Recommendation device and recommendation method

Abstract

This recommendation device recommends actions that promote an improvement in the health of a user, said device including: an acquisition unit for acquiring user information indicating attributes of and / or the state of the user; and a recommendation unit for recommending at least one action on the basis of the user information. The acquisition unit acquires the following: from among the recommended actions, an action accepted by the user; and results information indicating a health improvement effect due to the action. On the basis of the user information and the results information, the recommendation unit determines the action to be recommended.

Description

Recommendation device and recommendation method

[0001] The disclosed technology relates to a recommendation device and a recommendation method.

[0002] In recent years, there has been an increasing interest in health improvement, including states of discomfort that do not lead to diseases. For not only an individual but also, for example, a company, improving the health of employees is important because it leads to suppressing a decline in productivity. States of discomfort include, for example, mental, psychological, and physical ones such as tension, anxiety, depression, listlessness, fatigue, decreased concentration, drowsiness, stiff shoulders, low back pain, and eye strain.

[0003] In general, it is known that various actions such as massage, taking supplements, taking a short nap, and stretching are effective for improving discomfort. A system for promoting the health improvement of users by inducing such actions has been proposed. For example, Non-Patent Document 1 describes a method of presenting a message for inducing a healthy action by performing rule processing using user attribute data, as well as real-time action data and environmental data.

[0004] Yasuyuki Taniguchi, Azuma Tsubota, Koshi Sakurada, "Development of a Health Promotion Prompt System Using Behavior Change Theory", Proceedings of the Symposium, Vol. 1, Institute of Electronics, Information and Communication Engineers Communication Society Conference, 2018, p. 425

[0005] Various actions for improving discomfort may have individual differences in the effects when implemented. In rule processing using user attributes, actions, and environment as in the prior art, individual differences in effects cannot be considered.

[0006] The disclosed technology has been made in view of the above points, and an object thereof is to provide a recommendation device and a recommendation method that can recommend actions for promoting the health improvement of users, which have a high acceptance rate by users and are effective.

[0007] A first aspect of this disclosure is a recommendation device for recommending actions that promote a user's health improvement, comprising: an acquisition unit for acquiring user information indicating at least one of the user's attributes and status; and a recommendation unit for recommending at least one of the actions based on the user information, wherein the acquisition unit acquires, from among the recommended actions, the actions accepted by the user and result information indicating the health improvement effect corresponding to those actions; and the recommendation unit determines the actions to recommend based on the user information and the result information.

[0008] A second aspect of this disclosure is a recommendation method for recommending actions that promote a user's health improvement, comprising: acquiring user information indicating at least one of the user's attributes and status; recommending at least one of the actions based on the user information; and acquiring, from among the recommended actions, the action accepted by the user and result information indicating the effect of the action on improving health, wherein in recommending the action, a computer performs a process to determine the action to recommend based on the user information and the result information.

[0009] According to the disclosed technology, it is possible to provide a recommendation device and recommendation method that can recommend actions that promote user health improvement, have a high acceptance rate among users, and are effective.

[0010] This figure shows an example of the general configuration of the recommendation system. This block diagram shows an example of the hardware configuration of the recommendation device. This block diagram shows an example of the functional configuration of the recommendation device. This figure shows an example of an attribute table. This figure shows an example of a recommendation list table. This figure shows an example of an action table. This figure shows an example of a similarity table. This figure shows an example of a results table. This figure illustrates how to derive a score using result information. This figure illustrates how to derive a score using similarity. This figure shows an example of output content. This figure shows an example of output content. This is a flowchart showing the flow of the recommendation process.

[0011] An example of an embodiment of the disclosed technology will be described below with reference to the drawings. In each drawing, identical or equivalent components and parts are given the same reference numerals. Furthermore, the dimensional ratios in the drawings are exaggerated for illustrative purposes and may differ from actual ratios.

[0012] Figure 1 shows an example of the configuration of a recommendation system 100 to which the recommendation device 10 of this disclosure is applied. The recommendation system 100 includes the recommendation device 10 and a plurality of user terminals 90. The recommendation device 10 and each user terminal 90 are connected to each other in a manner that enables communication via a wired or wireless network (not shown). The network is, for example, the Internet, LAN (Local Area Network), or WAN (Wide Area Network).

