User Portrait Update Method, Device, Storage Medium, and Electronic Device

Abstract

An embodiment of the present application discloses a method, device, storage medium, and electronic device for updating a user profile. The method includes: processing the current behavior data of a target user in the current update cycle to obtain current profile tags, acquiring a set of tag logical relationships and the historical profile tags of the target user in the historical update cycle, and performing tag validity selection processing on the current profile tags and the historical profile tags according to the tag logical relationships in the set of tag logical relationships to obtain the valid profile tags of the target user in the current update cycle. Finally, the user profile of the target user is updated according to the valid profile tags. In the embodiment of the present application, the valid profile tags in the user profile consider both the historical profile tags and the current profile tags corresponding to the current behavior data, and also consider the tag logical relationships between the profile tags, making the determination of the valid profile tags more accurate and effectively improving the accuracy of the user profile.

Description

Technical Field

[0001] This application relates to the field of robot technology, and particularly to a method, device, storage medium and electronic device for updating user profiles. Background Art

[0002] A user profile refers to a set of tags (i.e., User Profiles) generated by a service platform based on data such as each user's behavior or opinions in products and services, used to describe user characteristics / intentions. A tag is a series of words that are condensed, refined, and carry specific meanings, used to describe the attribute characteristics of real users themselves, facilitating enterprises to conduct data statistical analysis. Common tags are divided into two major categories: relatively static tags (static tags) and changing tags (dynamic tags). Static tags are tags that are not easily changed, such as gender; dynamic tags are predicted based on user behavior, etc.

[0003] Since user behavior, etc. will change over time, this will lead to inaccurate dynamic tags, and thus inaccurate user profiles. For enterprises, a relatively accurate user profile is necessary, which can help enterprises quickly find accurate user groups and user needs to achieve customized user services. Therefore, how to improve the accuracy of user profiles is a technical problem that needs to be solved. Summary of the Invention

[0004] Embodiments of this application provide a method, device, storage medium and electronic device for updating user profiles, which can update user profiles and improve the accuracy of user profiles.

[0005] Embodiments of this application provide a method for updating user profiles, including:

[0006] Based on the correspondence between the behavior data in the target application scenario and each profile tag in the profile tag set, process the current behavior data of the target user in the current update cycle to obtain the current profile tag;

[0007] Obtain the tag logical relationship set and the historical profile tags of the target user in the historical update cycle, where the tag logical relationship set includes the tag logical relationships between the profile tags in the profile tag set;

[0008] According to the tag logical relationships in the tag logical relationship set, perform tag validity selection processing on the current profile tag and the historical profile tags to obtain the valid profile tags of the target user in the current update cycle;

[0009] Update the user profile of the target user according to the valid profile tags.

[0010] The embodiment of the present application further provides a user portrait updating device, including:

[0011] A current tag determination module, configured to process the current behavior data of a target user in the current update cycle based on the correspondence between the behavior data in the target application scenario and each portrait tag in the portrait tag set, so as to obtain a current portrait tag;

[0012] A first acquisition module, configured to acquire a tag logic relationship set and the historical portrait tags of the target user in the historical update cycle, where the tag logic relationship set includes the tag logic relationships between the portrait tags in the portrait tag set;

[0013] A portrait tag determination module, configured to perform tag validity selection processing on the current portrait tag and the historical portrait tags according to the tag logic relationships in the tag logic relationship set, so as to obtain valid portrait tags of the target user in the current update cycle;

[0014] An update module, configured to update the user portrait of the target user according to the valid portrait tags.

[0015] The embodiment of the present application further provides a computer-readable storage medium, which stores a computer program, and the computer program is suitable for being loaded by a processor to execute the steps in the user portrait updating method described in any of the above embodiments.

[0016] The embodiment of the present application further provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the steps in the user portrait updating method described in any of the above embodiments by calling the computer program stored in the memory.

[0017] The user portrait updating method, device, computer-readable storage medium, and electronic device provided by the embodiments of the present application obtain current portrait tags by processing the current behavior data of a target user in the current update cycle, acquire a set of tag logical relationships and the historical portrait tags of the target user in the historical update cycle, perform tag validity selection processing on the current portrait tags and the historical portrait tags according to the tag logical relationships in the set of tag logical relationships, obtain the valid portrait tags of the target user in the current update cycle, and update the user portrait of the target user according to the valid portrait tags. In the embodiments of the present application, by introducing a set of tag logical relationships and performing tag validity selection processing on the current portrait tags and the historical portrait tags according to the tag logical relationships in the set of tag logical relationships to determine the valid portrait tags of the target user, the determination of the valid portrait tags in the user portrait takes into account both the historical portrait tags and the current portrait tags corresponding to the current behavior data, and also considers the tag logical relationships between the portrait tags, making the determination of the valid portrait tags more accurate and effectively improving the accuracy of the user portrait. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those skilled in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0019] Figure 1 FIG. is a schematic diagram of an application scenario of the user portrait updating method provided by the embodiments of the present application.

[0020] Figure 2 FIG. is a schematic flowchart of the user portrait updating method provided by the embodiments of the present application.

[0021] Figure 3 FIG. is a schematic flowchart of determining valid portrait tags provided by the embodiments of the present application.

[0022] Figure 4 FIG. is another schematic flowchart of the user portrait updating method provided by the embodiments of the present application.

[0023] Figure 5 FIG. is a schematic diagram of the correspondence between the recommended level tags and the distance intervals provided by the embodiments of the present application.

[0024] Figure 6 FIG. is a schematic structural diagram of the user portrait updating device provided by the embodiments of the present application.

[0025] Figure 7 FIG. is another schematic structural diagram of the user portrait updating device provided by the embodiments of the present application.

[0026] Figure 8 A schematic structural diagram of an electronic device provided by an embodiment of the present application. Detailed implementation manners

[0027] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0028] The embodiments of the present application provide a user portrait update method, device, computer-readable storage medium, and electronic device. Specifically, the user portrait update method in the embodiments of the present application can be executed by an electronic device. Among them, the electronic device can be a terminal or a server, etc. The terminal can be a smart phone, a tablet computer, a laptop computer, a touch screen, a robot, a personal computer (PC), etc. The server can be an independent physical server, a server cluster composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services and cloud databases.

[0029] The user portrait update method in the embodiments of the present application can be applied to any scenario for updating the user portrait (that is, the target application scenario in the present application). For example, the target application scenario can be an activity promotion scenario, an outbound marketing scenario, etc.; specifically, such as a game promotion scenario, a training course intention collection scenario, a financial field activity promotion scenario (such as a new card opening promotion scenario), an outbound sales scenario for the operator field, a (5G) traffic package outbound marketing scenario, etc. In these scenarios, it is necessary to collect information and construct user portraits for the users who have been promoted or called outbound to assist enterprises in doing targeted promotion or outbound call data analysis.

[0030] To facilitate the understanding of the technical solutions in the embodiments of the present application, the following takes the (5G) traffic package outbound marketing scenario as an example for illustration. The implementation principles of other scenarios are similar, and the only difference is the specific content of the tags and behavior data.

[0031] Such as Figure 1As shown in the figure, it is a schematic diagram of an application scenario of the user portrait update method provided by an embodiment of the present application. In this application scenario, there are an electronic device 1 and an electronic device 2, and the electronic device 1 and the electronic device 2 can communicate with each other. Among them, the electronic device 1 is a device for implementing outbound calls, such as an outbound call robot, an outbound call terminal, etc. The electronic device 1 can make outbound calls to the user terminal through a landline number, a mobile phone number, or other means, and the electronic device 1 can collect outbound interaction information such as call interaction corpus data during the outbound call process. The electronic device 2 can be a server device of the electronic device 1. This server device is used to collect the data collected in all electronic devices 1, such as outbound interaction information, user information, and other data, and can also have other service functions. The user portrait update method provided by the embodiment of the present application can run independently in the electronic device 1, can also run independently in the electronic device 2, or can run partially in the electronic device 1 and partially in the electronic device 2. The embodiment of the present application will be described by taking the user portrait update method running independently in the electronic device 1 as an example.

