JP2015001927A - Function recommendation device and function recommendation method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To recommend an appropriate function on the basis of execution histories of functions.SOLUTION: A function recommendation device 1 includes: a function execution accepting unit 12 which accepts function identification information identifying an executed function; an environmental information acquisition unit 14 which acquires environmental information including at least one piece of time information and position information of the function recommendation device 1 at the time of acceptance; an execution history storage unit 10 in which execution history information relating the function identification information and the environmental information to each other is stored; a function recommendation model 15 in which function priority information relating the function identification information and priorities of recommendation of functions identified by the function identification information to each other per prescribed unit of environmental information is stored; a spreading information generation unit 16 which generates spreading information relating to a method of updating the function priority information on the basis of the execution history information, on the basis of the execution history information; and a function recommendation model update unit 18 which updates the function priority information on the basis of the execution history information and the spreading information.

Description

  The present invention eliminates the user's trouble of selecting a target function by recommending a function based on the user's operation history in accordance with environmental information such as date and time in the mobile terminal device, The present invention relates to technology for supporting the operation of mobile terminals.

  2. Description of the Related Art Conventionally, there has been known a device that presents a function to be provided on a screen as an icon associated with the function, and the function associated with the selected icon is executed by the user selecting the presented icon. It has been. For example, in order to select a recommended function to be presented, there is a technique for determining a recommended function to be presented to a user based on the date and position based on the function execution history of the user. As a result, it is possible to save time and effort for selecting a target function of the user and subsequent operation (see Patent Document 1 below).

  For function recommendation, a function recommendation model based on the function execution history is used. A value obtained by accumulating function execution history in a multi-dimensional space with the date and position as an axis is used as a function priority, and at the time of function recommendation, a function is recommended by comparing the function priority of each function for each date and position.

  In Patent Document 1, the resolution of environmental information such as date and position accumulated in one execution history is set in advance. If the set resolution is large, the range of date and position where the execution history is reflected in the function recommendation model will be larger, so the recommendation can be fluctuated, even if the function is different from the date and position where the function was executed It is possible to recommend. The accumulation of execution histories for different dates and positions and utilization for function recommendation is hereinafter referred to as history spillover.

JP 2011-160145 A

  However, in Patent Document 1, since the function execution history spillover method is static, the prediction accuracy may decrease due to an increase in the number of execution histories. For example, when the number of execution histories is small and the history spillover range is set to be narrow, it is not easy to recommend a function in the situation of date and position where there is no execution history. If the history spillover range is set large in order to cope with this, when the history is small, the history can be propagated at the date and position where there is no execution history, so it is possible to recommend functions.

  However, as the number of execution histories increases, if the ripple method when the history is small is applied as it is, the ripple range is too large, so functions executed in environment information such as other dates and positions more than necessary Is recommended, and it is not easy to make an appropriate function recommendation.

  Therefore, the present invention has been made in view of such a problem, and an object thereof is to provide a function estimation device and a function recommendation method capable of recommending an appropriate function based on a function execution history. .

  In order to solve the above problems, the function recommendation device of the present invention is a function recommendation device that recommends one or more specific functions from among a plurality of functions, and accepts function identification information that identifies the executed functions. Function execution accepting means, environment information obtaining means for obtaining environment information including at least one of time information when function identification information is accepted by the function execution accepting means and position information of the function recommendation device, and function execution Function execution history storage means for storing execution history information in which the function identification information received by the reception means is associated with the environment information acquired by the environment information acquisition means, the function identification information, and the function identification information are identified A function recommendation model in which function priority information associated with a predetermined unit of environment information is stored, and a function execution acceptance When the function identification information is accepted by the stage, the spread information generating means for generating the spread information on the method of updating the function priority information based on the execution history information based on the execution history information stored by the function execution history storage means And function recommendation model update means for updating function priority information stored in the function recommendation model based on the execution history information stored by the function execution history storage means and the spread information generated by the spread information generation means And a recommended function determining unit that determines a function to be recommended based on the function priority information stored in the function recommendation model, and a recommended function output unit that outputs information on the recommended function determined by the recommended function determining unit. .

  According to such a function recommendation device, the propagation information related to the method of updating the function priority information based on the execution history information is generated based on the execution history information, and based on the execution history information and the generated propagation information The function priority information is updated. That is, a method for dynamically updating the function priority information is determined based on the execution history information, and the function priority information is updated based on the method. Then, based on the updated function priority information, a recommended function is output. Therefore, it is possible to recommend an appropriate function based on the function execution history.

  In the function recommendation device of the present invention, it is preferable that the spread information generating means generates the spread information based on the cumulative number of execution history information stored by the function execution history storage means.

  In the function recommendation device of the present invention, it is preferable that the spread information generating means generates the spread information based on the distribution information of the execution history information stored by the function execution history storage means.

  In the function recommendation device of the present invention, the spread information generating means may generate the spread information for each environment information based on the distributed information for each environment information of the execution history information stored by the function execution history storage means. preferable.

  Further, in the function recommendation device of the present invention, the spread information generating means is based on the function identification information and the distributed information for each environment information of the execution history information stored by the function execution history storage means for each environment information. It is preferable to generate the spread information.