[0013] The user terminal 90 is a terminal device owned by each user. The user terminal 90 is used for inputting user information, inputting result information, and presenting recommended actions, as described later. As the user terminal 90, for example, a personal computer, smartphone, tablet device, and wearable device can be appropriately applied.

[0014] The recommendation device 10 is a device for recommending actions that promote the user's health improvement, including discomfort that does not lead to illness. Discomfort includes mental, psychological, and physical conditions such as tension, anxiety, depression, lethargy, fatigue, decreased concentration, drowsiness, stiff shoulders, lower back pain, and eye strain.

[0015] It is generally known that various actions such as massage, taking supplements, napping, and stretching are effective in improving physical ailments. However, some actions, such as massage and taking supplements, tend to be less readily accepted when recommended because they place a significant time and financial burden on the user.

[0016] Therefore, the recommendation device 10 recommends various actions for improving discomfort, such as napping and stretching, which the user can perform spontaneously and easily. Because such simple actions are easy for the user to take, the percentage of users who perform them when recommended (hereinafter referred to as the acceptance rate) tends to be high.

[0017] On the other hand, the effectiveness of various actions to improve health, including such simple actions, may vary from person to person. Therefore, the recommendation device 10 aims to recommend actions that are effective for each user by receiving feedback from the user on the effectiveness of the recommended actions. The recommendation device 10 according to this embodiment will be described in detail below. In the following description, actions that promote the improvement of the user's health may be simply referred to as "actions."

[0018] Figure 2 is a block diagram showing an example of the hardware configuration of the recommendation device 10. The recommendation device 10 includes a CPU (Central Processing Unit) 11, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 13, storage 14, an input unit 15, a display unit 16, and a communication interface 17. Each component is connected to the others via a bus 19 so as to be able to communicate with each other.

[0019] The CPU 11 is a central processing unit that executes various programs and controls various parts. Specifically, the CPU 11 reads a program from the ROM 12 or storage 14 and executes the program using the RAM 13 as a working area. The CPU 11 controls each of the above components and performs various calculations according to the program stored in the ROM 12 or storage 14. In this embodiment, the ROM 12 or storage 14 stores a recommendation program for executing the recommendation process described later.

[0020] ROM 12 stores various programs and data. RAM 13 temporarily stores programs or data as a working area. Storage 14 consists of storage devices such as HDD (Hard Disk Drive) and SSD (Solid State Drive) and stores various programs and data, including the operating system.

[0021] The input unit 15 includes a pointing device such as a mouse and a keyboard, and is used for various types of input. The display unit 16 is, for example, a liquid crystal display and displays various types of information. The display unit 16 may also function as the input unit 15 by employing a touch panel system.

[0022] Communication I / F 17 is an interface for communicating with other devices. For this communication, wired communication standards such as Ethernet (registered trademark) or FDDI (Fiber Distributed Data Interface), or wireless communication standards such as 4G, 5G, or Wi-Fi (registered trademark) can be used. As the recommended device 10, for example, a server computer can be appropriately applied.

[0023] Figure 3 is a block diagram showing an example of the functional configuration of the recommendation device 10. The recommendation device 10 includes, as a functional configuration, an acquisition unit 30, an extraction unit 32, a recommendation unit 34, an output unit 36, and an evaluation unit 38. Each functional configuration is realized by the CPU 11 reading a recommendation program stored in the ROM 12 or storage 14, expanding it into the RAM 13, and executing it.

[0024] The recommendation device 10 also includes an attribute table T1, a recommendation list table T2, an action table T3, a similarity table T4, and a result table T5. Each table is stored in ROM 12, storage 14, or an external storage device (not shown).

[0025] Figure 4 shows an example of attribute table T1. Attribute table T1 is a table in which a group ID is pre-assigned to each combination of multiple items that represent the user's attributes. User attributes include, for example, age, gender, preferences, and lifestyle. Lifestyle includes, for example, how leisure time is spent and work style such as the length of each break during work.

[0026] Figure 5 shows an example of the recommendation list table T2. The recommendation list table T2 contains pre-registered recommendation lists for each combination of group ID and multiple items indicating the user's status. The user's status includes, for example, health-related conditions such as the type of ailment, and environmental conditions such as time of day, location, and work content.