[0032] A user portrait update method, device, computer-readable storage medium, and electronic device provided by an embodiment of the present application will be described in detail below. It should be noted that the serial numbers of the following embodiments do not limit the preferred order of the embodiments.

[0033] Figure 2 is a schematic flowchart of a user portrait update method provided by an embodiment of the present application. Please refer to Figure 2 and the user portrait update method includes the following steps:

[0034] Step 101: Based on the correspondence between the behavior data in the target application scenario and each portrait label in the portrait label set, process the current behavior data of the target user in the current update cycle to obtain the current portrait label.

[0035] Among them, the target application scenario is described by taking the (5G) traffic package outbound marketing scenario as an example. The behavior data in the target application scenario includes outbound interaction information in the outbound marketing scenario, such as outbound interaction corpus data. The portrait label set is a set of portrait labels of all current users in the target application scenario. For example, the portrait labels in the (5G) traffic package outbound marketing scenario include: not using much traffic, upgrading 5G traffic is useless, not having a 5G mobile phone, wanting to use a mobile card, already having a mobile card, the user refuses without reason, the user agrees, no portrait label, traffic is sufficient, etc. In other application scenarios, the behavior data can also be the survey results selected in the questionnaire, etc. The corresponding behavior data in different application scenarios may be different, and the corresponding portrait label sets are also different.

[0036] For the (5G) traffic package outbound marketing scenario, the current behavior data can be determined according to the number of outbound calls, or according to time. It can be the outbound interaction information corresponding to this time, or the outbound interaction information within half an hour / the outbound interaction information of the current day, etc. Therefore, the corresponding current update cycle can be in units of times, in units of time, or in other units, etc. In the embodiments of the present application, it is described in units of times.

[0037] In the target application scenario, there is a corresponding relationship between the behavior data and each portrait label in the portrait label set, and this corresponding relationship can be calculated according to the behavior analysis algorithm. The behavior analysis algorithm is used to process the current behavior data of the target user in the current update cycle to obtain the current portrait label.

[0038] In one embodiment, the step of processing the current behavior data of the target user in the current update cycle to obtain the current portrait label includes: obtaining the current behavior data of the target user in the current update cycle in the target application scenario; performing behavior analysis on the current behavior data to obtain at least one candidate portrait label; when there is one candidate portrait label, directly use the candidate portrait label as the current portrait label corresponding to the current behavior data of the target user; when the obtained candidate portrait labels include multiple ones, and the label logical relationship among the multiple candidate portrait labels includes an inclusion relationship, delete the included portrait label in the inclusion relationship, and use the remaining candidate portrait labels as the current portrait label corresponding to the current behavior data of the target user.

[0039] Obtain the outbound interaction corpus data of the target user in the (5G) traffic package outbound marketing scenario, such as the outbound interaction text corresponding to the outbound interaction record, etc., as Figure 3 shown, obtain the outbound interaction records of different target users; then perform corpus behavior analysis on the outbound interaction corpus data, and the corpus behavior analysis can be performed on the interaction corpus data through an algorithm model to obtain at least one candidate portrait label.

[0040] Through observation and data analysis, there is an obvious phenomenon of portrait labels, that is, for the same target user, the same portrait label will not be expressed in the next outbound interaction of the target user. For example: the target user whose portrait label obtained after the previous outbound interaction is "not using much traffic" may not disclose any information in the next outbound interaction, that is, there is no portrait label, or there may be new portrait labels (not interested in the 5G traffic package, the surrounding environment does not support 5G, etc.). Different portrait labels have different meanings, and when updating portrait labels, it cannot be simply directly deleted or added, otherwise it will lead to inaccurate portrait labels.

[0041] At the same time, in a single outbound call interaction, the target user may also have completely inconsistent outbound call interaction corpora before and after, or similar outbound call interaction corpora, or duplicate outbound call interaction corpora. After using the algorithm model to perform corpus behavior analysis on the interaction corpus data, among the multiple candidate portrait labels obtained, there may be duplicate portrait labels, or one portrait label contains another portrait label. In these cases, it will also have a certain impact on the update of the portrait labels.

[0042] Therefore, in the embodiments of the present application, the label logical relationship between portrait labels is defined, where the label logical relationship includes the same relationship, the inclusion relationship, the included relationship, the parallel relationship, and the conflict relationship. Among them:

[0043] The same relationship means that the essential contents such as the text content / semantic content of two portrait labels are completely the same.

[0044] For the inclusion relationship / included relationship, the inclusion relationship and the included relationship between two portrait labels (the first portrait label and the second portrait label) exist simultaneously. The first portrait label is included by the second portrait label. Correspondingly, the second portrait label includes the first portrait label.

[0045] The parallel relationship is used to describe similar portrait labels. Two similar portrait labels can appear simultaneously, and the union is directly taken.

[0046] The conflict relationship refers to two portrait labels that definitely cannot appear on a target user at the same time. Take the latest non-conflicting portrait label.

[0047] For example, if the current portrait label corresponding to the current outbound call interaction is included by the historical portrait label corresponding to the historical outbound call interaction (such as the previous portrait label corresponding to the previous outbound call interaction), then take the historical portrait label corresponding to the historical outbound call interaction (such as the previous portrait label corresponding to the previous outbound call interaction). If the current portrait label corresponding to the current outbound call interaction includes the historical portrait label corresponding to the historical outbound call interaction (such as the previous portrait label corresponding to the previous outbound call interaction), then take the current portrait label corresponding to the current outbound call interaction. Among them, by default, the empty portrait label is included by all portrait labels, and the non-reason refusal is included by the reason-based refusal.

[0048] For example, assume there are portrait label A (the user refuses without reason) and portrait label B (not using much traffic), then Then there is an inclusion relationship and an included relationship between portrait label A and portrait label B. Portrait label A is included by portrait label B, and portrait label B includes portrait label A. In this label logical relationship, portrait label A is the included portrait label and portrait label B is the inclusion label.

[0049] For example, if the current portrait label corresponding to the current outbound call interaction is A and the previous portrait label corresponding to the previous outbound call interaction is B, then the portrait label B needs to be retained; if the current portrait label corresponding to the current outbound call interaction is B and the previous portrait label corresponding to the previous outbound call interaction is A, then the portrait label A needs to be deleted.

[0050] As shown in Table 1, the same relationship, inclusion relationship, juxtaposition relationship, and conflict relationship between portrait labels are illustrated by taking the first portrait label and the second portrait label as examples in Table 1.

[0051] Table 1 Label logical relationships between portrait labels

[0052]

[0053] It should be noted that when there are multiple obtained candidate portrait labels and the label logical relationships among the multiple candidate portrait labels include an inclusion relationship, delete the included portrait labels in the inclusion relationship, and use the remaining candidate portrait labels as the current portrait labels corresponding to the current behavior data of the target user. In this way, by deleting the included portrait labels in the inclusion relationship, the obtained current portrait labels can retain both truly useful portrait labels and avoid the influence of useless portrait labels on the user portrait.

[0054] In addition, when there are multiple obtained candidate portrait labels, only delete the included portrait labels in the inclusion relationship. If other label logical relationships are included among the multiple obtained candidate portrait labels, retain all the obtained multiple candidate portrait labels as the current portrait labels.

[0055] Such as Figure 3 shown, the current portrait labels obtained for user identifier 1111 are: no demand - sufficient traffic; the current portrait labels obtained for user identifier 1112 are: considering and hesitating.

[0056] In this step, process the current behavior data of the target user in the current update cycle to obtain the current portrait labels.

[0057] Step 102, obtain the label logical relationship set and the historical portrait labels of the target user in the historical update cycle, where the label logical relationship set includes the label logical relationships between each portrait label in the portrait label set.

[0058] Among them, if the current update period is in terms of the number of times, such as once, correspondingly, the current behavior data is the outbound call interaction information corresponding to this time. The historical update period includes all the times before this time, or all the times within a preset time period before this time, such as all the times within two days before this time. The historical portrait tags in the historical update period correspond to the portrait tags after the last update of the target user, or the portrait tags updated by the target user using the outbound call interaction information within the preset time period before this time.