  In the function recommendation device of the present invention, it is preferable that the propagation information generation unit generates the propagation information further based on the execution history information stored by the function execution history storage unit of another function recommendation device.

  Further, in the function recommendation device of the present invention, the propagation information generation means is based on the propagation information generated based on the cumulative number of execution history information stored by the function execution history storage means and the distribution information of the execution history information. It is preferable to generate the propagation information by weighting and combining the generated propagation information.

  In the function recommendation device of the present invention, the spread information generating means generates the spread information based on the execution history information of the environment information within a predetermined range among the execution history information stored by the function execution history storage means. Is preferred.

  By the way, the present invention can be described as the invention of the apparatus as described above, and also as the invention of the method as follows. This is substantially the same invention only in different categories, and has the same operations and effects.

  That is, the function recommendation method according to the present invention is a function recommendation device that recommends one or more specific functions from among a plurality of functions, and the function identification information for identifying the functions and the function identification information to be identified The function recommendation model includes a function recommendation model in which function priority information associated with each predetermined unit of environment information including at least one of time information and position information of the function recommendation device is stored. A function recommendation method executed by a function recommendation device, a function execution receiving step for receiving function identification information of an executed function, and an environment for acquiring environment information when the function identification information is received in the function execution receiving step The information acquisition step, the function identification information received in the function execution reception step, and the environment information acquired in the environment information acquisition step are associated with each other. Function execution history storage step for storing the executed execution history information, and when the function identification information is received in the function execution receiving step, the function execution history is transmitted as information on how to update the function priority information based on the execution history information. Based on the propagation information generation step generated based on the execution history information stored in the storage step, the execution history information stored in the function execution history storage step, and the propagation information generated in the propagation information generation step, the function A function recommendation model update step for updating function priority information stored in the recommendation model, a recommended function determination step for determining a function to be recommended based on the function priority information stored in the function recommendation model, and a recommended function determination A recommended function output that outputs information about the recommended function determined in the step Tsu including and up, the.

  According to the present invention, it is possible to recommend an appropriate function based on the function execution history.

It is a functional block diagram of a function recommendation device concerning an embodiment of the present invention. It is a figure which shows the example of a screen at the time of the function recommendation in the function recommendation apparatus which concerns on embodiment of this invention. It is a figure which shows the hardware constitutions in the function recommendation apparatus which concerns on embodiment of this invention. It is a figure showing the example of a function ID table. It is a figure which shows the example of a table of execution log information. It is a figure which shows Example 1 which calculates | requires the spreading parameter which shows a spreading range. It is a figure which shows Example 2 which calculates | requires the spreading parameter which shows a spreading range. It is a figure which shows the example which calculates | requires the ripple parameter which shows a ripple distribution function. It is an image figure of the spreading information in the function priority map when position information is represented by latitude and longitude. It is a figure which shows the example of a graph which represented the distance between positions in case positional information is represented by semantic data with a pseudo distance. It is a figure which shows the example which calculates a distance by mapping to the position two-dimensional space in case position information is represented by semantic data. It is a figure which shows the example which produces | generates a propagation parameter using the dispersion | distribution information of execution history information. It is a figure which shows the example which updates a function recommendation model using a propagation parameter. It is a flowchart which shows the process (function recommendation method) performed with the function recommendation apparatus which concerns on embodiment of this invention. It is a figure which shows the effectiveness of dynamic propagation method determination. It is a functional block diagram of the function recommendation apparatus which concerns on the modification 1 of 1st Embodiment of this invention. It is a functional block diagram of the function recommendation apparatus which concerns on the modification 2 of 1st Embodiment of this invention.

  A function recommendation device and a function recommendation method according to an embodiment of the present invention will be described with reference to the drawings. In the description of the drawings, the same elements are denoted by the same reference numerals, and redundant description is omitted.

[First Embodiment]
First, the configuration of the function recommendation device 1 according to the present embodiment will be described. FIG. 1 is a configuration diagram of a function recommendation device 1 according to the present embodiment.

  The function recommendation device 1 is a mobile terminal such as a smartphone. The function recommendation device 1 is not limited to a portable terminal, and may be a PC (Personal Computer) or the like. The function recommendation device 1 recommends one or more specific functions from among a plurality of functions included in the device. Specifically, the function recommendation device 1 recommends a function by presenting several functions to the user, and the user can receive a service based on the function by executing the function. . As an example of the function, a software application executed by the function recommendation device 1 can be given. In the function recommendation device 1, the function is associated with an icon. The function recommendation device 1 presents the function to be provided on the screen as an icon associated with the function, and the user selects the presented icon. The function associated with the selected icon is executed.

  The function recommendation device 1 learns probability information (likelihood and priority) of function execution (startup) for the date and position from the user's past operation history, and corresponds to a function having a high likelihood at the time of operation. Recommend by presenting. FIG. 2 is a diagram illustrating a screen example at the time of function recommendation in the function recommendation device 1. For example, in the screen example of FIG. 2, four icons corresponding to functions with a high likelihood by the user are displayed mainly when the screen is displayed. Specifically, as shown in FIG. 2, an icon I1 for using the route search service for train transfer guidance of the function recommendation device 1, a map icon I2 for executing the map function of the function recommendation device 1, and a function recommendation A telephone icon I3 for executing the telephone function of the device 1 and a weather icon I4 for executing the weather forecast browsing function of the function recommendation device 1 are displayed.

