Method and apparatus for recommending service product

Abstract

The application provides a service product recommendation method and device, the method comprises the following steps: obtaining service information to be processed of a user; wherein the service information comprises at least one information group, the information group comprises service data, generation time of the service data and device information of the service data; performing feature mining processing on the information group to obtain first service features of the information group and determine second service features corresponding to the service information; determining first probability information and second probability information according to the second service features and a preset identification model, wherein the first probability information is used to indicate a probability value of needing to use a service product to be recommended, and the second probability information is used to indicate a probability value of not needing to use the service product to be recommended; and if the first probability information is greater than the second probability information, the service product to be recommended is recommended to the user. The method provided in the application can improve the accuracy of the service product recommended to the user.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data processing, and in particular to a service product recommendation method and device. BACKGROUND

[0002] At present, with the continuous development of technology, the types of service products in the financial field that can be provided for users are also increasing. Among them, the service product in the actual application can be the software installed in the intelligent terminal, or a specific function embedded in the software of the intelligent terminal, etc.

[0003] Therefore, how to determine the service product with higher relevance to the user's service demand from a large number of service products is a problem that needs to be solved urgently. SUMMARY

[0004] The service product recommendation method and device provided by the present application are used to recommend the service product with higher relevance to the user's demand to the user.

[0005] In a first aspect, the present application provides a service product recommendation method, comprising:

[0006] Obtaining service information to be processed of a user: wherein the service information comprises at least one information group, and the information group has a service category identifier; the information group comprises service data, generation time of the service data, and device information of the service data;

[0007] Performing feature mining processing on the information group to obtain first service features of the information group, and determining second service features corresponding to the service information;

[0008] According to the second service features and a preset identification model, determining first probability information and second probability information, wherein the first probability information is used to indicate a probability value of needing to use a service product to be recommended; the second probability information is used to indicate a probability value of not needing to use the service product to be recommended; and the preset identification model is used to determine a probability of needing to use the service product to be recommended and a probability of not needing to use the service product to be recommended;

[0009] If the first probability information is greater than the second probability information, the service product to be recommended is recommended to the user.

[0010] In a possible implementation, the feature mining processing on the information group to obtain the first service features of the information group comprises:

[0011] determine a first vector set based on the service data of the information group and the generation time of the service data, wherein the first vector set comprises at least one first feature vector, and the first feature vector comprises the service data and the generation time of the service data, and the generation time of the service data included in different first feature vectors is different;

[0012] perform clustering processing on the first vector set to obtain a first clustering result, wherein the first clustering result comprises at least one first cluster and a first clustering feature corresponding to the first cluster, and the first cluster comprises at least one first feature vector;

[0013] determine a second vector set based on the service data of the information group and the device information of the service data, wherein the second vector set comprises at least one second feature vector, and the second feature vector comprises the service data and the device information of the service data, and the generation time of the service data included in different second feature vectors is different;

[0014] perform clustering processing on the second vector set to obtain a second clustering result, wherein the second clustering result comprises at least one second cluster and a second clustering feature corresponding to the second cluster, and the second cluster comprises at least one second feature vector;

[0015] determine the first clustering feature and the second clustering feature as the first service feature of the information group.

[0016] In a possible implementation, the clustering processing on the first vector set to obtain the first clustering result comprises:

[0017] perform filtering processing on the first vector set, and determine K first feature vectors obtained after the filtering processing as K first center vectors; K is a positive integer; and each first center vector corresponds to a first cluster;

[0018] for each first feature vector, determine a first difference value result of the first feature vector in the first vector set and the K first center vectors respectively, and determine a first cluster corresponding to a first center vector with a minimum first difference value result as a first cluster to which the first feature vector belongs;

[0019] for each first cluster, determine a first difference value sum corresponding to the first feature vector in the first cluster, wherein the first difference value sum is a sum of difference values of the first feature vector and each first feature vector in the first cluster, determine a second difference value sum of the first center vector corresponding to the first cluster and each first feature vector in the first cluster, and determine a first clustering feature corresponding to the first cluster based on the first difference value sum and the second difference value sum.

[0020] determining the first cluster feature of the first cluster as the first clustering result.

[0021] In a possible implementation, the first clustering feature corresponding to the first cluster is determined based on the first difference sum and the second difference sum, including:

[0022] if the first difference sum is less than the second difference sum, determining the first feature vector corresponding to the first difference sum as the first clustering feature corresponding to the first cluster;

[0023] if the first difference sum is greater than or equal to the second difference sum, determining the first center vector corresponding to the second difference sum as the first clustering feature corresponding to the first cluster.

[0024] In a possible implementation, the preset identification model is established based on the third service feature of each to-be-trained service information and the first use information corresponding to each to-be-trained service information; the third service feature of the to-be-trained service information is determined based on the fourth service feature of an information group contained in the to-be-trained service information; the fourth service feature of the information group is obtained based on feature mining processing on the information group; and the first use information is used to represent whether the user has used the to-be-recommended service product under the to-be-trained service information.

[0025] In a possible implementation, the preset identification model is established based on a Bayesian formula; wherein the third probability information, the fourth probability information, the fifth probability information and the sixth probability information included in the preset model are determined based on the third service feature of each to-be-trained service information and the first use information corresponding to each to-be-trained service information.

[0026] The third probability information is used to represent the probability of occurrence of the first use information representing use of the to-be-recommended service product in each first use information; the fourth probability information is used to represent the probability of occurrence of the first use information representing non-use of the to-be-recommended service product in each first use information; the fifth probability information is used to represent the probability of occurrence of the third service feature in the case of use of the to-be-recommended service product; and the sixth probability information is used to represent the probability of occurrence of the third service feature in the case of non-use of the to-be-recommended service product.