[0059] If the current update period is in terms of time, such as half a day, correspondingly, the current behavior data is the outbound call interaction information of the target user within half a day. The historical update period includes all the time before half a day, or within the preset time before half a day, such as within two days before half a day; the historical portrait tags correspond to the portrait tags updated by the target user using all the outbound call interaction information before half a day, or the portrait tags updated by the target user using the outbound call interaction information corresponding to within two days before half a day.

[0060] In the embodiments of the present application, taking the current update period as this time, the current portrait tag as the portrait tag corresponding to this outbound call interaction, the historical update period as all the times before this time, and the historical portrait tag as the portrait tag corresponding to all the outbound call interactions before this time as an example for illustration.

[0061] It should be noted that although the outbound call interaction records and corresponding outbound call interaction information of the target user each time will be saved, the portrait tag corresponding to a target user is unique, that is, the finally determined portrait tag (the effective portrait tag mentioned later in the text).

[0062] The label logic relationship set includes the label logic relationships between the portrait tags in the portrait tag set, and the label logic relationships correspond to logic relationship weights. Different label logic relationships correspond to different logic relationship weights.

[0063] For example, the logic relationship weight corresponding to the parallel relationship is 1, the logic relationship weight corresponding to the conflict relationship is -1, and the logic relationship weight corresponding to the same relationship is also 1; for the inclusion relationship, such as portrait tag A (the user refuses without reason) and portrait tag B (not using much traffic), then Define the logic relationship weight T(AB) = 1, where the current portrait tag is A, the historical portrait tag is B, and the historical portrait tag B needs to be retained; define the logic relationship weight T(BA) = 0, where the current portrait tag is B, the historical portrait tag is A, and the current portrait tag A needs to be deleted.

[0064] Among them, the label logic relationship set can be represented by a matrix, or by a vector, or in other appropriate ways.

[0065] In one embodiment, the set of label logical relationships includes a label transition matrix, and the label transition matrix includes matrix elements for characterizing the label logical relationships between portrait labels.

[0066] In one embodiment, before obtaining the set of label logical relationships, it further includes: constructing a label transition matrix according to all portrait labels in the portrait label set and the label logical relationships between each portrait label in all portrait labels.

[0067] In one embodiment, the step of constructing a label transition matrix according to all portrait labels in the portrait label set and the label logical relationships between each portrait label in all portrait labels includes: obtaining the number of labels N of all portrait labels in the portrait label set, where N is a positive integer greater than zero; constructing an initial label matrix with dimensions of N * N, where each dimension in the initial label matrix corresponds to a portrait label; filling the initial label matrix according to the label logical relationships between each portrait label in all portrait labels to obtain the label transition matrix.

[0068] For example, if the number of labels of all portrait labels in the portrait label set is 3, then an initial label matrix with dimensions of 3 * 3 is constructed, that is, the number of rows and columns of the initial label matrix is the same as the number of labels, and each dimension where the row is located corresponds to a portrait label, and each dimension where the column is located corresponds to a portrait label.

[0069] After obtaining the initial label matrix, fill the initial label matrix according to the label logical relationships between each portrait label in all portrait labels to obtain the label transition matrix.

[0070] In one embodiment, the step of filling the initial label matrix according to the label logical relationships between each portrait label in all portrait labels to obtain the label transition matrix includes: obtaining the logical relationship weights corresponding to the label logical relationships between each portrait label in all portrait labels; filling the initial label matrix according to the logical relationship weights to obtain the label transition matrix. In this embodiment, the corresponding label logical relationships are represented by logical relationship weights, and the label logical relationships are quantified for easy calculation.

[0071] For example, assume that the number of tags for all image tags in the image tag set is 3. Among them, image tag D = wants to use a mobile card, image tag E = doesn't use much data, and image tag F = already has a mobile card. At this time, image tag D and image tag E are in a parallel relationship, and the corresponding logical relationship weight is 1. Or it can also be understood that 1 is used to represent the parallel relationship between image tag D and image tag E; image tag D and image tag F are in a conflict relationship, and the corresponding logical relationship weight is -1. Or it is understood that -1 is used to represent the conflict relationship between image tag D and image tag F; image tag E and image tag F are in a parallel relationship, and the corresponding logical relationship weight is 1. Or it is understood that 1 is used to represent the parallel relationship between image tag E and image tag F. After filling the initial tag matrix according to the logical relationship weights, the obtained tag transition matrix is shown in Table 2. It should be noted that the table headers and other information in Table 2 are for easy understanding.

[0072] Table 2 Example 1 of Tag Transition Matrix

[0073]

[0074] For example, assume that the number of tags for all image tags in the image tag set is 3. Among them, image tag A = the user refuses without reason, image tag B = doesn't use much data, and image tag C = upgrading the 5G data package is useless, and doesn't have a 5G mobile phone. Among them, The logical relationship weight T(AB) = 1, the logical relationship weight T(BA) = 0, the logical relationship weight T(AC) = 1, and the logical relationship weight T(CA) = 0. Image tag B and image tag C are in a parallel relationship, and the corresponding logical relationship weight is 1. After filling the initial tag matrix according to the logical relationship weights, the obtained tag transition matrix is shown in Table 3.

[0075] Table 3 Example 2 of Tag Transition Matrix

[0076]

[0077] It should be noted that the above logical relationship weights are only for illustrative purposes. In other embodiments, other logical relationship weights can also be used, etc.

[0078] It can be understood that when the set of tag logical relationships includes a tag transition matrix, the tag transition matrix can be constructed before obtaining the tag transition matrix, that is, the tag transition matrix is constructed in advance. In this way, obtaining the tag transition matrix means obtaining the pre-constructed tag transition matrix. Since the tag transition matrix includes the logical relationship weights between the portrait tags in the portrait tag set, the tag transition matrix can be used when determining the valid portrait tags of each target user, improving the applicability of the tag transition matrix; and constructing the tag transition matrix in advance, when specifically determining the valid portrait tags of each target user, there is no need to construct the tag transition matrix and the tag transition matrix can be directly used, improving the determination speed of the valid portrait tags.

[0079] When a new portrait tag is added to the portrait tag set of the target application scenario, the tag transition matrix can be updated according to the tag logical relationship (the corresponding logical relationship weight) between the newly added portrait tag and other portrait tags. For example, when there is 1 newly added portrait tag, the tag transition matrix needs to add a row and a column of data.

[0080] Step 103: According to the tag logical relationships in the set of tag logical relationships, perform tag validity selection processing on the current portrait tag and the historical portrait tag to obtain the valid portrait tags of the target user in the current update cycle.

[0081] Among them, the valid portrait tags can also be understood as the final portrait tags of the target user. According to the tag logical relationships in the set of tag logical relationships, perform tag validity selection processing on the current portrait tag and the historical portrait tag to obtain the final portrait tags of the target user in the current update cycle. That is, the valid portrait tags of the target user are determined according to the logical relationship weights corresponding to the tag logical relationships. In this way, the determination of the valid portrait tags of the target user not only considers the current portrait tag, but also considers the historical portrait tag, and at the same time considers the tag logical relationship between the current portrait tag and the historical portrait tag, improving the accuracy of the determination of the valid portrait tags and further improving the accuracy of the user portrait.

[0082] In one embodiment, step 103 includes: according to the tag logical relationships in the set of tag logical relationships, find the target tag logical relationships between the current portrait tag and the historical portrait tag to obtain a set of target tag logical relationships; according to the set of target tag logical relationships, perform tag validity selection processing on the current portrait tag and the historical portrait tag to obtain the valid portrait tags of the target user in the current update cycle.

[0083] Among them, find the logical relationship of the target tags corresponding to the current portrait tag and the historical portrait tag in the tag transition matrix, and use all the logical relationships of the target tags as the target tag logical relationship set. Specifically, use the logical relationship weight corresponding to the overlapping position of the current portrait tag and the historical portrait tag in the tag transition matrix as the target tag logical relationship set (corresponding logical relationship weight). According to the target tag logical relationship set (corresponding logical relationship weight), perform tag validity selection processing on the current portrait tag and the historical portrait tag to obtain the valid portrait tags of the target user in the current update cycle.