[0027] Figure 6 shows an example of an action table T3. The action table T3 contains pre-registered actions designed to encourage user health improvement. In the example in Figure 6, an action ID is assigned to each recommendation statement describing the content of each action.

[0028] A recommendation list is a ranking of multiple different actions. In the example in Figure 5, the recommendation list ranks N action IDs from the action IDs defined in action table T3, where N is an integer greater than or equal to 2. Furthermore, the recommendation list is predetermined for at least one of the user's attributes and status. In other words, different user attributes and statuses will result in different types and rankings of actions included in the recommendation list.

[0029] The types and rankings of actions included in the recommendation list are determined according to their acceptance rate. Here, the acceptance rate is assumed to differ depending on the user's attributes and state. For example, the acceptance rate for listening to music is assumed to change depending on whether the user likes music or not. Also, the acceptance rate for each action is assumed to change depending on the user's state, such as whether they are sleepy or stressed, whether it is lunchtime or after work, or whether they are at home or in the office. In recommendation list table T2, such changes in acceptance rate are taken into account by defining a recommendation list for at least one of the user's attributes and state. The acceptance rate may be determined, for example, by statistical preliminary surveys of user attributes and states, or it may be determined arbitrarily based on hypotheses.

[0030] Figure 7 shows an example of a similarity table T4. The similarity table T4 contains pre-registered similarities between various behaviors for each combination of behaviors. In the example in Figure 7, the similarity is defined as a normalized value within the range of 0 to 1, with a higher value indicating greater similarity. The similarity can be determined as appropriate; it may be derived by the evaluation unit 38 described later, or it may be arbitrarily determined based on a hypothesis.

[0031] Figure 8 shows an example of the results table T5. The results table T5 records, for each user, the actions that were accepted (implemented) by the user from among the actions recommended by the recommendation device 10 in the past, and result information showing the effect of the health improvement corresponding to those actions. For example, a results table T5 is prepared for each user. In addition, data is added to the results table T5 each time the recommendation device 10 recommends an action.

[0032] In the example shown in Figure 8, a results table T5 for a single user U1 is displayed. The results table T5 records the recommendation ID, date and time, acceptance result, and the effect of the action in association with each other. The recommendation ID is an identification number assigned each time the recommendation device 10 makes a recommendation. The acceptance result includes the action ID of the action accepted by the user from among the various actions included in the recommendation list, and the rank of that action in the recommendation list at the time it was recommended.

[0033] The above table configurations are examples only, and various modifications are possible. For example, attribute table T1 and recommendation list table T2 may be combined into one. Alternatively, for example, the acceptance result items and effect items in result table T5 may be separated and recorded in different tables. Alternatively, for example, instead of preparing a separate result table T5 for each user, a user ID item may be included in result table T5 so that it can be extracted later, and all data for all users may be combined into a single table.

[0034] The acquisition unit 30 acquires user information that indicates the user's attributes and status. The attributes indicated by the user information are the same as those defined in attribute table T1. The status indicated by the user information is the same as the status defined in recommendation list table T2.

[0035] For example, the acquisition unit 30 may acquire user information manually entered by the user using the user terminal 90 from the user terminal 90. Alternatively, for example, the acquisition unit 30 may determine the time period, which is a type of user information indicating a state, using the time when other user information was entered. Alternatively, for example, the acquisition unit 30 may determine the location, which is a type of user information indicating a state, using the location information of the user terminal 90 at the time when other user information was entered.

[0036] Furthermore, in order to reduce the effort required to re-enter user information in subsequent processing, it is preferable for the acquisition unit 30 to register user information acquired once in a database (not shown) for storing user information. If user information is already registered in the database, the acquisition unit 30 may extract user information for the target user from the database. Alternatively, the acquisition unit 30 may combine these methods, for example, by extracting user information indicating attributes that is already stored in the database, and acquiring user information indicating status from the user terminal 90 that has been manually entered by the user using the user terminal 90.