[0027] In a possible implementation, the method further includes:

[0028] If it is determined that the data dimension of the device information in the to-be-processed service information is greater than the data dimension of the device information in the to-be-trained service information obtained in the preset model establishment process, and the to-be-recommended service product does not change, then the third service feature corresponding to the to-be-trained service information is adjusted according to the data dimension of the device information in the to-be-processed service information, to obtain an adjusted third service feature.

[0029] Based on the adjusted third service feature, the fifth probability information and the sixth probability information of the third service feature corresponding to the to-be-trained service information are adjusted, to obtain adjusted fifth probability information and adjusted sixth probability information.

[0030] Based on the third probability information, the fourth probability information, the adjusted fifth probability information, the adjusted sixth probability information, and the Bayes formula, the preset identification model is updated.

[0031] In a possible implementation, the method further includes:

[0032] If it is determined that the data dimension in the to-be-processed service information is greater than the data dimension in the to-be-trained service information obtained in the preset model establishment process, and the to-be-recommended service product changes, then the third service feature corresponding to the to-be-trained service information is adjusted according to the data dimension in the to-be-processed service information, to obtain an adjusted third service feature.

[0033] The second use information corresponding to the to-be-trained service information is obtained, where the second use information is used to represent whether the changed to-be-recommended service product is used.

[0034] The third probability information, the adjusted third probability information, the adjusted fourth probability information, the adjusted fifth probability information, and the adjusted sixth probability information are determined according to the second use information and the adjusted to-be-trained service feature information.

[0035] Based on the adjusted third probability information, the adjusted fourth probability information, the adjusted fifth probability information, the adjusted sixth probability information, and the Bayes formula, the preset identification model is updated.

[0036] In a possible implementation, the information group is obtained by performing cyclic shift processing on an initial information group; the initial information group includes initial service data, a generation time of the initial service data, and device information of the initial service data.

[0037] In a second aspect, the application provides a service product recommendation device, including:

[0038] The first obtaining unit is configured to obtain service information to be processed of a user, wherein the service information comprises at least one information group, and the information group comprises a service category identifier; and the information group comprises service data, a generation time of the service data, and device information of the service data.

[0039] The processing unit is configured to perform feature mining processing on the information group to obtain first service features of the information group.

[0040] The first determining unit is configured to determine second service features corresponding to the service information.

[0041] The second determining unit is configured to determine first probability information and second probability information according to the second service features and a preset identification model, wherein the first probability information is used to indicate a probability value of needing to use a service product to be recommended; the second probability information is used to indicate a probability value of not needing to use the service product to be recommended; and the preset identification model is used to determine a probability of needing to use the service product to be recommended and a probability of not needing to use the service product to be recommended.

[0042] The pushing unit is configured to recommend the service product to be recommended to the user if it is determined that the first probability information is greater than the second probability information.

[0043] In a possible implementation manner, the processing unit comprises:

[0044] The first determining module is configured to determine a first vector set based on the service data of the information group and the generation time of the service data, wherein the first vector set comprises at least one first feature vector, the first feature vector comprises the service data and the generation time of the service data, and different first feature vectors comprise different generation times of the service data.

[0045] The first processing module is configured to perform clustering processing on the first vector set to obtain a first clustering result, wherein the first clustering result comprises at least one first cluster and a first clustering feature corresponding to the first cluster in one-to-one manner, and the first cluster comprises at least one first feature vector.

[0046] The second determining module is configured to determine a second vector set based on the service data of the information group and the device information of the service data, wherein the second vector set comprises at least one second feature vector, the second feature vector comprises the service data and the device information of the service data, and different second feature vectors comprise different generation times of the service data.

[0047] The second processing module is configured to perform clustering processing on the second vector set to obtain a second clustering result, the second clustering result comprising at least one second cluster and a second clustering feature corresponding to the second cluster, and the second cluster comprising at least one second feature vector.

[0048] The third determining module is configured to determine that the first clustering feature and the second clustering feature are first service features of the information group.

[0049] In a possible implementation, the first processing module is specifically configured to:

[0050] perform filtering processing on the first vector set, and determine K first center vectors obtained after filtering as K first center vectors; K is a positive integer; and each first center vector corresponds to a first cluster;

[0051] For each first feature vector, determine first difference value results of the first feature vector and the K first center vectors in the first vector set respectively, and determine a first cluster of a first center vector with a minimum first difference value result as a first cluster to which the first feature vector belongs;

[0052] For each first cluster, determine a first difference value sum corresponding to the first feature vector in the first cluster, wherein the first difference value sum is a sum of differences between the first feature vector and each first feature vector in the first cluster; determine a second difference value sum of the first center vector corresponding to the first cluster and each first feature vector in the first cluster; and determine a first clustering feature corresponding to the first cluster based on the first difference value sum and the second difference value sum.

[0053] Determine the first cluster and the first clustering feature of the first cluster as the first clustering result.

[0054] In a possible implementation, the first processing module is specifically configured to:

[0055] If the first difference value sum is smaller than the second difference value sum, determine a first feature vector corresponding to the first difference value sum as the first clustering feature corresponding to the first cluster;

[0056] If the first difference value sum is greater than or equal to the second difference value sum, determine a first center vector corresponding to the second difference value sum as the first clustering feature corresponding to the first cluster.

[0057] In a possible implementation, the preset identification model is established based on third service features of each piece of to-be-trained service information and first use information corresponding to each piece of the to-be-trained service information; the third service feature of the to-be-trained service information is determined based on a fourth service feature of an information group contained in the to-be-trained service information; the fourth service feature of the information group is obtained based on feature mining processing on the information group; and the first use information is used to represent whether the user has used the to-be-recommended service product under the to-be-trained service information.