[0084] In one embodiment, the step of performing tag validity selection processing on the current portrait tag and the historical portrait tag according to the target tag logical relationship set to obtain the valid portrait tags of the target user in the current update cycle includes: performing first validity selection processing on the historical portrait tag according to the target tag logical relationship set to determine the historical impact portrait tags; performing second validity selection processing on the historical impact portrait tags and the current portrait tag to obtain the valid portrait tags of the target user in the current update cycle.

[0085] Among them, use the target tag logical relationship set (corresponding logical relationship weight) to perform first validity selection processing on the historical portrait tag to determine the historical impact portrait tags. In one embodiment, multiply the target tag logical relationship set by the set corresponding to the historical portrait tag to obtain the historical impact portrait tags. In other embodiments, the first validity selection processing can also be performed in other ways.

[0086] After obtaining the historical impact portrait tags, perform second validity selection processing according to the historical impact portrait tags and the current portrait tag to obtain the valid portrait tags. In one embodiment, perform union processing on the historical impact portrait tags and the current portrait tag, and use the portrait tags obtained after the union processing as the valid portrait tags of the target user in the current update cycle, and use the coefficient of the portrait tags obtained after the union processing as the weight coefficient of the corresponding portrait tags in the valid portrait tags. This weight coefficient can be used to represent the importance degree of the corresponding portrait tags. In other embodiments, the second validity selection processing can also be performed in other ways.

[0087] In one embodiment, after the step of obtaining the valid portrait tags of the target user within the current update cycle, the user portrait update method further includes: obtaining the weight coefficients of each portrait tag in the valid portrait tags; determining the importance degree of each portrait tag according to the weight coefficients, and selecting portrait tags from the valid portrait tags as the main portrait tags within the current update cycle according to the importance degree. For example, the portrait tag with the highest importance degree is selected from the valid portrait tags as the main portrait tag of the target user within the current update cycle. In this embodiment, the weight coefficients of each portrait tag in the valid portrait tags can be determined according to the weight coefficients, and the importance degree can be determined according to the weight coefficients. It is possible to output or display each portrait tag in the valid portrait tags according to the high or low importance degree, and the main portrait tag can also be determined according to the importance degree, etc.

[0088] In one embodiment, if the valid portrait tag is defined as now’, then now’ can be calculated according to the following operation method:

[0089] now’ = gen-eye(now) * now T + T_n_o * old T

[0090] where now represents the current portrait tag, and now T represents the transpose of the current portrait tag now, gen-eye(now) represents the generation of the identity matrix of the current portrait tag, old represents the historical portrait tag, and old T represents the transpose of the historical portrait tag, T_n_o represents the target tag logical relationship set, and the result of T_n_o * old T represents the historical influence portrait tag, and gen-eye(now) * now T represents the corresponding current portrait tag. In this embodiment, the matrix method is used to determine the valid portrait tag, which can greatly improve the speed of portrait output determination.

[0091] It should be noted that other operation methods can also be used to determine the valid output tag.

[0092] Taking the tag transition matrix shown in Table 2 as an example, assume that the current portrait tag is now = [D, E] = [want to use a mobile card, not use much traffic], and the historical portrait tag is old = [E, F] = [not use much traffic, already have a mobile card]. Search for the logical relationship weights corresponding to the overlapping positions of the current portrait tag (the row where portrait tag D is located and the row where portrait tag E is located) and the historical portrait tag (the column where portrait tag E is located, the column where portrait tag F is located) in Table 2, and use this logical relationship weight as the target tag logical relationship set (the corresponding logical relationship weight). Correspondingly, the target tag logical relationship set (the corresponding logical relationship weight) is shown in Table 4.

[0093] Table 4 Example 1 of the target label logical relationship set

[0094] 1 -1 1 1

[0095] Perform the first validity selection process on the historical portrait label old = [E, F] according to the target label logical relationship set, that is, multiply the target label logical relationship set by the transpose of the historical portrait label old, T_n_o * old T , and the obtained historical influence portrait label is 2E.

[0096] For the current portrait label now = [D, E], generate gen-eye(now), as shown in Table 5. The table header of Table 5 is for easy understanding.

[0097] Table 5 Example of gen-eye(now)

[0098] Current Image Label Image Label D Image Label E Image Label D 1 0 Image Label E 0 1

[0099] Use gen-eye(now) * now T to obtain the current portrait label A + B.

[0100] Then perform the union process according to the current portrait label and the historical influence portrait label to obtain now’ = D + E + 2E = D + 3E.

[0101] In this way, the effective portrait labels obtained are D (want to use a mobile card), E (not using much traffic). Among them, the weight coefficient of the portrait label D is 1, and the weight coefficient of the portrait label E is 3, which means that the portrait label E appears more frequently, and obviously has a greater influence on the target user. Taking the portrait label E as the main portrait label of the target user in the current update cycle, the effective portrait labels can be stored or displayed in the order of importance, for example, stored as portrait label E (not using much traffic), portrait label D (want to use a mobile card).

[0102] Taking the label transition matrix shown in Table 2 as an example, assume that the current label is now = [D] = [want to use a mobile card], and the historical labels are old = [E, F] = [not using much data, already have a mobile card]. Look up the logical relationship weights corresponding to the overlapping positions of the current portrait label (the row where portrait label D is located) and the historical portrait labels (the column where portrait label E is located, the column where portrait label F is located) in Table 2 to obtain the target label logical relationship set (the corresponding logical relationship weights) [1, -1]. Perform the first validity selection process on the historical portrait label old = [E, F] according to the target label logical relationship set [1, -1], and the obtained historical impact portrait label is E - F. And the current portrait label is D. In this way, the obtained now’ = D + E - F. Since -F indicates that portrait label D and portrait label F are in a conflict relationship, the final valid portrait labels are portrait label D (want to use a mobile card) and portrait label E (not using much data). Since the weights are the same, the corresponding importance levels are the same.

[0103] Taking the label transition matrix shown in Table 3 as an example for illustration, assume that the current portrait label is now = [A] = [user refuses without reason], and the historical portrait labels are old = [B, C] = [not using much data, upgrading the 5G data package is useless without a 5G mobile phone]. Look up the logical relationship weights corresponding to the overlapping positions of the current portrait label (the row where portrait label A is located) and the historical portrait labels (the column where portrait label B is located, the column where portrait label C is located) in Table 3 to obtain the target logical relationship set (the corresponding logical relationship weights) [1, 1]. Perform the first validity selection process on the historical portrait label old = [B, C] according to the target label logical relationship set [1, 1], and the obtained historical impact portrait label is B + C. And the current portrait label is portrait label A. In this way, now’ = A + B + C, that is, the valid portrait labels are portrait label A, portrait label B, and portrait label C.

[0104] It should be noted that the above portrait labels A, B, C, D, E, and F are just for illustrative purposes and do not constitute a limitation.

[0105] As Figure 3 shown, the valid portrait labels obtained for user identifier 1111 are: no demand - sufficient data; the valid portrait labels obtained for user identifier 1112 are: considering and hesitating.

[0106] The following takes the outbound marketing scenario of 5G data packages as an example for illustration. Among them, taking the current behavior data as the outbound interaction information for this time (including the outbound interaction record for this time), the historical behavior data as the outbound interaction information for the last time (including the outbound interaction record for the last time), the current portrait label as the portrait label for this time, and the historical portrait label as the portrait label for the last time as an example, the effective portrait labels finally determined according to the current portrait label and the historical portrait label are shown in Table 6.

[0107] Table 6 Example of Outbound Marketing Scenario of 5G Data Packages

[0108]

[0109]

[0110] Step 104, update the user portrait of the target user according to the effective portrait label.

[0111] After obtaining the effective portrait label, update the user portrait of the target user.

[0112] In one embodiment, when the effective portrait label of the target user is different from the portrait label (historical portrait label) in the user portrait, use the effective portrait label to update the portrait label in the user portrait; when the effective portrait label of the target user is the same as the portrait label in the user portrait, the portrait label in the user portrait can be updated using the effective portrait label, or no processing can be performed on the portrait label in the user portrait.

[0113] This embodiment determines the effective portrait label of the target user by introducing a set of label logical relationships and performing label validity selection processing on the current portrait label and the historical portrait label according to the label logical relationships in the set of label logical relationships, so that the determination of the effective portrait label in the user portrait takes into account both the historical portrait label and the current portrait label corresponding to the current behavior data, and also takes into account the label logical relationships between each portrait label, making the determination of the effective portrait label more accurate and effectively improving the accuracy of the user portrait.

