22. The log storage unit 31 stores the threshold value T calculated by the estimation unit 23 for each of the multiple users.

[0042] The preprocessing unit 22 calculates next access number of days data indicating the number of days until the user next accesses the target app for each of a plurality of days based on at least two days of log data stored in the log storage unit 31. The calculated next access number of days data is stored in the log storage unit 31.

[0043] The preprocessing unit 22 performs a process of calculating the next access date data for each user. The preprocessing unit 22 uses log data for at least two days to determine whether the user has accessed the target app on each day of the usage period from the user's start date of use of the target app to the present. For example, the preprocessing unit 22 calculates the next access date data for each day of the usage period from the user's start date of use of the target app to the present. ｉ If the log data indicates that the user has launched the target app at least N times, ｉ The preprocessing unit 22 determines that the user accessed the target application on, for example, day d. ｉ If the log data indicates that the target application has been launched less than N times, ｉ It is determined that the user has not accessed the target application on day d. N is, for example, 1. When N=1, the preprocessing unit 22 ｉ If there is no log data for date d ｉ It is sufficient to determine that the user has not accessed the target app in the specified time period. The preprocessing unit 22 calculates the next access number data for each day based on the result of the access determination for each of at least two days. The next access number data includes the user ID and the number of days until the next access for each day.

[0044] FIG. 2 is a diagram showing an example of next access days data. The first line of FIG. 2 shows the next access days data for user A, and the second line of FIG. 2 shows the next access days data for user B. In FIG. 2, the white blocks indicate the number of days until next access on days with access. The gray blocks indicate the number of days until next access on days without access. The numbers below the white blocks indicate the number of days until next access. Since the gray blocks are the number of days until next access corresponding to days without access, the number of days until next access is not set.

[0045] For example, the number of days for next access for user A on the far left is set to "1" because user A accessed the target app the following day. For example, the number of days for next access for user A, second from the left, is set to "2" because user A accessed the target app two days later. The number of days for next access for user A, third from the left, is shown as a gray block because user A did not access the target app.

[0046] The estimation unit 23 calculates a histogram showing the relationship between the number of days to next access and the frequency of the number of days to next access, based on the data on the number of days to next access calculated by the preprocessing unit 22. Based on the calculated histogram, the estimation unit 23 estimates a threshold T for the number of days to next access at which a user shows signs of abandoning the application. The estimation unit 23 calculates the threshold T for each user.

[0047] 3 is a histogram showing the relationship between the number of days to next access and the frequency of the number of days to next access for user A. In FIG. 3, the vertical axis represents frequency and the horizontal axis represents the number of days to next access. The estimation unit 23 calculates the histogram for user A by using the number of days to next access data for user A to tally the frequency of the number of days to next access for each of multiple number of days to next access.

[0048] FIG. 4 is a histogram showing the relationship between the number of days until next access and the frequency of the number of days until next access for user B. User B is a frequent user who accesses the target app almost every day. Therefore, in the histogram of FIG. 4, the most frequent value for the number of days until next access is 1 day. Therefore, if user B's number of days until next access becomes, for example, 2 days or more, there is a possibility that user B will abandon the target app. "Abandoning the target app" means that the user no longer uses the target app.

[0049] On the other hand, User A uses the target app less frequently than User B. Furthermore, User A does not access the target app regularly. Therefore, in the histogram of FIG. 3 , the most frequent value for the number of days to next access is three days. Furthermore, in the histogram of FIG. 3 , the number of days to next access is distributed so that the frequency gradually decreases as it moves away from the most frequent value of three days to the left or right. Therefore, User A may abandon the target app if, for example, the number of days to next access is six days or more. Thus, the threshold T for the number of days to next access, which indicates a sign of abandonment of the target app, varies depending on the user's access frequency to the target app. Therefore, the present disclosure estimates a threshold T appropriate for each user, taking into account the characteristics of each user with respect to accessing the target app.

[0050] The estimation unit 23 fits the histogram with a probability distribution that can be regarded as a normal distribution according to the central limit theorem. The estimation unit 23 estimates the number of days to next access when the percentile in the fitted probability distribution reaches a predetermined value as a threshold.

[0051] The estimation unit 23 uses a geometric distribution as the probability distribution. The time-series data of the number of days to next access consists of two trials, one where access is attempted and the other where no access is attempted, and the results of each trial are independent of each other, so it follows a Bernoulli trial. When Bernoulli trials are repeated, the number of trials until the first success follows a geometric distribution. Since the number of days to next access is considered to be the number of trials until the first success, a histogram showing the number of days to next access and its frequency follows a geometric distribution. Therefore, in the first embodiment, the histogram is fitted with a geometric distribution.