[0058] In a possible implementation, the preset identification model is established based on a Bayesian formula; wherein the third probability information, the fourth probability information, the fifth probability information and the sixth probability information included in the preset model are determined based on third service features of each piece of to-be-trained service information and first use information corresponding to each piece of the to-be-trained service information.

[0059] The third probability information is used to represent a probability of occurrence of first use information representing use of the to-be-recommended service product in each piece of first use information; the fourth probability information is used to represent a probability of occurrence of first use information representing non-use of the to-be-recommended service product in each piece of first use information; the fifth probability information is used to represent a probability of occurrence of the third service feature in the case of use of the to-be-recommended service product; and the sixth probability information is used to represent a probability of occurrence of the third service feature in the case of non-use of the to-be-recommended service product.

[0060] In a possible implementation, the apparatus further includes:

[0061] The third determination unit is configured to, if it is determined that the data dimension of the device information in the to-be-processed service information is greater than the data dimension of the device information in the to-be-trained service information obtained in the preset model establishment process and the to-be-recommended service product is not changed, adjust the third service feature corresponding to the to-be-trained service information according to the data dimension of the device information in the to-be-processed service information, to obtain an adjusted third service feature.

[0062] The first adjustment unit is configured to adjust the fifth probability information and the sixth probability information of the third service feature corresponding to the to-be-trained service information based on the adjusted third service feature, to obtain adjusted fifth probability information and adjusted sixth probability information.

[0063] The first update unit is configured to update the preset identification model based on the third probability information, the fourth probability information, the adjusted fifth probability information, the adjusted sixth probability information and the Bayesian formula.

[0064] In a possible implementation, the apparatus further includes:

[0065] The fourth determining unit is configured to, if it is determined that the data dimension in the to-be-processed service information is greater than the data dimension in the to-be-trained service information obtained in the preset model establishment process and the to-be-recommended service product changes, adjust the third service feature corresponding to the to-be-trained service information according to the data dimension in the to-be-processed service information, to obtain an adjusted third service feature.

[0066] The second obtaining unit is configured to obtain second usage information corresponding to the to-be-trained service information, where the second usage information is used to represent whether the changed to-be-recommended service product is used.

[0067] The fifth determining unit is configured to determine adjusted third probability information, adjusted fourth probability information, adjusted fifth probability information, and adjusted sixth probability information according to the second usage information and the adjusted to-be-trained service feature information.

[0068] The second updating unit is configured to update the preset identification model based on the adjusted third probability information, the adjusted fourth probability information, the adjusted fifth probability information, the adjusted sixth probability information, and a Bayes formula.

[0069] In a possible implementation, the information group is obtained by performing cyclic shift processing on an initial information group; and the initial information group includes initial service data, a generation time of the initial service data, and device information of the initial service data.

[0070] In a third aspect, the present application provides an electronic device, including: a memory, a processor;

[0071] a memory; and a memory for storing executable instructions of the processor;

[0072] The processor is configured to execute the method according to any one of the first aspect.

[0073] In a fourth aspect, the present application provides a computer readable storage medium, the computer readable storage medium stores computer execution instructions, and the computer execution instructions are executed by a processor to implement the method according to any one of the first aspect.

[0074] In a fifth aspect, the present application provides a computer program product, the computer program product includes a computer program, and the computer program is executed by a processor to implement the method according to any one of the first aspect.

[0075] This application provides a service product recommendation method and apparatus. The method includes: acquiring user's pending service information; wherein the service information includes: at least one information group, the information group having a service category identifier; the information group includes: service data, service data generation time, and service data device information; performing feature mining processing on the information group to obtain a first service feature of the information group, and determining a second service feature corresponding to the service information; determining first probability information and second probability information based on the second service feature and a preset recognition model, wherein the first probability information is used to indicate the probability value of needing to use the service product to be recommended; the second probability information is used to indicate the probability value of not needing to use the service product to be recommended; the preset recognition model is used to determine the probability of needing to use the service product to be recommended and the probability of not needing to use the service product to be recommended; if the first probability information is greater than the second probability information, then recommending the service product to be recommended to the user. Through the above method, by combining the first service feature corresponding to the information group generated by the service products of different service categories currently used by the user and the preset recognition model, the probability of whether the current user needs to use the service product to be recommended is determined, so as to further determine whether to recommend the service product to the user based on the determined probability value. In addition, the service information also includes device information corresponding to the service data, so as to obtain data generated by users using the same service category on different devices, thereby improving the completeness of the data and the accuracy of the final determined first probability information and second probability information, making the service products recommended to users more accurate. Attached Figure Description

[0076] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0077] Figure 1 A flowchart illustrating a method for recommending service products provided in an embodiment of this application;

[0078] Figure 2 A flowchart illustrating another method for recommending a service product provided in an embodiment of this application;

[0079] Figure 3 This is a schematic diagram of an application scenario provided by an embodiment of this application;

[0080] Figure 4 This application provides a schematic diagram of the structure of a service product recommendation device.

[0081] Figure 5 A schematic diagram of the structure of a recommending device for yet another service product provided in this application embodiment;

[0082] Figure 6 Fig. 1 is a schematic diagram of an electronic device according to an embodiment of the present application.

[0083] The above-described embodiments of the present application have been shown and described, and the following detailed description will be given with reference to the accompanying drawings. These drawings and detailed description are not intended to limit the scope of the present application in any way, but to explain the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0084] The exemplary embodiments will be described in detail with reference to the accompanying drawings. The following description is made with reference to the accompanying drawings, in which like numerals represent like elements, unless the context dictates otherwise. The embodiments described in the following exemplary embodiments are not meant to represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present application.