[0114] Based on the above user portrait update method, Figure 4 Another process schematic diagram of the user portrait update method provided by the embodiment of the present application is shown as Figure 4 shown, and the user portrait update method corresponding to this process schematic diagram includes the following steps:

[0115] Step 201, obtain the basic information of multiple users in the target application scenario, the business information of each user in the target application scenario, and the response information of each user to the business information.

[0116] The user basic information includes user age, gender, place of origin, etc.

[0117] The business information of the user in the target application scenario includes the promotion package information in the outbound marketing scenario, the current behavior data in the current update cycle, and the effective portrait tags in the corresponding user portrait, etc. The portrait tags can be the effective portrait tags obtained by the method described in any of the above embodiments. Among them, the current behavior data includes outbound interaction information, for example, interaction duration, interaction discourse, outbound interaction text, user interaction path, hang-up details, and other information.

[0118] The response information of the user to the business information represents the ordering / subscription information of the user for the promotion package information in the outbound marketing scenario, including: ordered / subscribed, not ordered / not subscribed, etc.

[0119] The user's basic information, business information, and response information can be as shown in Table 6. It should be noted that Table 6 is only an example and does not constitute a limitation.

[0120] Table 6 Example of User's Basic Information, Business Information, and Response Information

[0121]

[0122]

[0123] It should be noted that according to different application scenarios, the corresponding basic information, business information, and response information will also be different.

[0124] Step 202: Encode the basic information, business information, and response information of each user to construct the feature data corresponding to each user, and use the feature data corresponding to each user as each sample to obtain a sample set.

[0125] In one embodiment, the encoding of the basic information, business information, and response information of each user to construct the feature data corresponding to each user includes: encoding the digital information and enumerable text information in the basic information, business information, and response information of each user using the one-hot encoding method; encoding the non-enumerable (non-enumerable) text information in the basic information, business information, and response information of each user using the text vector method; and using the encoded information of each user as the feature data corresponding to each user.

[0126] For example, the interaction duration, interaction rounds, exceeding the charging standard, traffic, etc. in Table 6 correspond to digital information; gender, age, region, promotion package information, portrait tags, user paths, hang-up details, ordering / subscription information, etc. correspond to enumerable text information; and the outbound interaction text corresponds to non-enumerable text information.

[0127] For digital information and enumerable text information, one-hot encoding is used for encoding. For example, for gender, there are "male" and "female". Using one-hot encoding will construct a 1*2 matrix, using [0,1] to represent the feature data corresponding to the gender "male", and [1,0] to represent the feature data corresponding to the gender "female"; for portrait tags, assuming the number of portrait tags in the portrait tag set is 6, using one-hot encoding will construct a 1*6 matrix, using [1,0,0,0,0,0] to represent the feature data of the portrait tag corresponding to the current user, and the portrait tag corresponding to the current user is the first portrait tag in the portrait tag set. Using [1,0,0,1,1,0] to represent the feature data of the portrait tag corresponding to the current user, and the portrait tag corresponding to the current user includes the first portrait tag, the fourth portrait tag, and the fifth portrait tag in the portrait tag set.

[0128] For non-enumerable text information, text vector encoding is used. For example, for outbound interaction text, term frequency–inverse document frequency (TF-IDF) can be used for encoding. Assuming there are 5000 words in the text library, then a 1*5000 matrix is set correspondingly, and then the probability of each word in the current outbound interaction text is calculated using the TF-IDF method, such as 0.9, etc. Then, the probability of each word in the current outbound interaction text is filled in the corresponding position of the matrix to obtain the feature data corresponding to the outbound interaction text.

[0129] For non-enumerable text information, word vectors and other methods can also be used for encoding.

[0130] The above encoding method is just one way to obtain the corresponding feature data, and any other method can be used to obtain the corresponding feature data.

[0131] The feature data of each user is obtained, and each user is used as a sample. In this way, a sample set is obtained.

[0132] Step 203, obtain the feature data of the target sample with the response information of subscribed from the sample set, and determine the recommended level label of the user portrait of the user corresponding to the to-be-classified sample with the response information of unsubscribed in the sample set according to the feature data of the target sample. This recommended level label is used to recommend the corresponding service to the user corresponding to the to-be-classified sample.

[0133] For example, obtain the feature data of the target sample with the response information of subscribed / ordered from the sample set, and use the feature data of the target sample as a seed to determine the recommended level label of the unsubscribed users.

[0134] Among them, the recommended level label is used to represent the label of the level for recommending the corresponding service to unsubscribed / undeployed users in the future. The recommended level label can include four labels: high, medium, low, and none; or it can also be understood that the subscription intention level of unsubscribed / undeployed users is determined according to the corresponding feature data. The higher the subscription intention, the higher the corresponding recommended level; the lower the subscription intention, the lower the corresponding recommended level; if there is no subscription intention, the corresponding recommended level is none.

[0135] In one embodiment, step 203 includes: initializing a first prototype vector, which includes target samples. The first prototype vector is used to store multiple samples with the recommended level label being the first recommended level label; according to the feature data and the first recommended level label of the target samples in the first prototype vector, perform recommended clustering analysis on the to-be-classified samples in the sample set with the response information being unsubscribed, so as to determine the target recommended level label of the users corresponding to the to-be-classified samples, and cluster the to-be-classified samples into the sample cluster matching the target recommended level label according to the target recommended level label.

[0136] Among them, the initialized first prototype vector is a vector including target samples. The target samples are subscribed / deployed samples, and the first recommended level label corresponding to the subscribed / deployed target samples is high. If the recommended level label includes four recommended levels, the first prototype vector only corresponds to one recommended level label.

[0137] In one embodiment, the step of performing recommended clustering analysis on the to-be-classified samples in the sample set with the response information being unsubscribed according to the feature data and the first recommended level label of the target samples in the first prototype vector to determine the target recommended level label of the users corresponding to the to-be-classified samples includes: determining the distance between the feature data of the to-be-classified samples in the sample set with the response information being unsubscribed and the feature data of the target samples in the first prototype vector; dividing distance intervals according to the distance. The number of the distance intervals is the same as the number of preset multiple recommended level labels. Each distance interval determines a corresponding recommended level label according to the corresponding distance size and the first recommended level label; determining the target recommended level label of the user portrait corresponding to the to-be-classified samples according to the distance interval to which the distance belongs.

[0138] In one embodiment, the distance between the feature data of the user corresponding to the sample to be classified and the feature data of the user corresponding to the target sample is calculated, and distance intervals are divided according to the distance. Among them, the number of distance intervals is the same as the number of preset multiple recommended level labels, for example, four. The division of the distance intervals is not in equal parts. The calculated distance is divided into four distance intervals. The distance interval closest to the first prototype vector is determined as the distance interval with a high recommended level label, the distance interval second farthest from the first prototype vector is determined as the distance interval with a medium recommended level label, the distance interval third farthest from the first prototype vector is determined as the distance interval with a low recommended level label, and the distance interval farthest from the first prototype vector is determined as the distance interval with a none recommended level label. In this way, the recommended level label corresponding to the user of the unsubscribed sample to be classified can be determined according to the adjusted distance between the data.

[0139] In one embodiment, after determining the recommended level label, the samples to be classified are clustered into sample clusters matching the recommended level label according to the recommended level label. For example, the samples to be classified with a none recommended level label are clustered into one sample cluster, and the samples to be classified with a medium recommended level label are clustered into another sample cluster, etc.

[0140] As Figure 5 shown, the distance between the feature data of the sample to be classified and the feature data of the target sample is calculated, four distance intervals are divided according to the distance, such as distance interval 1, distance interval 2, distance interval 3, and distance interval 4. According to the distance interval of the distance between the feature data of the sample to be classified and the feature data of the target sample, four different recommended level labels corresponding to the distance interval are determined, such as high, medium, low, none, etc. According to the recommended level label, the corresponding samples to be classified are respectively clustered into four different sample clusters, such as sample cluster 1, sample cluster 2, sample cluster 3, and sample cluster 4.