[0037] Furthermore, the acquisition unit 30 retrieves a predetermined number of result information from the result table T5 for the target user, sorted by newest first. The acquisition unit 30 may also perform filtering by date and time before retrieving a predetermined number of result information. For example, if the current time is 2 PM, the unit may filter the results to include only those retrieved between 12 PM and 4 PM, and then retrieve a predetermined number of result information in chronological order.

[0038] Furthermore, the acquisition unit 30 determines that it is the first processing if no result information exists in the result table T5 for the target user.

[0039] The extraction unit 32 extracts recommendation lists from the recommendation list table T2 that correspond to the acquired user information. Specifically, the extraction unit 32 identifies a group ID from the attribute table T1 using the user attributes included in the user information. Next, the extraction unit 32 extracts a recommendation list from the recommendation list table T2 using the identified group ID and the user status included in the user information.

[0040] The recommendation unit 34 recommends at least one action based on the user information. Specifically, the recommendation unit 34 determines which action to recommend based on the recommendation list extracted from the recommendation list table T2 according to the user information. For example, the recommendation unit 34 decides to recommend a predetermined number of actions from the N actions specified in the recommendation list, in order of highest rank.

[0041] Here, if it is the first time a recommendation process is performed for a target user, the recommendation unit 34 uses the recommendation list extracted from the recommendation list table T2 as is to determine the recommended action. On the other hand, if it is the second or subsequent time a recommendation process is performed for a target user, the recommendation unit 34 updates the ranking of actions in the recommendation list extracted from the recommendation list table T2 and determines the recommended action based on the updated recommendation list. In other words, the recommendation unit 34 determines the recommended action based on result information in addition to user information. The method for updating the recommendation list in the second and subsequent recommendation processes will be explained below.

[0042] Recommendation unit 34 derives a score indicating the level of effectiveness for each action based on the result information. Referring to FIG. 9, an example of a method for deriving a score using the result information will be described. In this example, as shown on the left side of FIG. 9, it is assumed that result information for eight cases regarding a certain user has been acquired.

[0043] Recommendation unit 34 calculates a weighted inverse rank for each of the eight pieces of result information. The weighted inverse rank r is obtained by the following formula using the effect a and the rank b of the action included in the result information. Here, for the effect a, when there is an effect, it is set to 1, and when there is no effect, it is set to 0.5. The rank b is the rank in the recommendation list. b is an integer from 1 to N. r = a / b

[0044] The values of the weighted inverse ranks obtained from each of the eight pieces of result information are shown in the upper right part of FIG. 9. For example, in the record with the recommendation ID "R11", since the effect is "none" and the rank in the recommendation list is the 4th, the weighted inverse rank r is 0.5 / 4, which is 0.125.

[0045] Next, recommendation unit 34 calculates the average of the weighted inverse ranks for each action ID. This average value becomes the score indicating the level of effectiveness for each action. The values of the scores for each action ID are shown in the lower right part of FIG. 9. The higher the score value, the higher the effect of the action. For example, among the action IDs "A1" to "A3" in FIG. 9, since the score of "A2" is the highest, it can be seen that the action indicated by "A2" is the most effective.

[0046] In this way, for the actions included in the result information (that is, actions with a recently received history), recommendation unit 34 derives a score indicating the level of effectiveness for each action based on the result information. On the other hand, for actions not included in the result information (that is, actions with no recently received history), it is not possible to appropriately derive a score based only on the result information. Therefore, recommendation unit 34 may derive a score for actions not included in the result information based on the similarity between actions and the scores derived for the actions included in the result information.

[0047] Referring to FIG. 10, an example of a method for deriving a score using similarity will be described. In this example, it is assumed that scores for actions with action IDs "A1" to "A3" have been derived by the process shown in FIG. 9 (illustrated in the upper left part of FIG. 10). Also, it is assumed that the similarity between actions is predefined in the similarity table T4 (illustrated in the upper right part of FIG. 10).

[0048] First, the recommendation unit 34 refers to the already derived scores and identifies the action ID with the highest score.

[0049] Next, the recommendation unit 34 identifies from the similarity table T4 other action IDs with relatively high similarity to the action ID with the highest score. For example, the recommendation unit 34 identifies a predetermined number of action IDs in descending order of similarity. Here, among the other action IDs, those for which scores have already been derived using the result information are excluded.