[0085] At present, with the continuous development of science and technology, the types of service products that can be provided in the financial field are also increasingly diversified, thereby making it difficult for users to find service products that meet their own needs in a short period of time. For example, although the current bank can provide users with a variety of different service products, due to the diversity of the current information types, the user can not be able to timely learn the update information of the latest service products of the bank, resulting in the user being unable to select the service products suitable for themselves in the service products. Therefore, how to recommend service products that meet the needs of users is a problem that needs to be solved urgently.

[0086] The service product recommendation method and device provided by the present application are used to solve the above technical problems.

[0087] The technical solutions of the present application and how the technical solutions of the present application solve the above technical problems will be described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes can not be described in some embodiments. The embodiments of the present application will be described below with reference to the accompanying drawings.

[0088] It should be noted that the data related to the user obtained in the present application is legally collected data.

[0089] Figure 1 Fig. 1 is a schematic diagram of an electronic device according to an embodiment of the present application. Figure 1 As shown in the figure, the method comprises the following steps:

[0090] S101, acquire service information to be processed of a user: wherein the service information comprises at least one information group, the information group has a service category identifier; the information group comprises service data, generation time of the service data, and device information of the service data.

[0091] Exemplarily, the execution subject of the embodiment can be a service product recommendation device, wherein the recommendation device can be a server, an electronic device, a smart terminal, a chip, etc., which is not specifically limited in the application.

[0092] The service information to be processed of the user in the embodiment can be understood as the information generated by the user when using the service product in the current period. Specifically, the service information can include at least one information group, wherein one information group corresponds to one service category identifier, and the service category identifier can be understood as the product identifier of the service product used by the user in the current period. For example, the service category identifier can include periodic payment service, remote payment service, fast payment service, positioning service, etc. In the information group, the service data generated under the service category, the generation time of the service data, and the device information corresponding to the service data are included. It can be understood that the information group can include multiple pieces of service data, each piece of service data corresponds to its own generation time and device information corresponding to the service data. The device information of the service data can be understood as the device identifier used by the user when using the service product of the service category. For example, the device can be a mobile terminal device of the user, or a POS (point of sale) device, an ATM (Automated Teller Machine) device, etc. That is, in the acquired service information of the user, the data can be divided into multiple information groups based on the service category identifier corresponding to the data generation time.

[0093] In addition, step S101 can be started to execute under the trigger of the user clicking the service product recommendation button set in the smart terminal, so as to recommend the service product to the user based on the acquired service information under the trigger of the user. It can also be started to execute when it is detected that there is a new service product on the market, so as to recommend the new service product to the user in time when it is determined that the new service product meets the needs of the user. It can also be started to execute when it is detected that the type of service product used by the user changes. The embodiment is not specifically limited.

[0094] S102, perform feature mining processing on the information group to obtain a first service feature of the information group, and determine a second service feature corresponding to the service information.

[0095] Exemplarily, in the embodiment, when the service information to be processed is acquired, at this time, for each information group in the service information to be processed, the first service feature corresponding to the information group can be obtained by performing feature mining processing on the information group. Then, the first service feature corresponding to each information group included in the service information to be processed is taken as the second service feature corresponding to the service information to be processed.

[0096] The manner of the feature mining processing on the information group in the embodiment is not specifically limited, for example, in actual application, random forest, principal component analysis, compressed sensing and the like in related technologies can be used.

[0097] S103, determining first probability information and second probability information according to the second service feature and a preset identification model, wherein the first probability information is used to indicate a probability value of needing to use the service product to be recommended; the second probability information is used to indicate a probability value of not needing to use the service product to be recommended; and the preset identification model is used to determine the probability of needing to use the service product to be recommended and the probability of not needing to use the service product to be recommended.

[0098] Exemplarily, the preset identification model in the embodiment can be used to predict, based on the input service feature, the probability that the user currently needs to use the service product to be recommended corresponding to the preset identification model and the probability that the user does not need to use the service product to be recommended corresponding to the preset model. It should be noted that the preset identification model is a model corresponding to the service product to be recommended at present, so that whether the service product to be recommended is pushed to the user can be determined based on the preset identification model and the acquired second service feature. The preset identification model can be trained according to a plurality of second service features corresponding to a plurality of groups of historical period service information and whether the service corresponding to the model is used under the plurality of second service features, and the specific model training method is not specifically limited in the embodiment.

[0099] When the second service feature corresponding to the service information to be processed is acquired, the second service feature can be input into the preset identification model, and then the first probability information and the second probability information output by the preset identification model are obtained. The first probability information can be understood as a probability value that the user needs to use the service product to be recommended; and the second probability information can be understood as a probability value that the user does not need to use the service product to be recommended.

[0100] S104, if the first probability information is greater than the second probability information, recommending the service product to be recommended to the user.

[0101] Exemplarily, when the first probability information and the second probability information are acquired through the preset identification model, if it is determined that the probability value corresponding to the first probability information is greater than the probability value corresponding to the second probability information, it indicates that the demand of the user for using the service product to be recommended corresponding to the preset identification model is greater, which is determined based on the second service feature corresponding to the current user, and then the service product to be recommended is recommended to the user. It should be noted that the recommendation method is not limited in the embodiment.

[0102] If it is determined that the probability value corresponding to the first probability information is less than or equal to the probability value corresponding to the second probability information, it indicates that the demand of the user for using the service product corresponding to the preset identification model is smaller, which is determined based on the second service feature corresponding to the current user, and then the service product to be recommended does not need to be recommended to the user.