  Next, the configuration of the function recommendation device 1 will be described. In FIG. 1, the functional block diagram which shows the structure of the function recommendation apparatus 1 which concerns on this embodiment is shown. As shown in FIG. 1, the function recommendation device 1 includes a function execution receiving unit 12 (function execution receiving unit), an environment information acquisition unit 14 (environment information acquisition unit), a spread information generation unit 16 (spread information generation unit), and an execution history. Storage unit 10 (function execution history storage unit), function recommendation model update unit 18 (function recommendation model update unit), function recommendation model 15 (function recommendation model), recommended function determination unit 22 (recommended function determination unit), and recommended function The display unit 24 (recommended function output means) is included.

  The function recommendation device 1 is composed of hardware such as a CPU. FIG. 3 is a diagram illustrating an example of a hardware configuration of the function recommendation device 1. As shown in FIG. 3, the function recommendation device 1 shown in FIG. 1 physically includes a CPU 100, a RAM 101 and a ROM 102 as main storage devices, an input / output device 103 such as a numeric keypad and a display, a communication module 104, And an auxiliary storage device 105 and the like.

  The function of each functional block of the function recommendation device 1 shown in FIG. 1 is controlled by the CPU 100 by loading predetermined computer software on the hardware such as the CPU 100 and the RAM 101 shown in FIG. This is realized by operating the communication module 104 and the auxiliary storage device 105 and reading and writing data in the RAM 101.

  Hereinafter, each functional block of the function recommendation device 1 shown in FIG. 1 will be described.

  The function execution receiving unit 12 receives function identification information (function ID) for identifying the executed function. Specifically, the function execution receiving unit 12 first receives a function execution by the user. In function execution, when the user selects an icon of a function displayed on the display screen of the function recommendation device 1, the function execution receiving unit 12 recognizes that the function is executed. Next, the function execution accepting unit 12 refers to the function ID table stored in advance in the function recommendation device 1 in which the function name is associated with the function ID that is identification information of the function, and corresponds to the selected icon. The function ID of the function to be output is output. FIG. 4 is a diagram illustrating a table example of the function ID table. For example, when the screen shown in FIG. 2 is displayed and the user selects a telephone icon, the function execution accepting unit 12 refers to the function ID table shown in FIG. 4 and outputs “0001” which is the telephone function ID. To do. The function accepted by the function execution accepting unit 12 may be recommended by the function recommendation device 1 or may not be recommended.

  The environment information acquisition unit 14 acquires environment information including at least one of time information when the function identification information is received by the function execution receiving unit 12 and position information of the function recommendation device 1. The environment information acquisition unit 14 may acquire environment information at an arbitrary timing such as a timing indicated by the function recommendation device 1. Specific examples of the environment information acquired by the environment information acquisition unit 14 include the current date, day of the week, time, and the position where the function recommendation device 1 exists. The location information is acquired using GPS (Global Positioning System) or mobile phone base station information. The position information may be latitude / longitude data or semantic data. Semantic data is data representing the meaning of a location, such as a commercial facility, hospital, or library. In order to acquire semantic data, it is necessary to perform semantic classification using known knowledge data. Specifically, for example, after acquiring data that can be acquired by the terminal, such as latitude / longitude and APN (Access Point Name), the data is meaningful by a reference table stored in the function recommendation device 1 in advance. Converted to data. Note that the environment information may be continuous or discretely expressed by sampling.

  The execution history storage unit 10 stores execution history information in which the function identification information received by the function execution receiving unit 12 and the environment information acquired by the environment information acquisition unit 14 are associated with each other. Specifically, the execution history storage unit 10 receives, as inputs, the function ID of the function executed by the user and the environment information acquired by the environment information acquisition unit 14 received by the function execution receiving unit 12. Is stored in the database. The execution history storage unit 10 executes execution history information stored in a database or execution history statistical information after a predetermined calculation process such as statistical processing is performed on the execution history information in the spread information generation process described later. Is output. The database may be managed by either a memory in the function recommendation device 1 or a server on the network. FIG. 5 is a diagram illustrating an example of a table of execution history information stored by the execution history storage unit 10. As shown in FIG. 5, the function ID corresponding to the executed function, the date, time, and latitude / longitude at the time of execution are stored. The information stored by the execution history storage unit 10 may be the same as the environment information used in the function recommendation model 15 described later.

  The function recommendation model 15 is a database in which function priority information in which function identification information and a priority of recommendation of a function identified by the function identification information are associated with each predetermined unit of environment information is stored. In addition, it shows that the priority which recommends a corresponding function is so high that the value of a priority is large.

  As the function recommendation model 15, a function priority map and / or a certainty map may be stored.