[0050] Next, the recommendation unit 34 newly derives the score of the identified action ID using the score of the action ID with the highest score and the similarity between the action ID with the highest score and the identified action ID. The score y of the identified action ID is obtained by the following formula using the score x of the action ID with the highest score and the similarity s. y = x × s

[0051] In the example of FIG. 10, "A2" is identified as the action ID with the highest score. When identifying one other action ID with high similarity to "A2", "A4" is identified. Note that since scores for "A1" to "A3" have already been derived using the result information, they are excluded. The score of "A4" is 0.19 because the score of "A2" is 0.63 and the similarity between "A2" and "A4" is 0.3. The scores derived using such similarity can be simply compared with the scores derived using the result information.

[0052] The recommendation unit 34 updates the ranking of actions in the recommendation list extracted from the recommendation list table T2 based on the scores. Specifically, the recommendation unit 34 rearranges the ranking of each action defined in the recommendation list in descending order of score. The recommendation unit 34 may also add actions that were not included in the recommendation list before the update and that have been derived with a high score. Conversely, the recommendation unit 34 may delete actions that were included in the recommendation list before the update but have been derived with a low score or have no score derived.

[0053] The lower right side of Figure 10 shows examples of recommendation lists before and after the update. In the updated recommendation list, "A4," which was not included in the previous list, has been added. Also, "A5," which was included in the previous list, has been removed. Furthermore, "A1" through "A3," which were included in the previous recommendation list, and the newly added "A4" have been sorted in descending order of score.

[0054] The update of the recommendation list is completed as described above. The recommendation department 34 decides which actions to recommend based on the updated recommendation list.

[0055] The output unit 36 ​​outputs a recommendation statement to the user terminal 90 for the action determined by the recommendation unit 34. Specifically, the output unit 36 ​​extracts a recommendation statement from the action table T3 using the action ID of the action determined by the recommendation unit 34 and outputs it to the user terminal 90. If multiple actions are recommended, the ranking of those actions in the recommendation list may also be output.

[0056] Figure 11 shows an example of screen D1 displayed on the user terminal 90's display in response to the output of the output unit 36. Screen D1 displays three recommendation statements in a ranked order. In the example in Figure 11, the "recommended order" is the same as the order specified in the recommendation list.

[0057] (Acquisition of Results) Next, the process of acquiring results, which is performed after the recommendation, will be explained. The acquisition unit 30 acquires result information for the recommendation of an action. The acquisition unit 30 also records the result information by adding data to the result table T5 based on the acquired result information. For example, the acquisition unit 30 acquires the action selected on screen D1 in Figure 11 as result information indicating the action accepted by the user.

[0058] For example, after one action is selected on screen D1, the acquisition unit 30 queries whether or not that action had an effect on improving health, and acquires the result of the query as result information indicating the effect on improving health. Figure 12 shows an example of screen D2 displayed on the user terminal 90's display in response to a query from the acquisition unit 30. Screen D2 displays buttons for evaluating, with two options, whether or not the action "Rotate your shoulders" had an effect on improving health. The acquisition unit 30 acquires whether or not "yes" was selected on screen D2 as result information indicating the effect on improving health.

[0059] Furthermore, the effect of improving health is not limited to being shown by a binary evaluation of "yes" or "no," but may also be shown by various commonly used evaluation scales and methods. For example, evaluation may be done in multiple stages such as "yes," "somewhat yes," "average," "almost no," or "no," or on a five-star scale. In these cases, the recommendation unit 34 may set the value of effect a in the calculation of the weighted inverse rank r described above to, for example, 1, 0.8, 0.6, 0.4, 0.2 in order of highest evaluation. Alternatively, the value of effect a may be set according to whether it is above or below a predetermined standard, for example, to 1 if the evaluation is "average" or higher, and 0.5 if the evaluation is below "average." Alternatively, a quantitative evaluation may be performed, such as assigning points in increments of 1 in the range of 0 to 10 points, and effect a may be calculated based on those points. In addition, the result information is not limited to information manually entered by the user using screens D1 and D2. For example, the acquisition unit 30 may make a separate inquiry using email.