[0103] It can be understood that in the embodiment, the probability of whether the current user needs to use the service product to be recommended is determined by combining the first service feature corresponding to the information group generated by the different service categories of service products used by the current user and the preset identification model, so as to further determine whether to recommend the service product to the user based on the size of the determined probability value. In addition, the device information corresponding to the service data is also included in the service information, so that the data generated by the user using the same service category of service product on different devices can be acquired, thereby improving the completeness of the data and the accuracy of the finally determined first probability information and second probability information, so that the service product finally recommended to the user is more accurate.

[0104] Figure 2 Another flowchart of a service product recommendation method provided by the embodiment is shown in FIG. 6. Figure 2 The method comprises the following steps:

[0105] S201, acquiring service information to be processed of a user: wherein the service information comprises at least one information group, and the information group has a service category identifier; the information group comprises service data, generation time of the service data, and device information of the service data.

[0106] In one example, the information group is obtained by performing cyclic shift processing on an initial information group; the initial information group comprises initial service data, generation time of the initial service data, and device information of the initial service data.

[0107] Exemplarily, the execution subject in the embodiment can be a service product recommendation device, wherein the recommendation device can be a server, an electronic device, a smart terminal, a chip, etc., which is not limited in the application. The electronic device is taken as an example for description in the embodiment.

[0108] When the electronic device needs to obtain the service information to be processed of the user, the electronic device can determine by receiving the information reported by the device receiving each service data.

[0109] For example, Figure 3 An application scenario provided by the embodiment of the application is shown in the figure, in which the electronic device can establish a communication connection with a plurality of different service devices. In the figure, the service device can be understood as a device used by the user when using a service product. When each service device reports data to the electronic device, the service device can perform cyclic shift processing on the obtained initial service data, the generation time of the initial service data and the device information of the initial service data, and then report the shifted data to the electronic device. That is, the initial service data is subjected to cyclic shift processing to obtain service data; the generation time of the initial service data is subjected to cyclic shift processing to obtain the generation time of the service data; and the device information of the initial service data is subjected to cyclic shift processing to obtain the device information of the service data. In addition, the number of bits corresponding to the cyclic shift processing can be determined by a random number generated by the electronic device, and the value range of the random number is not specifically limited in the embodiment.

[0110] For example, the original initial service data represented by a binary number can be "1101000011110001", and the service data obtained after left cyclic shift by 4 bits can be "0000111100011101"

[0111] It should be noted that, in order to facilitate the subsequent processing of the data reported by each service device by the electronic device, the data dimensions of the three types of data, i.e. the initial service data, the generation time of the initial service data and the device information of the initial service data, are adjusted before the initial service data is subjected to cyclic shift. In this way, the data dimensions reported by each service device are the same. When the data dimension is small, 0 can be added to the high bit of the data to ensure the consistency of the data dimension.

[0112] It can be understood that, in the embodiment, the original data collected by the service device is processed by means of cyclic shift processing, so as to ensure the security in the data transmission process. In addition, the cyclic shift processing does not change the statistical characteristics of the data, which is conducive to the accuracy of the subsequent feature mining processing. In some possible implementation manners, the number of bits subjected to cyclic shift processing by the service device can be sent by the remaining encryption server (i.e. the electronic device which does not finally obtain the service information in the embodiment), and the encryption server and the service device can adopt a unidirectional communication link for unidirectional communication, i.e. only the encryption server is allowed to send a randomly generated random value to the service device, so as to reduce the probability of data leakage.

[0113] S202, determining a first vector set based on the service data in the information group and the generation time of the service data, wherein the first vector set comprises at least one first feature vector, and the first feature vector comprises the service data and the generation time of the service data.

[0114] For example, after receiving the service information to be processed in the embodiment, for each information group in the service information, the service data in the information group and the generation time corresponding to the service data can be combined as the first vector set. For example, when the information group comprises a piece of service data and the generation time corresponding to the service data, the service data and the corresponding generation time can be taken as the first feature vector. When the information group comprises multiple pieces of service data and the generation time corresponding to each piece of service data, the service data and the corresponding generation time can be taken as a first feature vector, and the generation time of the service data corresponding to different first feature vectors obtained by the above combination is different, and the first vector set is obtained.

[0115] S203, performing clustering processing on the first vector set to obtain a first clustering result, wherein the first clustering result comprises at least one first cluster and a first clustering feature corresponding to the first cluster, and the first cluster comprises at least one first feature vector.

[0116] For example, after obtaining the first vector set, in order to avoid the problem that the amount of data in the first vector set is large and the subsequent calculation amount of the electronic device is large, the first vector set is subjected to clustering processing in the embodiment, that is, the first feature vectors contained in the first vector set are subjected to clustering processing, and at least one first cluster after clustering is obtained, and each first cluster also has a corresponding first clustering feature. The first clustering feature can be understood as any first feature vector contained in the first cluster corresponding thereto, which is not limited in the embodiment.

[0117] In one example, when performing clustering processing, a clustering method in machine learning in related technologies can be used, such as K-means clustering, mean shift clustering, density-based clustering method, etc.

[0118] In one example, step S203 comprises the following steps:

[0119] The first step of step S203: performing filtering processing on the first vector set, and determining K first feature vectors obtained after filtering as K first center vectors; K is a positive integer; wherein each first center vector corresponds to a first cluster.

[0120] The second step of step S203: for each first feature vector, determine the first difference value result of the first feature vector in the first vector set and the K first center vectors respectively; determine the first class cluster of the first center vector corresponding to the first difference value result with the smallest value as the first class cluster to which the first feature vector belongs.