  Here, the function priority map is a map in which the priority of each recommended function is set for the value (slot) of environmental information to be considered. A slot is a range formed by a minimum unit of an axis such as date and position. For example, when a function priority map is formed by day of the week and time, and the day of the week is sampled every day and the time is sampled every hour, the slot of the function priority map corresponds to “14:00 on Tuesday” The range is 1 slot.

  The certainty factor map is a map in which each slot has a certainty factor calculated based on the function priority of each recommended function with respect to the value of the environmental information to be considered. The function priority map and the space are the same, but the value is not a priority but a certainty. The certainty factor is, for example, a value indicating the ratio of the number of executions for each function under each slot condition. As an example, 0.14 is given as the probability (confidence) that a telephone call will be used at 9:00 am.

  When the function identification information is received by the function execution receiving unit 12, the transmission information generating unit 16 executes the transmission information related to the method for updating the function priority information based on the execution history information, stored in the execution history storage unit 10. Generate based on history information. Specifically, the propagation information generation unit 16 outputs a propagation parameter (propagation information) used by the function recommendation model update unit 18 based on past execution history information acquired from the execution history storage unit 10. The propagation parameter may be a value indicating the number of slots in a range in which the history is propagated to different slots of the function priority map, or may be a function having environment information forming the function priority map as an argument.

  When the propagation parameter is a value indicating the number of slots in the range in which the history is propagated to different slots in the function priority map, the position (spreading source) serving as the reference of the range is the time when the function is executed (function execution reception unit 12). Is set based on the environmental information at the time of acceptance. Moreover, it is preferable that the addition value for the function priority information within the propagation range is the same as or smaller than the addition value of the execution history of the propagation source. Accordingly, it is possible to consider that the probability that the function in the environment information different from the transmission source is executed is not higher than the probability that the function is executed in the environmental information of the transmission source.

  The past execution history information acquired from the execution history storage unit 10 by the spread information generation unit 16 may be the cumulative number of execution histories. In this case, the propagation parameter is determined based on the cumulative number of execution histories. When the propagation parameter indicates the range in which the execution history is propagated, the propagation information generation unit 16 may narrow the range to be propagated as the cumulative number of execution histories increases. Thereby, it is possible to prevent the density of the function priority map from becoming excessively large with respect to the increase in the cumulative number of histories, and it is possible to generate a function priority map having an appropriate density.

  Here, the density of the function priority map is the number of slots indicating a high priority among the slots of the function priority map. For example, when there are many slots whose function priority is 0, the density is low, and when there are many slots with a high function priority, the density is high. The function priority map density is excessively high. Since sufficient function priorities are stored for each slot, there is a difference in function priority between the slots even though the execution history count is large. It refers to the state that is difficult to stick.

  Further, the spread information generation unit 16 may fix the spread range when the cumulative number becomes constant. Thereby, it is possible to prevent the spillover range of the history from becoming excessively narrow, and to have fluctuations in the function recommendation.

  Hereinafter, the function of the spread information generation unit 16 will be described using a specific example. In FIG. 6, the propagation parameter D is the number of slots in the range in which the history is propagated to different slots in the function priority map, and the propagation parameter is obtained by the equation of FIG. The ripple constant k is a constant required when calculating the ripple range, and may be a resolution value of the space axis of the function recommendation model 15. For example, if the function recommendation model 15 has a 24-hour axis with resolution every other hour, the ripple constant is 24. The propagation parameter is obtained by dividing the propagation constant by the cumulative number of function executions.

  Specifically, as shown in FIG. 6B, when the function execution cumulative number N is 12, the propagation parameter D = 24 × (1/12) = 2. In the case of FIG. 6B, the priority of the function is incremented by 1 for a slot of ± 2 hours with the time (the transmission source) when the function identification information is received by the function execution receiving unit 12 as the reference time 0. The propagation information generating unit 16 generates the propagation parameter. On the other hand, as shown in FIG. 6C, when the cumulative function execution count N is 24, the propagation parameter D = 24 × (1/24) = 1. In the case of FIG. 6C, a propagation parameter that adds 1 to the priority of the function for a slot of ± 1 hour, with the time when the function identification information is received by the function execution receiving unit 12 as the reference time 0. Generated by the spread information generation unit 16.

  When the propagation parameter is set as the history propagation range, the magnitude of the addition value for the function priority information within the propagation range cannot be represented by the propagation parameter. In this case, for example, as shown in FIG. 6, it may be the same value as the addition value of the transmission source, or may be changed linearly or nonlinearly so that the addition value becomes 0 at the end of the transmission range indicated by the transmission parameter. Also good. FIG. 7 is an example in which the addition value is changed linearly in the same spreading range as in FIG. In the graph shown in FIG. 7, the addition value decreases linearly with the ripple source as the maximum value.

  On the other hand, FIG. 8 shows a specific example in the case where the propagation parameter is expressed by a function having environment information forming the space of the function priority map as an argument. In this case, as the cumulative number of execution histories increases, the distribution of the normal distribution of the function priority map decreases and the spread range also decreases. Regarding the magnitude of the priority to be added, it can be regarded as a probability density function of 1 in total, and the priority can be calculated by adding the value of normal distribution in each slot. The normal distribution function may be modified so that In FIG. 8, the normal distribution based on P represents the distribution of ripples. Further, as shown in FIG. 8, the magnitude of the added value in the ripple range decreases nonlinearly with the ripple source as the maximum value.