[0052] The probability that the conditions for determining that a user has accessed a target app are met is defined as p (geometric distribution probability). In this case, based on the definition of geometric distribution, the histogram follows a geometric distribution whose expected value E is expressed by the following formula (1) and whose variance V is expressed by the following formula (2). Assume that the mean value μ of the histogram matches the expected value E. In this case, the probability p is expressed as p = 1 / μ.

[0053] E=1 / p (1) V=(1-p) / p ２ (2) Therefore, the estimation unit 23 performs the following process. First, the estimation unit 23 calculates the average value μ of the number of days to next access from the histogram. Next, the estimation unit 23 calculates the probability p from the reciprocal of the average value μ. Next, the estimation unit 23 calculates the expected value E (= average value μ) and variance V of the geometric distribution based on the probability p. The expected value E and variance V are parameters that determine the shape of the geometric distribution. As a result, a geometric distribution that fits the histogram is obtained.

[0054] When the histogram follows a geometric distribution, the probability of the number of days to next access occurring being n days or more can be calculated. In a standard normal distribution, the probability of an occurrence greater than the mean value plus the standard deviation is 15.9%. Therefore, the estimation unit 23 estimates, as the threshold T, the integer value closest to the number of days to next access where the occurrence probability is 15.9% or more in the fitted geometric distribution, i.e., the number of days to next access where the percentile is 84.1% or more.

[0055] Alternatively, the estimation unit 23 may estimate, as the threshold T, the integer value closest to the number of days to next access at which the occurrence probability is equal to or greater than the mean value + 2 × standard deviation in the fitted geometric distribution, i.e., the number of days to next access at which the percentile is equal to or greater than 97.7%. The above 84.1% and 97.7% are common reference values ​​in statistics, but other values ​​may be used. The estimation unit 23 may estimate, as the threshold T, the integer value closest to the number of days to next access at which the occurrence probability is equal to or greater than the mean value + k × standard deviation in the fitted geometric distribution, i.e., the number of days to next access at which the percentile is equal to or greater than a predetermined value, where k is a real number greater than or equal to 0. The estimation unit 23 stores the estimated threshold T in the log storage unit 31.

[0056] The profiling unit 24 performs user profiling based on the threshold value T. The profiling unit 24 acquires next access days data from the log storage unit 31 and excludes days on which the user has not accessed the target app from the acquired next access days data. The profiling unit 24 acquires the most recent number of days to next access from the next access days data from which days on which no access has occurred have been excluded. The profiling unit 24 determines that the user is exhibiting a dropout symptom if the most recent number of days to next access is greater than the threshold value T. On the other hand, the profiling unit 24 determines that the user is not exhibiting a dropout symptom if the most recent number of days to next access indicated by the next access days data is equal to or less than the threshold value T. However, this is just one example, and the profiling unit 24 may also determine that the user is exhibiting a dropout symptom if, in the next access days data from which days on which no access has occurred have been excluded, the average number of days to next access from the most recent number of days to next access from a predetermined number of days before the next access is greater than the threshold value T. The predetermined number of days may be any appropriate value, such as 3 days or 7 days.

[0057] The output unit 25 outputs the results of user profiling performed by the profiling unit 24. When the profiling unit 24 obtains a user profiling result indicating that a user has signs of withdrawal, the output unit 25 may transmit the user profiling result to the administrator's terminal using a communication interface. The administrator may be, for example, a department of a local government that promotes the health of local residents. After viewing the user profiling result, the administrator notifies the user who has signs of withdrawal with a message or the like to avoid withdrawal. This prevents the user who has signs of withdrawal from the target app. Alternatively, when the profiling unit 24 obtains a user profiling result indicating that a user has signs of withdrawal, the output unit 25 may transmit the user profiling result to the terminal 4 of the corresponding user using a communication interface.

[0058] 5 is a flowchart showing an example of processing performed by the information processing device 1 according to the first embodiment to estimate the threshold value T. The following processing is performed for each user. In step S1, the preprocessing unit 22 acquires log data of the user during the period in which the target app was used from the log storage unit 31, calculates the number of days until next access for each day during the period from the acquired log data, and calculates number-of-days-until-next-access data indicating the number of days until next access for each day.

[0059] Next, in step S2, the estimation unit 23 calculates a histogram showing the relationship between the number of days to next access and the frequency of the number of days to next access from the data of the number of days to next access calculated in step S1.

[0060] Next, in step S3, the estimation unit 23 calculates the mean value μ of the histogram and calculates the probability p of the geometric distribution by taking the reciprocal of the mean value μ.

[0061] Next, in step S4, the estimation unit 23 calculates the expected value E and variance V of the geometric distribution from the probability p using the above-mentioned formulas (1) and (2), and fits the histogram with the geometric distribution.

[0062] Next, in step S5, the estimation unit 23 calculates, as a threshold T, an integer value closest to the number of days until next access at which the percentile is equal to or greater than a predetermined value (for example, 84.1%) in the fitted geometric distribution.