[0121] The third step of step S203: for each first class cluster, determine the first difference value sum corresponding to the first feature vectors in the first class cluster, wherein the first difference value sum is the sum of the difference values of the first feature vector and each first feature vector in the first class cluster; determine the second difference value sum of the first center vector corresponding to the first class cluster and each first feature vector in the first class cluster; based on the first difference value sum and the second difference value sum, determine the first clustering feature corresponding to the first class cluster.

[0122] The fourth step of step S203: determine the first class cluster and the first clustering feature of the first class cluster as the first clustering result.

[0123] Exemplarily, the embodiment provides a manner that can be used for clustering processing of the first vector set. When it is needed to cluster process the first vector set, first, K first feature vectors can be screened out from the first vector set, and the K first feature vectors are respectively taken as K first center vectors, wherein each first center vector one-to-one corresponds to a first class cluster, that is, the first feature vectors contained in the first vector set can be divided into K first class clusters, and then each first center vector corresponds to a first class cluster. Wherein, the value of K is not specifically limited in the embodiment, and can be determined according to the number of the first feature vectors contained in the first vector set, or can be artificially specified.

[0124] After the K first center vectors are determined, the first feature vectors in the first vector set can be respectively subtracted from the K first center vectors to obtain K first difference value results. Then, the sizes of the K first difference value results are compared, and the first class cluster corresponding to the first center vector corresponding to the first difference value result with the smallest value is taken as the first class cluster to which the first feature vector belongs. Further, the above steps are repeated for each first feature vector, and the first class cluster corresponding to each first feature vector can be determined.

[0125] After the first feature vectors included in each first-type cluster are determined, the first clustering feature corresponding to the first-type cluster can be further determined. Specifically, in this embodiment, when determining the first clustering feature, the sum of the differences between the first center vector corresponding to the first-type cluster and each first feature vector included in the first-type cluster can be determined as a second difference sum. The sum of the differences between any first feature vector in the first-type cluster and the remaining first feature vectors in the first-type cluster can be determined as a first difference sum. The first clustering feature corresponding to the first-type cluster can be determined by comparing the size of the first difference sum and the second difference sum. For example, when the difference between the first difference sum and the second difference sum is less than a preset threshold, the first center vector corresponding to the first-type cluster can be directly taken as the first clustering feature corresponding to the first-type cluster; otherwise, any first feature vector selected when determining the first difference sum can be taken as the first clustering feature.

[0126] In addition, in a possible implementation, the number of first difference sums can be multiple, that is, multiple first feature vectors can be selected from the first-type cluster, and the first difference sum corresponding to the selected first feature vector can be determined. Then, the first clustering feature can be determined by comparing the size of the multiple first difference sums and the first center vector. The number of selected first feature vectors is not specifically limited, and can be part of the first feature vectors in the first vector set or all the feature vectors, which is not specifically limited here.

[0127] It can be understood that, in this embodiment, when performing clustering processing, the first center vector is pre-selected, and the first feature vector included in each first-type cluster is determined by calculating the difference from the first center vector. The above-mentioned method is relatively simple to implement, does not need to occupy many resources to store the clustering model, and is beneficial to reducing the storage resources of the electronic device. In addition, when determining the first clustering feature corresponding to the first-type cluster, the first clustering feature can be determined by comparing the first difference sum and the second difference sum, so that the determined first clustering feature can represent the feature of the first-type cluster.

[0128] In one example, when performing the third step of step S203, the following method can be used: if the first difference sum is less than the second difference sum, the first feature vector corresponding to the first difference sum is determined as the first clustering feature corresponding to the first-type cluster; if the first difference sum is greater than or equal to the second difference sum, the first center vector corresponding to the second difference sum is determined as the first clustering feature corresponding to the first-type cluster.

[0129] Exemplarily, when the first clustering feature corresponding to the first type of cluster is determined based on the first sum of differences and the second sum of differences, the vector corresponding to the minimum value between the first sum of differences and the second sum of differences can be taken as the first clustering feature corresponding to the first type of cluster in this embodiment, that is, when the first sum of differences is smaller than the second sum of differences, the first feature vector corresponding to the first sum of differences is taken as the first clustering feature, otherwise, the originally selected first center vector is taken as the first clustering feature of the first type of cluster.

[0130] It can be understood that in this embodiment, when the first clustering feature is determined, the vector corresponding to the smaller value between the first sum of differences and the second sum of differences can be taken as the first clustering feature, and thus the accuracy of the screened first clustering feature can be determined while clustering, so that the first clustering feature can better represent the characteristics of each first feature vector in the first type of cluster.

[0131] S204, determining a second vector set based on the service data of the information group and the device information of the service data, wherein the second vector set includes at least one second feature vector, the second feature vector includes the service data and the device information of the service data, and the service data included in different second feature vectors is generated at different times.

[0132] Exemplarily, after receiving the service information to be processed, for each information group in the service information, the service data in the information group and the device information corresponding to the service data can be combined to obtain the second vector set. For example, when an information group includes a piece of service data and the device information corresponding to the service data, the service data and the corresponding device information can be taken as a second feature vector at this time. When an information group includes multiple pieces of service data and device information corresponding to the multiple pieces of service data one by one, the service data and the corresponding device information can be taken as a second feature vector, and thus multiple second feature vectors are obtained, wherein the service data included in different second feature vectors is generated at different times, and the obtained multiple second feature vectors are taken as the second vector set.

[0133] S205, performing clustering processing on the second vector set to obtain a second clustering result, the second clustering result including at least one second type of cluster and a second clustering feature corresponding to each second type of cluster, and the second type of cluster including at least one second feature vector.