[0060] Alternatively, the acquisition unit 30 may estimate the actions performed by the user (i.e., actions accepted by the user) by detecting the user's movements using sensors such as accelerometers and detection means such as cameras. The detection means may be, for example, those provided by the user terminal 90, or an external device connected to the recommendation device 10 or the user terminal 90.

[0061] Alternatively, the acquisition unit 30 may quantitatively measure the degree of effectiveness by, for example, monitoring the user's biological information such as heart rate and body temperature, checking the results of calculation problems, or conducting medical or physiological tests such as blood tests. In this case, the acquisition unit 30 may determine the level of effectiveness by comparing it with a threshold or past data.

[0062] (Similarity) The evaluation unit 38 evaluates the similarity between actions for each combination of actions. Similarity is evaluated, for example, during setup and when the content of actions in the action table T3 is updated. The evaluation unit 38 stores the evaluated similarity in the similarity table T4.

[0063] Publicly known techniques can be applied as appropriate to evaluate similarity. For example, based on the idea that similar actions tend to result in similar recommendation letters, publicly known techniques for evaluating the similarity of natural language can be applied as appropriate.

[0064] One method for evaluating the similarity of natural language texts is to use tf-idf (term frequency-inverse document frequency). tf-idf is derived as a value corresponding to the frequency and specificity of words. For example, the evaluation unit 38 may derive a vector for each recommendation sentence that has dimensions equal to the number of words in all recommendation sentences and consists of tf-idf for each word. Alternatively, the evaluation unit 38 may derive the similarity of the vectors between recommendation sentences using, for example, cosine similarity. In this case, the evaluation unit 38 stores the derived similarity of the vectors between recommendation sentences as the similarity between actions in the similarity table T4.

[0065] Other methods include those using machine learning models for natural language processing, such as BERT (Bidirectional Encoder Representations from Transformers) and Large Language Models (LLM). For example, the evaluation unit 38 may input each recommendation into such a machine learning model to derive a vector consisting of the feature quantities of each word contained in the recommendation for each recommendation. In other words, each recommendation may be encoded. The evaluation unit 38 may also derive the similarity of the vectors between recommendation sentences using, for example, cosine similarity. In this case, the evaluation unit 38 stores the derived similarity of the vectors between recommendation sentences as the similarity between actions in the similarity table T4.

[0066] Next, the operation of the recommendation device 10 will be explained. Figure 13 is a flowchart showing the flow of the recommendation process by the recommendation device 10. The recommendation process is performed when the CPU 11 reads the recommendation program from the ROM 12 or storage 14, loads it into the RAM 13, and executes it. Note that the recommendation process is an example of the recommendation method of this disclosure.

[0067] In step S10, the CPU 11, as the acquisition unit 30, acquires user information indicating at least one of the user's attributes and status. In step S12, the CPU 11, as the extraction unit 32, extracts a recommendation list corresponding to the user information acquired in step S10. In step S14, the CPU 11, as the acquisition unit 30, determines whether the recommendation process for the target user is the second or subsequent time, or the first time.

[0068] If it is the second time or later (i.e., step S14 is Y), the process proceeds to step S16. In step S16, the CPU 11, as the acquisition unit 30, acquires a predetermined number of result information items related to the target user, in chronological order. In step S18, the CPU 11, as the recommendation unit 34, derives a score indicating the effectiveness of each action based on the result information acquired in step S16. In step S20, the CPU 11, as the recommendation unit 34, updates the ranking of actions in the recommendation list extracted in step S12 based on the score derived in step S18.

[0069] In step S22, the CPU 11, acting as a recommendation unit 34, determines the recommended action based on the recommendation list updated in step S20. In step S24, the CPU 11, acting as an output unit 36, outputs a recommendation statement for the action determined in step S22 to the user terminal 90. In step S26, the CPU 11, acting as an acquisition unit 30, acquires the action accepted by the user in accordance with the recommendation statement output in step S24, and result information showing the health improvement effect corresponding to that action. The CPU 11, also acting as an acquisition unit 30, records the acquired result information in the result table T5.

[0070] On the other hand, in the case of the first time (i.e., step S14 is N), steps S16 to S20 are skipped, and the process proceeds to step S22. In step S22, the CPU 11, as the recommendation unit 34, decides on the recommended action based on the recommendation list extracted in step S12. Steps S24 and S26 are the same as in the case of the second time and subsequent times described above. Once step S26 is completed, the recommendation process ends.