  In addition, a case will be described in which position information is included in the environment information forming the function priority map space of the function recommendation model 15. As described above, the position information may be expressed as latitude / longitude or as semantic data.

  When the position information is expressed by latitude / longitude, two axes of the function priority map are configured by the latitude and longitude. For each function, a three-dimensional function priority map indicating a priority value only by position information may be used, or a function priority map of four or more dimensions may be combined with time and day of the week. FIG. 9 shows an image in which history is propagated to a position using a three-dimensional function priority map. The cone shown in FIG. 9 indicates the added value of the function priority information in a range centered on a predetermined latitude (x axis) and longitude (y axis) (z axis indicates priority).

  When the position information is expressed by semantic data, the position information may be structured in a graph and the distance between the classification positions may be set in a pseudo manner. FIG. 10 is a diagram illustrating a graph example in which numbers between nodes are represented by pseudo distances. In the example of FIG. 10, when “2” is set as the propagation parameter indicating the propagation range and the function A is executed at the accommodation facility, it is an entertainment facility and a commercial facility that are within a distance of 2 from the accommodation facility. The addition value 1 is spread only to the function priority maps of the entertainment facility and the commercial facility of function A. In addition, each classification position may be plotted on a map composed of distances in the X-axis direction and the Y-axis direction as shown in FIG. 11, and the Euclidean distance may be used to calculate the distance. In addition, a table having a distance to each position may be held for each position.

  The past execution history information acquired from the execution history storage unit 10 by the spread information generation unit 16 may be distributed information of the past execution history of the user. Since the distributed information varies greatly depending on the number of execution histories, it is preferable to use a distributed value of a certainty factor map obtained by normalizing the function priority map. Specifically, a conditional posterior probability value indicating the probability that each function is executed under the condition of each slot in the environment information of the function priority map is obtained as the certainty factor. Therefore, the variance value is 0 to 1.

  FIG. 12 shows a specific example in which the propagation parameter is obtained using the distribution information of the past execution history of the user. The graph shown in FIG. 12 is a normal distribution function based on the variance value of the Gaussian distribution indicating the ripple distribution based on the variance value of the confidence calculated with the confidence map. However, the ripple parameter is not a Gaussian distribution value, but may be a slot number indicating the ripple range, a function having environment information as an argument, or a combination thereof. Here, the variance value of the certainty factor is obtained by using an average and a variance calculated from the certainty factor to be considered as inputs. The certainty factor in the certainty factor map can be determined by calculating the variance from the certainty factor of all slots in all functions, or by limiting the environmental information that forms the certainty factor map, You may ask for it.

  The past execution history information acquired from the execution history storage unit 10 by the propagation information generation unit 16 may be distributed information of the execution history obtained for each environment information forming the function priority map space. Thereby, the correlation value with respect to each axis forming the function priority map is known for each environment information, and the spreading method can be changed.

  The past execution history information acquired from the execution history storage unit 10 by the spread information generation unit 16 is the distribution information of the execution history obtained for each function that is a recommendation candidate and for each environment information that forms the function priority map space. It is good. Thus, for each function, the correlation value for each axis forming the function priority map is known, and the spreading method can be changed. For example, since the map function has a strong correlation with the position, an effect of narrowing the spread range can be expected.

  When the past execution history information acquired from the execution history storage unit 10 by the spread information generation unit 16 is used as the distribution information of the past execution history, the distributed information filtered in a certain period may be used. The certain period may be the most recent period or a past certain period. Thereby, it is possible to cope with a change in function execution pattern due to a change in time.

  The spread information generation unit 16 acquires the cumulative number of past execution histories and the past execution history distribution information from the execution history storage unit 10, determines the spread parameters, and weights and combines the two spread parameters. One propagation parameter may be determined. As described above, there are a plurality of representations of the propagation parameter. However, when two or more propagation parameters are weighted and synthesized, the propagation parameter is expressed in the same way. For example, the number of slots indicating the propagation range or the normal distribution distribution is represented. The variance value of the distribution. As a specific example, when the propagation parameter indicates the number of slots indicating the propagation range, the propagation parameter obtained from the cumulative number of execution histories is 2, and the propagation parameter obtained from the execution history distribution information is 4. The two parameters are synthesized in a one-to-one relationship, and the ripple parameter is 3. At this time, the weight ratio is not 1: 1 and may be changed according to the importance of the execution history information.

  The function recommendation model update unit 18 stores the function priority information stored in the function recommendation model 15 based on the execution history information stored in the execution history storage unit 10 and the propagation parameter generated by the propagation information generation unit 16. Update. The function recommendation model updating unit 18 stores the function priority stored in the function recommendation model 15 based on the function executed by the user received by the function execution receiving unit 12 and the propagation parameter generated by the propagation information generating unit 16. The degree information may be updated. The function recommendation model updating unit 18 uses the execution history information stored in the execution history storage unit 10 in order to determine the function priority information to be updated among the function priority information stored in the function recommendation model 15. Use.