[0063] In this way, the threshold value T for each user is obtained.

[0064] 6 is a flowchart showing an example of processing performed by the information processing device 1 in embodiment 1 when determining a withdrawal sign. This processing is performed for each user.

[0065] In step S11, the profiling unit 24 acquires the next access date data from the log storage unit 31, excludes days with no access from the acquired next access date data, and acquires the latest next access date from the next access date data excluding the days with no access. ０ ) to 2 days ago (d ２) are days with no access. Therefore, the profiling unit 24 ３ ) is acquired as the latest access date. In the example of user B in FIG. ０ ) is a day with no access. Therefore, the profiling unit 24 １ ) is obtained as the number of days since the last access.

[0066] Next, in step S12, the profiling unit 24 determines whether the most recent number of days since next access is greater than a threshold value T or not.

[0067] If the number of days since the most recent next access is greater than the threshold T (YES in step S12), the profiling unit 24 proceeds to step S13. On the other hand, if the number of days since the most recent next access is equal to or less than the threshold T (NO in step S12), the profiling unit 24 ends the process.

[0068] In step S13, the profiling unit 24 outputs the results of the user profiling, which indicate that there are signs of withdrawal, to the administrator's terminal using the communication interface.

[0069] If the result of step S12 is NO, the profiling unit 24 may output the result of user profiling indicating that there are no signs of withdrawal to the terminal of the administrator using the communication interface. In this way, users who have signs of withdrawal are identified.

[0070] As described above, according to the information processing device 1 of the first embodiment, next access days data indicating the number of days until the next access until the user next accesses the application is calculated for each day based on at least two days' worth of log data indicating the user's access history to the application. A histogram indicating the relationship between the number of days to next access and the frequency of the number of days to next access is calculated based on this next access days data. Based on this histogram, a threshold T for the number of days to next access at which the user shows signs of dropping out of the application is estimated. User profiling is performed based on this threshold T. This allows a threshold T appropriate for the user to be obtained, thereby realizing highly accurate user profiling of the user's signs of dropping out.

[0071] (Embodiment 2) In embodiment 2, the threshold T is estimated directly from the histogram without fitting the histogram with a geometric distribution. In embodiment 2, the same components as those in embodiment 1 are denoted by the same reference numerals, and their description will be omitted.

[0072] 7 is a flowchart showing an example of processing when the information processing device 1 according to the second embodiment estimates the threshold value T. The processing in steps S21 and S22 is the same as steps S1 and S2 in FIG.

[0073] In step S23, the estimation unit 23 identifies the number of days to next access when the percentile in the histogram calculated in step S22 is equal to or greater than a predetermined value, and estimates the identified number of days to next access as the threshold value T. As the predetermined value, for example, a percentile value corresponding to the mean value + k × standard deviation can be used. For example, the predetermined value corresponding to the mean value + standard deviation is 84.1%, and the predetermined value corresponding to the mean value + 2 × standard deviation is 97.7%.

[0074] The process of the estimation unit 23 will be described in detail below. The estimation unit 23 normalizes the histogram so that the total area of ​​the histogram becomes 1. Hereinafter, the normalized histogram will be referred to as a normalized histogram. Here, in the normalized histogram, an arbitrary number of days to next access is expressed as d ｊj is an index indicating the number of days to the next access, and takes the value of j=1, 2, 3, ..., J. J is the number of days to the next access in the normalized histogram d ｊ The estimation unit 23 estimates the next access days d ｊ Each time j is increased to 1, 2, ..., the number of days until the next access d ｊ Percentile P ｊ The estimation unit 23 calculates the percentile P ｊ Each time you calculate percentile P ｊ The estimation unit 23 determines whether the percentile P ｊ The number of days d for the next access when exceeds a predetermined value ｘ The estimation unit 23 determines the number of days to next access d ｘ＋１ is estimated as a threshold T.

[0075] As described above, according to the information processing device 1 of the second embodiment, the histogram is directly used to calculate the threshold value for the number of days until the next access, so that the threshold value T can be calculated with a small number of processing steps.

[0076] (Modification) In the first embodiment, the histogram is fitted using a geometric distribution. However, it may be fitted using a probability distribution other than the geometric distribution. When there is a sufficient number of data (e.g., 400), the histogram can be fitted using a probability distribution that can be regarded as a normal distribution according to the central limit theorem. In addition to the geometric distribution, examples of probability distributions that can be regarded as normal distributions according to the central limit theorem include the Poisson distribution and the binomial distribution. These probability distributions have a single peak and a shape that asymptotically approaches 0 as the random variable approaches positive infinity and negative infinity. These probability distributions approach a normal distribution as the number of data increases. Therefore, the present disclosure allows the Poisson distribution and the binomial distribution to be adopted instead of the geometric distribution.