[0071] As described above, the recommendation device 10 according to this embodiment is a recommendation device that recommends actions to promote the user's health improvement, and includes an acquisition unit 30 that acquires user information indicating at least one of the user's attributes and state, and a recommendation unit 34 that recommends at least one action based on the user information. The acquisition unit 30 acquires the actions accepted by the user from among the recommended actions, and result information indicating the health improvement effect corresponding to those actions. The recommendation unit 34 determines the action to recommend based on the user information and result information. By recommending actions based on user information and result information in this way, it is possible to recommend actions that are highly accepted by the user and are effective.

[0072] Furthermore, the recommendation device 10 according to this embodiment may further include an extraction unit 32 that extracts a recommendation list corresponding to the acquired user information from a predetermined recommendation list for at least one of the user's attributes and status, in which a plurality of different actions are ranked. By extracting recommendation lists using user information in this way, it is possible to recommend actions that have a high acceptance rate with the user, even from the first time, for example.

[0073] Furthermore, in the recommendation device 10 according to this embodiment, the recommendation unit 34 may derive a score indicating the effectiveness of each action based on the result information, update the ranking of actions in the extracted recommendation list based on the score, and decide which action to recommend based on the updated recommendation list. By updating the recommendation list using the result information in this way, it is possible to recommend actions that are more effective for the target user.

[0074] Furthermore, in the recommendation device 10 according to this embodiment, the recommendation unit 34 may derive a score indicating the degree of effectiveness for each action included in the result information based on the result information. On the other hand, the recommendation unit 34 may derive a score for actions not included in the result information based on the similarity between actions and the score derived for actions included in the result information. By deriving a score using similarity in this way, it is possible to consider whether to recommend actions that have not been selected in the past, thereby preventing the recommended actions from becoming monotonous. Therefore, an improvement in the user retention rate can be expected.

[0075] In the above embodiment, the acquisition unit 30 was described as acquiring user information indicating the user's attributes and status, but it is not limited to this. The user information only needs to indicate at least one of the user's attributes and status, or it may indicate only one of them. This is because some users may refuse to input attributes or status.

[0076] If the user information indicates only one of the user's attributes or status, the method of extracting the recommendation list by the extraction unit 32 may be changed as appropriate. For example, a recommendation list corresponding only to attributes and a recommendation list corresponding only to status may be predetermined. Also, for example, a recommendation list may be predetermined for combinations in the attribute table T1 and the recommendation list table T2 where at least one of the items indicating attributes or status is left blank. Furthermore, for example, a recommendation list may be extracted in the same manner as in the above embodiment by assigning a default value to items that cannot be obtained.

[0077] Furthermore, although the above embodiment describes a configuration in which the recommendation unit 34 derives a score using similarity, it is not limited to this configuration. For example, if the score for all actions can be calculated solely from the result information, the derivation of the score using similarity may be omitted.

[0078] Furthermore, in the above embodiment, when the recommendation unit 34 derives a score using similarity, the target of score deriving was limited to a predetermined number of actions with relatively high similarity. However, this is not limited to this. For example, the recommendation unit 34 may derive a score using similarity for all actions not included in the result information.

[0079] Furthermore, in the above embodiment, it is desirable that the recommendation unit 34 derives the score each time a recommendation is made. This is because the score is derived based on the most recent result information, and therefore the score changes as the number of recommendations increases. On the other hand, the recommendation unit 34 may reduce the amount of processing by deriving the score not each time, but every predetermined number of times (for example, three times).

[0080] Furthermore, although the above embodiment described a configuration in which the output unit 36 ​​outputs a recommendation text for action to the user terminal 90, the output content is not limited to recommendation text. The output content can be anything that can recommend actions that promote the user's health improvement, and may be words indicating actions such as "shoulder rotations," "deep breathing," "stretching," and "napping." Alternatively, it may be, for example, an audio recording of the same content as the recommendation text.

[0081] Furthermore, although the above embodiment describes a configuration in which the recommendation device 10 acquires (receives) user information and result information from the user terminal 90 and outputs (transmits) the action determined based on them to the user terminal 90, the embodiment is not limited to this. For example, the user terminal 90 may include at least a part of each functional configuration and each table T1 to T5 included in the recommendation device 10.