[0134] Exemplarily, after obtaining the second feature vector set, in order to avoid the problem that the subsequent electronic device has a large amount of calculation caused by a large amount of data in the second vector set, the second vector set is subjected to clustering processing in this embodiment, that is, the second feature vectors contained in the second vector set are subjected to clustering processing, and at least one second cluster after clustering is obtained, and each second cluster also has its corresponding second clustering feature. The second clustering feature can be understood as any first feature vector contained in the first cluster corresponding to the second clustering feature, which is not specifically limited in this embodiment.

[0135] In one example, when clustering processing is performed, a clustering method in machine learning in related technologies can be used, such as K-means clustering, mean shift clustering, density-based clustering method, etc.

[0136] It should be noted that the clustering processing of the second vector set can also refer to the principle of clustering processing of the first vector set in step S203, which will not be described here.

[0137] S206, determining the first clustering feature and the second clustering feature as the first service feature of the information group.

[0138] Exemplarily, after determining the first clustering feature of the first cluster in the first vector set and the second clustering feature of the second cluster in the second vector set, the first clustering feature and the second clustering feature can be combined to determine the first service feature of the information group.

[0139] For example, when the service category identifier is the geographic coordinates of the positioning service, at this time, the data in the information group collected by the user in the current period can include: service data d1: 180.3271, generation time t1 of service data d1: 0101, device information e1 of service data d1: 103; service data d2: 179.200, generation time t2 of service data d2: 0102, device information e2 of service data d2: 112; at this time, the above data can be converted into binary data and subjected to shift processing to form a first feature vector (d1 t1) and another first feature vector (d2 t2); then, a second feature vector (d1 e1) and a second feature vector (d2 e2) are formed; all the first feature vectors are clustered to obtain the corresponding first clustering feature, and all the second feature vectors are clustered to obtain the corresponding second clustering feature, and the vector composed of the obtained first clustering feature and the second clustering feature can be used as the first service feature of the information group.

[0140] It can be understood that, in the embodiment, by constructing the first vector set and the second vector set in the information group respectively, and then determining the first clustering feature between the service data and the generation time of the service data, and the second clustering feature between the service data and the device information of the service data in subsequent clustering, feature mining in multiple dimensions is realized, so that the obtained first service feature can better reflect the use habit of the service product currently used by the user, and the accuracy of the service product to be recommended pushed to the user is improved.

[0141] S207, determine the second service feature corresponding to the service information.

[0142] For example, the specific principle of step S207 can be referred to step S102, and the embodiment is not limited specifically.

[0143] S208, according to the second service feature and the preset identification model, determine the first probability information and the second probability information, wherein the first probability information is used to indicate the probability value of needing to use the service product to be recommended; the second probability information is used to indicate the probability value of not needing to use the service product to be recommended; and the preset identification model is used to determine the probability of needing to use the service product to be recommended and the probability of not needing to use the service product to be recommended.

[0144] For example, the specific principle of step S208 in the embodiment can be referred to step S103, which will not be repeated here.

[0145] In one example, the preset identification model is established based on the third service feature of each training service information and the first use information corresponding to each training service information; the third service feature of the training service information is determined based on the fourth service feature of the information group contained in the training service information; the fourth service feature of the information group is obtained based on the feature mining processing of the information group; and the first use information is used to represent whether the user has used the service product to be recommended under the training service information.

[0146] Exemplarily, in the present example, a manner of establishing a preset identification model is provided. When establishing a preset identification model corresponding to a service product to be recommended, a plurality of groups of historical service information to be trained can be acquired first, wherein the service information includes at least one information group, and the information group has a service category identifier; the information group includes service data, generation time of the service data, and device information of the service data. Then, fourth service features of the information group are obtained through feature mining processing of the information group included in each service information to be trained, and the fourth service features of the information group in the service information to be trained are taken as third service features of the service information to be trained. The implementation manner of the feature mining processing can refer to steps S202-S206, which will not be described here again. In addition, usage information of a user corresponding to each service information to be trained needs to be acquired, wherein the first usage information is used to represent whether the user has used the service product to be recommended within the service data generation time corresponding to each service information to be trained. That is, the service information to be trained corresponds to one first usage information, and the service information to be trained corresponds to one third service feature. Then, based on the third service feature and the first usage information, model training is performed, and then a preset identification model corresponding to the service product to be recommended is obtained.

[0147] It can be understood that the present embodiment provides a training manner of a preset identification model, so that subsequent determination can be made on whether to recommend a service product to be recommended to a user based on the trained preset identification model and service information to be processed of the user, so as to improve user satisfaction.

[0148] In one example, the preset identification model is established based on a Bayesian formula; wherein the third probability information, the fourth probability information, the fifth probability information, and the sixth probability information included in the preset model are determined based on the third service features of the service information to be trained and the first usage information corresponding to the service information to be trained; wherein the third probability information is used to represent the probability of occurrence of the first usage information representing use of the service product to be recommended in the first usage information; the fourth probability information is used to represent the probability of occurrence of the first usage information representing non-use of the service product to be recommended in the first usage information; the fifth probability information is used to represent the probability of occurrence of the third service feature in the case of use of the service product to be recommended; and the sixth probability information is used to represent the probability of occurrence of the third service feature in the case of non-use of the service product to be recommended.

[0149] Exemplarily, the preset identification model in the present embodiment is established based on a Bayesian formula.