[0077] (Application Examples) Application examples of the present disclosure will be described below.

[0078] Many applications are installed on smartphones or personal computers. However, users do not continue to use all of the installed applications, and stop using the applications midway, i.e., dropout occurs. Therefore, if it is possible to predict users who are likely to dropout in order to encourage continued use of the applications, notifications encouraging use can be provided to those users, thereby preventing dropout. As an example of an application, a health promotion application (hereinafter referred to as a "health promotion application") that records the number of steps taken to promote health can be adopted. Below, an example of a specific process will be described in order with reference to the configurations of Embodiments 1 and 2.

[0079] A resident (user) in a certain community uses a smartphone as terminal 4. A health promotion app is installed on terminal 4. Acquisition unit 21 acquires log data indicating whether or not the user has launched the health promotion app. Log storage unit 31 can be, for example, a database built on the cloud. Acquisition unit 21 saves the log data acquired from terminal 4.

[0080] The batch processing for estimating withdrawal signs is executed, for example, at 11:00 PM every day. The preprocessing unit 22 acquires log data from the log storage unit 31 and calculates next access days data indicating the number of days until next access for each day from the acquired log data. The estimation unit 23 calculates a histogram showing the relationship between the number of days until next access and its frequency from the next access days data. The estimation unit 23 calculates a threshold T based on this histogram. The profiling unit 24 determines whether or not the user has withdrawal signs by comparing the threshold T calculated by the estimation unit 23 with the most recent number of days until next access. If the profiling unit 24 determines that the user has withdrawal signs, the output unit 25 notifies the administrator of an organization aiming to promote health of the user profiling results indicating the presence of withdrawal signs. The administrator is, for example, a department in a local government that promotes health promotion for local residents in the local government that has adopted this health promotion app. The administrator calls residents who have been determined to have withdrawal signs to encourage them to continue using the app.

[0081] When a health promotion app is an application that a local government requests residents to install, residents are not highly motivated to actively use the health promotion app. This disclosure is important for estimating user dropout signs for such applications that potentially have low user motivation.

[0082] The present disclosure is useful in performing user profiling.

Claims

1. An information processing device comprising: an acquisition unit that acquires log data indicating a user's access history to an application for at least two days; a pre-processing unit that calculates, on a daily basis based on the log data, next access days data that indicates the number of days until the user next accesses the application; an estimation unit that calculates, based on the next access days data, a histogram that indicates the relationship between the number of days to next access and the frequency of the number of days to next access, and estimates, based on the histogram, a threshold value for the number of days to next access at which the user shows signs of abandoning the application; and a profiling unit that performs user profiling based on the threshold value.

2. The information processing device of claim 1, wherein the estimation unit performs the following: fitting the histogram with a probability distribution that can be considered a normal distribution according to the central limit theorem; and estimating, as the threshold value, the number of days to next access when the percentile in the fitted probability distribution becomes a predetermined value.

3. The information processing device according to claim 2, wherein the probability distribution is a geometric distribution.

4. An information processing device as described in claim 3, wherein the fitting includes: calculating the average value of the number of days to next access from the histogram; calculating the probability of the geometric distribution from the average value of the number of days to next access; and calculating the expected value and variance of the geometric distribution based on the probability.

5. The information processing device according to claim 1, wherein the estimation unit identifies the number of days to next access at which a percentile in the histogram is equal to or greater than a predetermined value, and estimates the identified number of days to next access as the threshold value.

6. The information processing device according to claim 1 or 2, further comprising an output unit that outputs the results of the user profiling.

7. An information processing device as described in claim 1 or 2, wherein the profiling unit excludes days on which the user did not access the application from the next access days data, and if the latest number of days to next access indicated by the excluded next access days data is greater than the threshold value, determines that the user is showing signs of withdrawal.

8. An information processing method in a computer, comprising: acquiring log data indicating a user's access history to an application for at least two days; calculating, on a daily basis based on the log data, next access days data indicating the number of days until the user next accesses the application; calculating, based on the next access days data, a histogram indicating the relationship between the number of days to next access and the frequency of the number of days to next access, and estimating, based on the histogram, a threshold value for the number of days to next access at which the user shows signs of abandoning the application; and performing user profiling based on the threshold value.

9. An information processing program that causes a computer to perform the following operations: acquire log data indicating a user's access history to an application for at least two days; calculate, on a daily basis based on the log data, next access days data indicating the number of days until the next access until the user next accesses the application; calculate, based on the next access days data, a histogram indicating the relationship between the number of days to next access and the frequency of the number of days to next access, and, based on the histogram, estimate a threshold value for the number of days to next access at which the user shows signs of abandoning the application; and perform user profiling based on the threshold value.

Images