[0141] This embodiment determines the recommended level label of the user corresponding to the sample to be classified in an unsupervised clustering manner, and can quickly determine the recommended level label of the corresponding user.

[0142] After determining the recommended level label corresponding to the unsubscribed sample to be classified, the user portrait can be updated, and corresponding services can be recommended to the user corresponding to the sample to be classified according to the recommended level label. For example, priority outbound calls can be made to users with a high recommended level label. For users with a low recommended level label, the outbound call strategy can be changed to quickly convert the users. In this way, according to the recommended level label, it is helpful for relevant personnel to conduct targeted analysis and optimization, and to guide subsequent outbound call strategies, etc., to guide user marketing, which helps the enterprise quickly find the accurate user group and user needs to achieve customized user services.

[0143] Any combination of the above technical solutions can form an alternative embodiment of the present application, which will not be elaborated herein one by one.

[0144] To facilitate better implementation of the user profile update method in the embodiments of the present application, the embodiments of the present application further provide a user profile update device.

[0145] Please refer to FIG. 6, Figure 6 which is a schematic structural diagram of the user profile update device provided by the embodiments of the present application. The user profile update device may include a current label determination module 301, a first acquisition module 302, a profile label determination module 303, and an update module 304.

[0146] Among them, the current label determination module 301 is configured to process the current behavior data of the target user in the current update cycle based on the correspondence between the behavior data in the target application scenario and each profile label in the profile label set, so as to obtain the current profile label.

[0147] In one embodiment, when the current label determination module 301 executes the step of processing the current behavior data of the target user in the current update cycle to obtain the current profile label, it specifically executes: obtaining the current behavior data of the target user in the target application scenario in the current update cycle; performing behavior analysis on the current behavior data to obtain at least one candidate profile label; when the candidate profile label is one, using the candidate profile label as the current profile label corresponding to the current behavior data of the target user; when the candidate profile labels include multiple and the label logical relationship among the multiple candidate profile labels includes an inclusion relationship, deleting the included profile label in the inclusion relationship, and using the remaining candidate profile labels as the current profile label corresponding to the current behavior data of the target user.

[0148] Among them, the first acquisition module 302 is configured to acquire a label logical relationship set and the historical profile labels of the target user in the historical update cycle, where the label logical relationship set includes the label logical relationships among the profile labels in the profile label set.

[0149] In one embodiment, the label logical relationship set includes a label transition matrix, and the label transition matrix includes matrix elements, and the matrix elements are used to represent the label logical relationships among the profile labels.

[0150] In one embodiment, as Figure 6 shown, the user profile update device further includes a construction module 305. Among them, the construction module 305 is configured to construct a label transition matrix according to all the profile labels in the profile label set and the label logical relationships among the profile labels in all the profile labels.

[0151] In one embodiment, when the construction module 305 executes the step of constructing the label transition matrix according to all the portrait labels in the portrait label set and the label logic relationships between the portrait labels in all the portrait labels, it specifically executes: obtaining the number of labels N of all the portrait labels in the portrait label set, where N is a positive integer greater than zero; constructing an initial label matrix with a dimension of N*N, where each dimension in the initial label matrix corresponds to a portrait label; and filling the initial label matrix according to the label logic relationships between the portrait labels in all the portrait labels to obtain the label transition matrix.

[0152] In one embodiment, when the construction module 305 executes the step of filling the initial label matrix according to the label logic relationships between the portrait labels in all the portrait labels to obtain the label transition matrix, it specifically executes: obtaining the logical relationship weights corresponding to the label logic relationships between the portrait labels in all the portrait labels; and filling the initial label matrix according to the logical relationship weights to obtain the label transition matrix.

[0153] In one embodiment, the label logic relationships include inclusion relationship, being included relationship, parallel relationship, and conflict relationship.

[0154] Among them, the portrait label determination module 303 is configured to perform label validity selection processing on the current portrait label and the historical portrait label according to the label logic relationships in the label logic relationship set, so as to obtain the valid portrait labels of the target user in the current update cycle.

[0155] In one embodiment, when the portrait label determination module 303 executes the step of performing label validity selection processing on the current portrait label and the historical portrait label according to the label logic relationships in the label logic relationship set to obtain the valid portrait labels of the target user in the current update cycle, it specifically executes: searching for the target label logic relationship between the current portrait label and the historical portrait label according to the label logic relationships in the label logic relationship set to obtain a target label logic relationship set; and performing label validity selection processing on the current portrait label and the historical portrait label according to the target label logic relationship set to obtain the valid portrait labels of the target user in the current update cycle.

[0156] In one embodiment, when the portrait label determination module 303 executes the step of performing label validity selection processing on the current portrait label and the historical portrait label according to the target label logical relationship set to obtain the valid portrait labels of the target user in the current update cycle, it specifically executes: performing first validity selection processing on the historical portrait label according to the target label logical relationship set to determine the historical impact portrait label; performing second validity selection processing on the historical impact portrait label and the current portrait label to obtain the valid portrait labels of the target user in the current update cycle.

[0157] In one embodiment, the second validity selection processing includes union processing. When the portrait label determination module 303 executes the step of performing second validity selection processing on the historical impact portrait label and the current portrait label to obtain the valid portrait labels of the target user in the current update cycle, it specifically executes: performing union processing on the historical impact portrait label and the current portrait label; using the portrait labels obtained after the union processing as the valid portrait labels of the target user in the current update cycle, and using the coefficients of the portrait labels obtained after the union processing as the weight coefficients of the corresponding portrait labels in the valid portrait labels.

[0158] In one embodiment, after the portrait label determination module 303 executes the step of obtaining the valid portrait labels of the target user in the current update cycle, it further executes: obtaining the weight coefficients of each portrait label in the valid portrait labels; determining the importance degree of each portrait label according to the weight coefficients, and selecting portrait labels from the valid portrait labels as the main portrait labels of the target user in the current update cycle according to the importance degree.

[0159] Among them, the update module 304 is used to update the user portrait of the target user according to the valid portrait labels.

[0160] In one embodiment, as Figure 7 shown, the user portrait update device may further include a second module 306, an encoding module 307, and a recommended label determination module 308.

[0161] Among them, the second acquisition module 306 is used to acquire the basic information of multiple users in the target application scenario, the business information of each user in the target application scenario, and the response information of each user to the business information. The business information includes the current behavior data in the current update cycle and the valid portrait labels corresponding to the user portrait. The valid portrait labels are obtained according to the user portrait update method described in any of the above embodiments, or are obtained according to the user portrait update device described in any of the above embodiments.

[0162] Among them, the encoding module 307 is used to encode the basic information, the service information, and the response information of each user to construct the feature data corresponding to each user, and use the feature data corresponding to each user as each sample to obtain a sample set.

[0163] In one embodiment, the encoding module 307 is specifically configured to encode the digital information and the enumerable text information in the basic information, the service information, and the response information of each user by using a one-hot encoding method; encode the non-enumerable text information in the basic information, the service information, and the response information of each user by using a text vector method; and use the information after encoding each user as the feature data corresponding to each user.

[0164] Among them, the recommended label determination module 308 is used to obtain the feature data of the target sample whose response information is subscribed from the sample set, and determine the recommended level label of the user portrait of the user corresponding to the to-be-classified sample whose response information is unsubscribed in the sample set, where the recommended level label is used to recommend the corresponding service to the user corresponding to the to-be-classified sample.

[0165] In one embodiment, the recommended label determination module 308 is specifically configured to initialize a first prototype vector, where the first prototype vector includes the target sample, and the first prototype vector is used to store multiple samples whose recommended level label is the first recommended level label; perform recommended clustering analysis on the to-be-classified samples whose response information is unsubscribed in the sample set according to the feature data of the target sample and the first recommended level label in the first prototype vector to determine the target recommended level label of the user portrait of the user corresponding to the to-be-classified sample, and cluster the to-be-classified samples into a sample cluster matching the target recommended level label according to the target recommended level label.