[0082] Furthermore, the recommendation process that the CPU reads and executes in the above embodiment may be executed by various processors other than the CPU. Examples of such processors include dedicated electrical circuits, which are processors with circuit configurations specifically designed to execute particular processes, such as FPGAs (Field-Programmable Gate Arrays), PLDs (Programmable Logic Devices) whose circuit configurations can be changed after manufacturing, and ASICs (Application Specific Integrated Circuits). The recommendation process may be executed by one of these various processors, or by a combination of two or more processors of the same or different types (for example, multiple FPGAs, and a combination of a CPU and an FPGA). More specifically, the hardware structure of these various processors is an electrical circuit that combines circuit elements such as semiconductor elements.

[0083] Furthermore, although the above embodiment describes a configuration in which the recommended program is pre-stored (installed) in ROM 12 or storage 14, the invention is not limited to this. The program may be provided in a form stored on a non-transitor storage medium such as CD-ROM (Compact Disk Read Only Memory), DVD-ROM (Digital Versatile Disk Read Only Memory), and USB (Universal Serial Bus) memory. Alternatively, the program may be provided in a form that is downloaded from an external device via a network.

[0084] The following additional information is disclosed regarding the embodiments described above.

[0085] (Addendum 1) A recommendation device for recommending actions that promote a user's health improvement, comprising: a memory; and at least one processor connected to the memory, wherein the processor is configured to acquire user information indicating at least one of the user's attributes and status; recommend at least one of the actions based on the user information; acquire the actions accepted by the user from among the recommended actions, and result information indicating the health improvement effect corresponding to the actions; and determine the actions to recommend based on the user information and the result information.

[0086] (Appendix 2) A non-temporary storage medium storing a computer-executable program to perform a recommendation process for recommending actions that promote a user's health improvement, wherein the recommendation process includes: acquiring user information indicating at least one of the user's attributes and status; recommending at least one of the actions based on the user information; acquiring the actions accepted by the user from among the recommended actions, and result information indicating the health improvement effect corresponding to the actions, and in recommending the actions, the non-temporary storage medium includes determining the actions to recommend based on the user information and the result information.

[0087] 10 Recommendation device 11 CPU 12 ROM 13 RAM 14 Storage 15 Input unit 16 Display unit 17 Communication interface 19 Bus 30 Acquisition unit 32 Extraction unit 34 Recommendation unit 36 ​​Output unit 38 Evaluation unit 90 User terminal 100 Recommendation system D1, D2 Screen T1 Attribute table T2 Recommendation list table T3 Behavior table T4 Similarity table T5 Result table

Claims

1. A recommendation device for recommending actions that promote a user's health improvement, comprising: an acquisition unit that acquires user information indicating at least one of the user's attributes and state; and a recommendation unit that recommends at least one of the actions based on the user information, wherein the acquisition unit acquires the actions accepted by the user from among the recommended actions, and result information indicating the health improvement effect corresponding to the actions; and the recommendation unit determines the actions to recommend based on the user information and the result information.

2. The recommendation device according to claim 1, further comprising an extraction unit that extracts a recommendation list corresponding to the acquired user information from among a predetermined recommendation list for at least one of the user's attributes and status, in which a plurality of different actions are ranked, wherein the recommendation unit derives a score indicating the degree of effectiveness for each action based on the result information, updates the ranking of the actions in the extracted recommendation list based on the score, and determines the action to be recommended based on the updated recommendation list.

3. The recommendation device according to claim 2, wherein the recommendation unit derives a score indicating the degree of effectiveness for each action included in the result information based on the result information, and derives a score for actions not included in the result information based on the similarity between the actions and the score derived for the actions included in the result information.

4. A recommendation method for recommending actions that promote a user's health improvement, comprising: acquiring user information indicating at least one of the user's attributes and status; recommending at least one of the actions based on the user information; and acquiring the actions accepted by the user from among the recommended actions, and result information indicating the health improvement effect corresponding to those actions, wherein, in recommending the actions, a computer performs a process to determine the actions to recommend based on the user information and the result information.

Images