  FIG. 13A is a diagram illustrating an example of a function priority map of telephone functions stored in the function recommendation model 15. For example, when the telephone function is executed at 12:00 and the propagation parameter indicating the propagation range is 2, the function recommendation model updating unit 18 first performs the function of the telephone function shown in FIG. Get function priority map. Since the propagation parameter is 2, priority 1 is added to the slot from 10:00 to 14:00. The function priority map after the addition is shown in FIG.

  Here, the priority added to the 12:00 slot of the transmission source may be a value other than 1, and the priority added to the slot within the transmission range is the priority added to the slot of the transmission source 12:00. May be equal to or smaller than. Further, the calculation for the priority may be subtraction or integration instead of addition. When the function priority map space becomes a dimensional space larger than the two-dimensional space as shown in FIG. 13, the priority is added to the slots satisfying both the propagation ranges of the environment information constituting the map space. Alternatively, the priority may be added as long as one of the spreading ranges is satisfied. For example, when the environment information constituting the map space is a day of the week and time, and the day of the week has a spread range of the day of the week, the spread range of the time is 1 hour. If a function at 14:00 is executed, the slot becomes a slot at Monday or Tuesday or Wednesday and at 13-15 o'clock. On the other hand, the latter is a slot that satisfies either the day of the week or the time range.

  Further, the propagation parameter generated by the propagation information generation unit 16 is a propagation parameter for the execution function received by the function execution reception unit 12, but the execution history information (propagation) stored by the execution history storage unit 10. The function recommendation model 15 may be newly regenerated by retroactively spreading the execution history information (applicable parameters). The function recommendation model 15 that more appropriately considers the execution history information is generated by applying the propagation parameter not only to the execution history received by the function execution receiving unit 12 but also to the stored execution history information. can do.

  The recommended function determination unit 22 determines a function to be recommended based on the function priority information stored in the function recommendation model 15. Specifically, the recommended function determination unit 22 receives the environment recommendation acquired from the function recommendation model 15 and the environment information acquisition unit 14 and determines a function recommended to the user. More specifically, a function to be recommended is determined by referring to a function priority map of each function of the function recommendation model 15 and comparing priorities based on environment information. For example, when the function priority map of the function recommendation model 15 is composed of two dimensions of day of the week and time, and a function is recommended at 8:10 on Sunday, it corresponds to 8:10 on Sunday of each function. Compare priorities. Thereafter, the functions are extracted in descending order of priority, and the function IDs of the functions are output. When the function recommendation model 15 holds a certainty factor map obtained by normalizing the function priority map, the certainty factor of each function may be compared to determine a recommended function. The number of functions to be recommended may be determined in advance by a predetermined recommended function determination parameter. When the recommended function is determined by the certainty factor, a threshold value is set in advance and is equal to or greater than the threshold value. All of these functions may be determined as recommended functions.

  The timing at which the function recommendation is performed by the recommended function determination unit 22 is arbitrary. For example, when the user activates a smartphone or when using this function recommendation system.

  The recommended function display unit 24 outputs information on the recommended function determined by the recommended function determination unit 22. Specifically, the recommended function display unit 24 displays an icon corresponding to the function ID of the recommended function determined by the recommended function determination unit 22 on the display screen of the function recommendation device 1. The recommended function display unit 24 may output information related to the recommended function to another server via the network.

  Next, processing of the function recommendation method in the function recommendation device 1 according to the present embodiment will be described using the flowchart shown in FIG.

  First, the function execution accepting unit 12 obtains the function ID of the function executed by the user (S1, function execution accepting step). Next, the current environmental information is acquired by the environmental information acquisition unit 14 (S2, environmental information acquisition step). Next, the execution history storage unit 10 stores the function ID acquired in S1 and the environment information acquired in S2 as execution history information, and also acquires past execution history information (S3, function execution history storage). Step). Next, the propagation information generation unit 16 generates a propagation parameter based on the function ID acquired in S1 and the execution history information acquired in S3 (S4, propagation information generation step). Next, the function recommendation model update unit 18 updates the function recommendation model 15 based on the function ID acquired in S1, the environment information acquired in S2, and the propagation parameter acquired in S4 (S5). , Function recommendation model update step). Next, the recommended function determination unit 22 refers to the function recommendation model 15 updated in S5, and determines a function recommended to the user (S6, recommended function determination step). Next, an icon corresponding to the function is displayed on the screen of the function recommendation device 1 by the recommended function display unit 24, so that the function is recommended to the user (S7, recommended function output step).

  Then, an effect | action and effect of the function recommendation apparatus 1 which concern on this embodiment are demonstrated.

  According to the function recommendation device 1 according to the present embodiment, first, the function execution receiving unit 12 receives the function execution. Next, based on the execution history information acquired from the execution history storage unit 10, the propagation information generation unit 16 generates propagation information for updating the function recommendation model 15. Based on the propagation parameter generated by the propagation information generation unit 16, the function ID of the function executed by the user, and the environment information acquired by the environment information acquisition unit 14, the function recommendation model update unit 18 generates the function recommendation model 15. Updated. With such a configuration, the propagation parameter is dynamically determined based on the execution history information. That is, it is possible to appropriately determine the spreading method according to the execution history information, and it is possible to perform function recommendation utilizing the history that is currently held.