[0166] In one embodiment, when the recommended label determination module 308 executes the step of performing recommended clustering analysis on the to-be-classified samples whose response information is unsubscribed in the sample set according to the feature data of the target sample and the first recommended level label in the first prototype vector to determine the target recommended level label of the user portrait of the user corresponding to the to-be-classified sample, it specifically executes: determining the distance between the feature data of the to-be-classified sample whose response information is unsubscribed in the sample set and the feature data of the target sample in the first prototype vector; dividing distance intervals according to the distance, where the number of the distance intervals is the same as the number of preset multiple recommended level labels, and each distance interval determines a corresponding recommended level label according to the corresponding distance size and the first recommended level label; and determining the target recommended level label of the user portrait of the user corresponding to the to-be-classified sample according to the distance interval to which the distance belongs.

[0167] Any combination of the above technical solutions can form an optional embodiment of the present application, which will not be elaborated one by one here.

[0168] Correspondingly, an embodiment of the present application further provides an electronic device, which can be a terminal or a server. As Figure 8 shown, Figure 8 is a schematic structural diagram of the electronic device provided by the embodiment of the present application. The electronic device 400 includes a processor 401 with one or more processing cores, a memory 402 with one or more computer-readable storage media, and a computer program stored on the memory 402 and executable on the processor. Among them, the processor 401 is electrically connected to the memory 402. Those skilled in the art can understand that the structural diagram of the electronic device shown in the figure does not constitute a limitation on the electronic device, and it may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0169] The processor 401 is the control center of the electronic device 400, connecting various parts of the entire electronic device 400 through various interfaces and lines. By running or loading software programs (computer programs) and / or modules stored in the memory 402, and calling the data stored in the memory 402, it executes various functions of the electronic device 400 and processes data, thereby monitoring the entire electronic device 400.

[0170] In the embodiment of the present application, the processor 401 in the electronic device 400 will load the instructions corresponding to the processes of one or more application programs into the memory 402 according to the following steps, and the processor 401 will run the application programs stored in the memory 402 to implement various functions:

[0171] Based on the correspondence between the behavior data in the target application scenario and each portrait label in the portrait label set, process the current behavior data of the target user in the current update cycle to obtain the current portrait label; obtain the label logic relationship set and the historical portrait labels of the target user in the historical update cycle, where the label logic relationship set includes the label logic relationships between the portrait labels in the portrait label set; according to the label logic relationships in the label logic relationship set, perform label validity selection processing on the current portrait label and the historical portrait labels to obtain the valid portrait labels of the target user in the current update cycle; update the user portrait of the target user according to the valid portrait labels.

[0172] For the specific implementation of each of the above operations, reference can be made to the previous embodiments, which will not be elaborated here.

[0173] Optionally, as Figure 8As shown, the electronic device 400 further includes: a touch display screen 403, a radio frequency circuit 404, an audio circuit 405, an input unit 406, and a power supply 407. Among them, the processor 401 is electrically connected to the touch display screen 403, the radio frequency circuit 404, the audio circuit 405, the input unit 406, and the power supply 407 respectively. Those skilled in the art can understand that Figure 8 the structure of the electronic device shown in

[0174] does not limit the electronic device, and may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements. The touch display screen 403 can be used to display a graphical user interface and receive operation instructions generated by the user acting on the graphical user interface. The touch display screen 403 may include a display panel and a touch panel. Among them, the display panel can be used to display information input by the user or information provided to the user, as well as various graphical user interfaces of the electronic device. These graphical user interfaces can be composed of graphics, text, icons, videos, and any combination thereof. Optionally, the display panel can be configured in the form of a liquid crystal display (LCD), an organic light-emitting diode (OLED), etc. The touch panel can be used to collect touch operations of the user on or near it (such as operations of the user using a finger, a stylus, or any suitable object or accessory on or near the touch panel), and generate corresponding operation instructions, and the operation instructions execute the corresponding program. The touch panel can cover the display panel. After the touch panel detects a touch operation on or near it, it transmits it to the processor 401 to determine the type of touch event. Subsequently, the processor 401 provides a corresponding visual output on the display panel according to the type of touch event. In the embodiments of the present application, the touch panel and the display panel can be integrated into the touch display screen 403 to implement input and output functions. However, in some embodiments, the touch panel and the touch panel can be implemented as two independent components to implement input and output functions. That is, the touch display screen 403 can also be used as a part of the input unit 406 to implement the input function.

[0175] In the embodiments of the present application, the touch display screen 403 is used to present a graphical user interface and receive operation instructions generated by the user acting on the graphical user interface.

[0176] The radio frequency circuit 404 can be used to transmit and receive radio frequency signals to establish wireless communication with a network device or other electronic devices, and transmit and receive signals with the network device or other electronic devices.

[0177] The audio circuit 405 can be used to provide an audio interface between the user and the electronic device through a speaker and a microphone. The audio circuit 405 can convert the received audio data into an electrical signal and transmit it to the speaker, which converts it into a sound signal for output; on the other hand, the microphone converts the collected sound signal into an electrical signal, which is received by the audio circuit 405, converted into audio data, and then the audio data is output to the processor 401 for processing, and then sent to another electronic device, for example, through the radio frequency circuit 404, or the audio data is output to the memory 402 for further processing. The audio circuit 405 may also include an earphone jack to provide communication between the peripheral earphone and the electronic device.

[0178] The input unit 406 can be used to receive input digital, character information or user feature information (such as fingerprint, iris, face information, etc.), and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function controls.

[0179] The power supply 407 is used to supply power to each component of the electronic device 400. Optionally, the power supply 407 can be logically connected to the processor 401 through a power management system, so as to realize functions such as management of charging, discharging, and power consumption management through the power management system. The power supply 407 may also include any components such as one or more DC or AC power supplies, a recharge system, a power failure detection circuit, a power converter or inverter, and a power status indicator.

[0180] Although Figure 8 not shown in the figure, the electronic device 400 may also include a camera, a sensor, a Wi-Fi module, a Bluetooth module, etc., which will not be elaborated here.

[0181] In the above embodiments, the descriptions of the various embodiments have their own focuses. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0182] Those of ordinary skill in the art can understand that all or part of the steps in the above methods of the embodiments can be completed by instructions, or by controlling related hardware through instructions. The instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0183] Therefore, the embodiments of the present application provide a computer-readable storage medium, in which multiple computer programs are stored. The computer programs can be loaded by a processor to execute the steps in any of the user portrait update methods provided by the embodiments of the present application. For example, the computer program can execute the following steps:

[0184] Based on the correspondence between the behavior data in the target application scenario and each portrait label in the portrait label set, process the current behavior data of the target user in the current update cycle to obtain the current portrait label; obtain the label logic relation set and the historical portrait labels of the target user in the historical update cycle, where the label logic relation set includes the label logic relations between the portrait labels in the portrait label set; perform label validity selection processing on the current portrait label and the historical portrait labels according to the label logic relations in the label logic relation set to obtain the valid portrait labels of the target user in the current update cycle; update the user portrait of the target user according to the valid portrait labels.

[0185] For the specific implementation of each of the above operations, reference may be made to the previous embodiments and will not be elaborated here.

[0186] Among them, the computer-readable storage medium may include: read-only memory (ROM, Read Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk, etc.

[0187] Since the computer program stored in this storage medium can execute the steps in any of the user portrait update methods provided in the embodiments of the present application, the beneficial effects achievable by any of the user portrait update methods provided in the embodiments of the present application can be realized. For details, refer to the previous embodiments and will not be elaborated here.

[0188] The above has introduced in detail a user portrait update method, device, computer-readable storage medium, and electronic device provided in the embodiments of the present application. Specific examples are used in this article to elaborate on the principle and implementation manner of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those skilled in the art, according to the idea of the present application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation of the present application.

Claims

1. A method for updating user profiles, characterized in that, Including: Based on the correspondence between the behavior data in the target application scenario and each portrait label in the portrait label set, process the current behavior data of the target user in the current update cycle to obtain the current portrait label; Obtain the label logical relationship set and the historical portrait labels of the target user in the historical update cycle, where the label logical relationship set includes the label logical relationships between each portrait label in the portrait label set, and the label logical relationships include inclusion relationship, included relationship, parallel relationship, and conflict relationship; According to the label logical relationships in the label logical relationship set, find the target label logical relationship between the current portrait label and the historical portrait labels to obtain the target label logical relationship set; According to the target label logical relationship set, perform a first validity selection process on the historical portrait labels to determine the historical impact portrait labels; Perform a union process on the historical impact portrait labels and the current portrait label; Use the portrait labels obtained after the union process as the valid portrait labels of the target user in the current update cycle, and use the coefficients of the portrait labels obtained after the union process as the weight coefficients of the corresponding portrait labels in the valid portrait labels; Update the user portrait of the target user according to the valid portrait labels.

2. The user profile update method according to claim 1, wherein After the step of obtaining the valid portrait labels of the target user in the current update cycle, it further includes: Obtain the weight coefficients of each portrait label in the valid portrait labels; Determine the importance degree of each portrait label according to the weight coefficients, and select portrait labels from the valid portrait labels as the main portrait labels of the target user in the current update cycle according to the importance degree.

3. The user profile update method according to claim 1, wherein The label logical relationship set includes a label transition matrix, and the label transition matrix includes matrix elements, where the matrix elements are used to represent the label logical relationships between the portrait labels; The step of obtaining the label logical relationship set includes: obtaining the label transition matrix.

4. The user profile update method according to claim 3, wherein Before the step of obtaining the label transition matrix, it further includes: Construct the label transition matrix according to all the portrait labels in the portrait label set and the label logical relationships between each portrait label in all the portrait labels.

5. The user profile update method according to claim 4, wherein The step of constructing the label transition matrix according to all the portrait labels in the portrait label set and the label logical relationships between each portrait label in all the portrait labels includes: Obtain the label quantity N of all the portrait labels in the portrait label set, where N takes a positive integer greater than zero; Construct an initial label matrix with dimensions of N*N, where each dimension in the initial label matrix corresponds to a portrait label; Fill the initial label matrix according to the label logical relationships between each portrait label to obtain the label transition matrix.

6. The user portrait updating method according to claim 5, wherein The step of filling the initial label matrix according to the label logical relationships between each portrait label to obtain the label transition matrix includes: Obtain the logical relationship weights corresponding to the label logical relationships between each portrait label; Fill the initial label matrix according to the logical relationship weights to obtain the label transition matrix.

7. The user profile update method according to claim 1, wherein The step of processing the current behavior data of the target user in the current update cycle to obtain the current portrait label includes: Obtain the current behavior data of the target user in the target application scenario in the current update cycle; Conduct behavior analysis on the current behavior data to obtain at least one candidate portrait label; When there is one candidate portrait label, use the candidate portrait label as the current portrait label corresponding to the current behavior data of the target user; When there are multiple candidate portrait labels and the label logical relationship among the multiple candidate portrait labels includes an inclusion relationship, delete the included portrait labels in the inclusion relationship, and use the remaining candidate portrait labels as the current portrait label corresponding to the current behavior data of the target user.

8. The user profile update method according to claim 7, characterized in that, The step of obtaining the current behavior data of the target user in the target application scenario in the current update cycle includes: Obtain the outbound interaction corpus data of the target user in the outbound marketing scenario in the current update cycle; The step of conducting behavior analysis on the current behavior data includes: conducting corpus behavior analysis on the outbound interaction corpus data.

9. The user profile update method according to any one of claims 1 to 8, characterized in that It further includes: Obtain the basic information of multiple users, the business information of each user in the target application scenario, and the response information of each user to the business information, where the business information includes the current behavior data in the current update cycle and the valid portrait labels in the corresponding user portraits; Encode the basic information, the business information, and the response information of each user to construct the feature data corresponding to each user, and use the feature data corresponding to each user as a sample to obtain a sample set; Obtain the feature data of the target sample with the response information being subscribed from the sample set, and determine the recommended level label of the user portrait of the user corresponding to the unsubscribed sample to be classified in the sample set according to the feature data of the target sample; where the recommended level label is used to recommend the corresponding business to the user corresponding to the sample to be classified.

10. The user portrait updating method according to claim 9, wherein, The step of obtaining the feature data of the target sample with the response information being subscribed from the sample set and determining the recommended level label of the user portrait of the user corresponding to the unsubscribed sample to be classified in the sample set according to the feature data of the target sample includes: Initialize a first prototype vector, where the first prototype vector includes the target sample, and the first prototype vector is used to store multiple samples with the recommended level label being the first recommended level label; Conduct recommended clustering analysis on the unsubscribed samples to be classified in the sample set according to the feature data of the target sample in the first prototype vector and the first recommended level label to determine the target recommended level label of the user portrait of the user corresponding to the sample to be classified, and cluster the samples to be classified into the sample cluster matching the target recommended level label according to the target recommended level label.

11. The user profile update method according to claim 10, characterized in that The step of performing recommended clustering analysis on the to-be-classified samples with unsubscribed response information in the sample set according to the feature data of the target samples in the first prototype vector and the first recommended level label to determine the target recommended level label of the user portrait corresponding to the to-be-classified samples includes: Determining the distance between the feature data of the to-be-classified samples with unsubscribed response information in the sample set and the feature data of the target samples in the first prototype vector; Dividing distance intervals according to the distance, where the number of distance intervals is the same as the number of preset multiple recommended level labels, and each distance interval determines a corresponding recommended level label according to the corresponding distance size and the first recommended level label; Determining the target recommended level label of the user portrait corresponding to the to-be-classified samples according to the distance interval to which the distance belongs.

12. The user portrait updating method according to claim 9, wherein The step of encoding the basic information, the service information, and the response information of each user to construct the feature data corresponding to each user includes: Encoding the digital information and the enumerable text information in the basic information, the service information, and the response information of each user by using the one-hot encoding method; Encoding the non-enumerable text information in the basic information, the service information, and the response information of each user by using the text vector method; Taking the information after encoding each user as the feature data corresponding to each user.

13. The user portrait updating method according to claim 9, wherein The service information further includes the promotion package information of the target application scenario; the response information of the user to the service includes the ordering information or subscription information of the user to the promotion package information.

14. A user profile update device, characterized in that, Including: A current label determination module, configured to process the current behavior data of the target user in the current update cycle based on the correspondence between the behavior data in the target application scenario and each portrait label in the portrait label set to obtain the current portrait label; A first acquisition module, configured to acquire the label logic relationship set and the historical portrait labels of the target user in the historical update cycle, where the label logic relationship set includes the label logic relationships between the portrait labels in the portrait label set, and the label logic relationships include the inclusion relationship, the included relationship, the parallel relationship, and the conflict relationship; A portrait label determination module, configured to find the target label logic relationship between the current portrait label and the historical portrait labels according to the label logic relationships in the label logic relationship set to obtain the target label logic relationship set, perform a first validity selection process on the historical portrait labels according to the target label logic relationship set to determine the historical influence portrait labels, perform a union process on the historical influence portrait labels and the current portrait labels, take the portrait labels obtained after the union process as the valid portrait labels of the target user in the current update cycle, and take the coefficients of the portrait labels obtained after the union process as the weight coefficients of the corresponding portrait labels in the valid portrait labels; An update module, configured to update the user portrait of the target user according to the valid portrait labels.

15. The user profile updating device according to claim 14, characterized in that, Further including: A second acquisition module, configured to acquire basic information of multiple users in the target application scenario, business information of each user in the target application scenario, and response information of each user to the business information, where the business information includes current behavior data in the current update period and valid portrait tags in the corresponding user portrait, and the valid portrait tags are obtained by the user portrait updating device according to claim 14; An encoding module, configured to encode the basic information, the business information, and the response information of each user to construct feature data corresponding to each user, and use the feature data corresponding to each user as each sample to obtain a sample set; A recommended tag determination module, configured to obtain the feature data of a target sample whose response information is subscribed from the sample set, and determine a recommended level tag of the user portrait of the user corresponding to a to-be-classified sample whose response information is unsubscribed in the sample set according to the feature data of the target sample; wherein, the recommended level tag is used to recommend corresponding services to the user corresponding to the to-be-classified sample.

16. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and the computer program is suitable for being loaded by a processor to execute the steps in the user portrait updating method according to any one of claims 1-13.

17. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory stores a computer program, and the processor executes the steps in the user portrait updating method according to any one of claims 1-13 by calling the computer program stored in the memory.

Images