  A specific example of the function and effect of the function recommendation device 1 according to the present embodiment will be described based on the function priority map example shown in FIG. In FIG. 15, the function priority map is formed only with respect to time, and the propagation parameter indicates the number of slots in the range to be propagated, and the size of the propagation is linearly reduced. Here, an example is shown in which the propagation constant is 8, the functions to be propagated are all functions, and the function recommended at 14:00 is obtained.

  FIG. 15B shows a function priority map when the cumulative number of execution histories is two. Specifically, a state in which the function A has been executed twice at 12:00 and 16:00 in the past is shown. The spreading parameter is 8 × (1/2) = 4 according to the formula shown in FIG. FIG. 15B shows a function priority map updated according to the propagation parameter. As shown in the function priority map shown in FIG. 15B, the function A can be recommended at 14:00 when there is no execution history.

  On the other hand, FIG. 15C shows a function priority map when the cumulative number is 3. Specifically, after the state shown in FIG. 15B, a state in which the function B is executed at 15:00 is shown. The propagation parameter is 8 × (1/3) = 3 according to the equation shown in FIG. FIG. 15C shows a function priority map that is updated according to the propagation parameter. As shown in the function priority map shown in FIG. 15C, the function B is recommended because the function B has a higher priority than the function A at 14:00.

  As described above, the function recommendation device 1 can recommend the function A that has been executed in the time zone close to 14:00 when the number of execution histories is small as shown in FIG. Here, in the state where the number of execution histories is large as shown in FIG. 15C, in the related art where the spillover range is not based on the number of execution histories, the spillover range is large. The priority of the function A that is executed more frequently in the time zone (12 o'clock and 16 o'clock) that is farther than the function B that is executed at the same time becomes higher, and the function A is recommended. On the other hand, in the function recommendation device 1, in the state where the number of execution histories is large as shown in FIG. 15C, the spread range becomes small as described above, so that the addition value of the function A becomes small at 14:00, Function B has a higher priority than A, and function B is recommended. Thus, in the function recommendation apparatus 1, it becomes possible to perform the expected appropriate prediction by dynamically determining the propagation parameter.

[First Modification of First Embodiment]
Next, a function recommendation device 1A according to Modification 1 of the first embodiment of the present invention will be described. The function recommendation device 1A is almost the same as the function of the function recommendation device 1, and only the difference will be described.

  FIG. 16 is a functional block diagram showing the configuration of the function recommendation device 1A according to this modification. The function recommendation device 1A differs from the function recommendation device 1 in the other user execution history storage unit 13A and the spread information generation unit 16A.

  The other user execution history storage unit 13A stores the execution history of other users, and outputs a dispersion value of the statistical data of the other users' execution history stored therein. The data stored in the other user execution history storage unit 13A may be acquired in advance in the terminal, or statistical data of other user execution history may be acquired as appropriate via the network. The variance value is calculated using the certainty factor in the function recommendation model 15 of another user as described above. The variance value may be obtained as one value from all priorities of the function recommendation model 15, or may be obtained for each environment information as an axis constituting the certainty factor map, and further, the environment information for each function. The variance value for each may be obtained.

  The spread information generation unit 16A spreads based on the distribution value of the statistical data of the execution history of other users acquired from the other user execution history storage unit 13A and the function ID of the function executed by the user received by the function execution reception unit 12. Generate parameters. The propagation information generation unit 16A may generate the propagation parameter only when the variance value of the statistical data of the execution history of other users is smaller than a predetermined value.

  Then, an effect | action and effect of 1 A of function recommendation apparatuses which concern on this embodiment are demonstrated.

  According to the function recommendation device 1A according to the present embodiment, by using the execution history of another user, even if the cumulative number of execution histories of the user is poor, the propagation parameter is set for the propagation parameter that is not user-dependent. Can be determined. For example, if the position of the function recommendation model 15 has a dimension, if the variance of the statistical data of the certainty factor for the position information of the alarm function is a small value close to 0, the usage tendency of the alarm depends on the user. And it does not depend on the position. In this case, when the user executes the alarm function, the function recommendation model 15 can be updated by maximizing the spillover range for the position. In other words, even if the cumulative number of execution histories of the user is small or the variance value is not stable, the propagation parameters can be generated using the execution histories of other users and function recommendations can be made appropriately. .

[Modification 2 of the first embodiment]
Next, a function recommendation device 1B according to Modification 2 of the first embodiment of the present invention will be described. The function recommendation device 1B is substantially the same as the function of the function recommendation device 1, and only the difference will be described.

  FIG. 17 is a functional block diagram showing the configuration of the function recommendation device 1B according to this modification. The function recommendation device 1B is different from the function recommendation device 1 in the propagation information generation unit 16B.

  The spread information generation unit 16B includes the environment information acquired by the environment information acquisition unit 14, the function ID of the function executed by the user acquired by the function execution receiving unit 12, and the execution history information acquired from the execution history storage unit 10. The propagation parameter is generated based on the above. At this time, the propagation parameter is generated based on the execution history obtained by filtering the environment information in the divided range. For example, when the environment information of time constitutes a function priority map and the telephone function is executed at 4 o'clock, the propagation parameter indicating the propagation range is determined based on the execution history at 0-6 o'clock, and the function recommendation model Used for 15 updates. The range of environment information used for generating the propagation parameter may be determined in advance, or may be automatically determined based on execution history information such as the number of execution histories and distributed information.

  Next, operations and effects of the function recommendation device 1B according to the present embodiment will be described.

  According to the function recommendation device 1B according to the present embodiment, an appropriate propagation parameter can be generated when the function execution tendency varies depending on the value of the environment information. Specifically, if the number of execution histories of the function from 12:00 to 18:00 is small, but the number of execution histories in the time zone from 6:00 to 12:00 is enormous, the function is executed as in the first embodiment. If the propagation parameter is generated based on the cumulative number of times, the propagation range is set to be small, so even if the function is executed in the time zone from 12:00 to 18:00, the propagation range becomes small. For such a problem, by dividing the range of the history used for generating the transmission parameter, the transmission range of 6:00 to 12:00 is set to be small, and the transmission range of 12:00 to 18:00 is set to be large. It becomes possible.

  DESCRIPTION OF SYMBOLS 1 * 1A * 1B ... Function recommendation apparatus, 10 ... Execution history storage part, 12 ... Function execution reception part, 13A ... Other user execution history storage part, 14 ... Environment information storage part, 15 ... Function recommendation model, 16 * 16A * 16B ... Ripple information generation unit, 18 ... Function recommendation model update unit, 22 ... Recommended function determination unit, 24 ... Recommended function display unit.

Claims (9)

A function recommendation device that recommends one or more specific functions from among a plurality of functions,
Function execution accepting means for accepting function identification information for identifying the executed function;
Environmental information acquisition means for acquiring environmental information including at least one of time information when the function identification information is received by the function execution reception means and position information of the function recommendation device;
A function execution history storage unit that stores execution history information in which the function identification information received by the function execution reception unit and the environment information acquired by the environment information acquisition unit are associated;
A function recommendation model in which the function priority information associated with the function identification information and the function recommendation priority identified by the function identification information is associated with each predetermined unit of the environment information;
When the function identification information is accepted by the function execution accepting unit, the propagation information on the method for updating the function priority information based on the execution history information is generated based on the execution history information stored by the function execution history storage unit. A ripple information generating means for
Function recommendation model update for updating function priority information stored in the function recommendation model based on the execution history information stored by the function execution history storage unit and the propagation information generated by the propagation information generation unit Means,
Based on function priority information stored in the function recommendation model, recommended function determining means for determining a function to recommend;
Recommended function output means for outputting information on the recommended function determined by the recommended function determination means;
A function recommendation device comprising: The spread information generating means generates spread information based on the cumulative number of execution history information stored by the function execution history storage means.
The function recommendation device according to claim 1. The spread information generating means generates spread information based on the distribution information of the execution history information stored by the function execution history storage means.
The function recommendation device according to claim 1. The spread information generating means generates spread information for each environment information based on the distributed information for each environment information of the execution history information stored by the function execution history storage means.
The function recommendation device according to claim 1. The spread information generating means generates spread information for each environment information for each function based on the function identification information and the distributed information for each environment information of the execution history information stored by the function execution history storage means.
The function recommendation device according to claim 1. The spread information generating means generates the spread information further based on the execution history information stored by the function execution history storage means of another function recommendation device.
The function recommendation device according to any one of claims 3 to 5, wherein The spread information generating unit weights the spread information generated based on the cumulative number of execution history information stored by the function execution history storage unit and the spread information generated based on the distributed information of the execution history information. Produce ripple information by combining,
The function recommendation device according to claim 1. The spread information generating unit generates the spread information based on the execution history information of the environment information in a predetermined range among the execution history information stored by the function execution history storage unit.
The function recommendation device according to claim 1, wherein the function recommendation device is a device. A function recommendation device that recommends one or more specific functions from among a plurality of functions, wherein the function identification information for identifying the function and the priority of the function recommendation identified by the function identification information are time information. And a function recommendation method executed by a function recommendation device having a function recommendation model in which function priority information associated with each predetermined unit of environment information including at least one of position information of the function recommendation device is stored. There,
A function execution accepting step for accepting function identification information of the executed function;
An environment information acquisition step of acquiring environment information when function identification information is received in the function execution reception step;
A function execution history storage step for storing execution history information in which the function identification information received in the function execution reception step and the environment information acquired in the environment information acquisition step are associated;
When function identification information is received in the function execution receiving step, propagation information on a method for updating function priority information based on the execution history information is generated based on the execution history information stored in the function execution history storing step. A ripple information generation step to perform,
Function recommendation model update for updating function priority information stored in the function recommendation model based on the execution history information stored in the function execution history storage step and the propagation information generated in the propagation information generation step Steps,
A recommended function determining step for determining a function to be recommended based on the function priority information stored in the function recommendation model;
A recommended function output step of outputting information on the recommended function determined in the recommended function determination step;
Function recommendation method including

Images