[0150] For example, the following formula is a model formula of a preset identification model provided by the present embodiment:

[0151]

[0152] wherein S is used to represent the first usage information, when S takes the value of 1, it represents that the user uses the service product to be recommended; when S takes the value of 0, it represents that the user does not use the service product to be recommended. F is used to represent the third service feature corresponding to the service information to be trained. Wherein, P(S=1) can be used as the third probability information, that is, in the first usage information corresponding to each service information to be trained, the probability of the first usage information representing that the service product to be recommended is used; P(S=1) can be used as the fourth probability information, that is, in the first usage information corresponding to each service information to be trained, the probability of the first usage information representing that the service product to be recommended is not used;

[0153] P(F|S=1) can be used as the fifth probability information, that is, representing the probability of the third service feature F appearing under the condition that the user uses the service product to be recommended; P(F|S=1) can be used as the sixth probability information, that is, representing the probability of the third service feature F appearing under the condition that the user does not use the service product to be recommended.

[0154] After obtaining the third service feature F and the first usage information S corresponding to each service information to be trained, the fifth probability information and the sixth probability information corresponding to each third service feature F can be determined according to the meanings of the above-mentioned probability information.

[0155] After obtaining the second service feature corresponding to the service information to be processed, the second service feature can be matched with the third service feature to find the third service feature that is the same as the second service feature (here, the same can be understood as that the similarity between the two service features is higher than a preset value), and then based on the fifth probability information and the sixth probability information corresponding to the third service feature, the third probability information and the fourth probability information determined in the training stage, and substituted into the above-mentioned formula, P(S=0|F) and P(S=1|F) can be obtained, wherein P(S=0|F) can be regarded as the second probability information in the present application, and P(S=1|F) can be regarded as the first probability information in the present application.

[0156] In addition, in some embodiments, due to the limitation of the data amount of the service information to be trained, the types of the third service features obtained in the training process are also limited, at this time, the types of the third service features can be expanded, and the fifth probability information and the sixth probability information corresponding to the expanded third service features can be determined. When determining the above-mentioned probability information, since the values of each dimension in the service feature F are independent of each other, the probability value of each dimension can be obtained by using the probability formula of independent events, that is, it can be obtained by the following way:

[0157]

[0158] wherein D can be regarded as the number of information groups contained in the service information, F i characterizes the fourth service feature corresponding to the i-th information group, dim(F) characterizes the dimension of the third service feature, F k is used to characterize the value corresponding to the k-th dimension in the third service feature F. P(F) can be regarded as the probability of the third service feature. Further, the original third service feature obtained based on the service information to be trained can be extended by the calculation manner of the probability of the above-mentioned independent event. Wherein, the P(F k ) corresponding to the extended third service feature can be determined by counting based on the value of the k-th dimension in the existing third service feature. The P(F|S=1) and P(F|S=0) corresponding to the extended third service feature can also be obtained by the principle of splitting and multiplying product provided in the above formula, which will not be described here.

[0159] wherein dim(F) can be represented by the following formula:

[0160]

[0161] wherein K1 is the number of the first type of clusters corresponding to the service information to be trained, K2 is the number of the second type of clusters corresponding to the service information to be trained, x i characterizes the dimension of the service data contained in the i-th information group in the service information to be trained, m1 is used to characterize the dimension of the generation time of the service data, and m2 is used to characterize the dimension of the device information of the service data.

[0162] It can be understood that, in the embodiment, the preset recognition model is trained by using the Bayesian work, which is relatively simple to implement, so that whether the service product to be recommended is recommended to the user can be determined based on the preset recognition model subsequently.

[0163] In one example, on the basis of the above-mentioned example, before step S208, the following steps can also be included:

[0164] First step: if it is determined that the data dimension of the device information in the service information to be processed is greater than the data dimension of the device information in the service information to be trained obtained in the preset model establishment process, and the service product to be recommended is not changed, then the third service feature corresponding to the service information to be trained is adjusted according to the data dimension of the device information in the service information to be processed, to obtain an adjusted third service feature.

[0165] The second step is to adjust the fifth probability information and the sixth probability information corresponding to the third service feature of the service information to be trained based on the adjusted third service feature.

[0166] The third step is to update the preset identification model based on the third probability information, the fourth probability information, the adjusted fifth probability information, the adjusted sixth probability information, and the Bayes formula.

[0167] For example, the embodiment provides a preset model optimization method. In actual application, as the number of users using service products increases, the number of devices also increases, and the data dimension of the corresponding device information also increases. For example, when the number of devices increases, the original device information is only required to be supplemented with 0 at the increased binary bit (for example, the original device information is 11010100, and if the device information is increased by four bits, it becomes 000011010100).

[0168] That is, the third service feature in the service information to be trained obtained in the preset model establishment process can be modified based on the data dimension of the device information in the currently obtained service information to be processed, and the adjusted third service feature is obtained. Then, according to the definition of the fifth probability information and the sixth probability information mentioned in the above example, the fifth probability information and the sixth probability information corresponding to each adjusted third service feature are determined in the plurality of adjusted third service features by counting, that is, the adjusted fifth probability information and the adjusted sixth probability information.

[0169] In a possible implementation, for the expanded third service feature, the expanded and adjusted third service feature is obtained after adjusting the data dimension of the device information. Since the values of each dimension in the expanded and adjusted third service feature are independent, the probability corresponding to the expanded and adjusted third service feature can be represented by the following formula:

[0170]

[0171] Wherein, F' represents the expanded and adjusted third service feature, L is the change amount of the device information data dimension. P(F l ) represents the probability value of the dimension corresponding to the data dimension adjustment part. That is, in determining the probability corresponding to the expanded and adjusted third service feature, a part of the result obtained in the previous training process can be directly reused, and P(F l ) can be directly determined in the adjusted third service feature by using the technical means.

[0172] Similarly, the P(F' | S=1) corresponding to the third service feature after the expansion adjustment (i.e., the fifth probability information after the adjustment corresponding to the third service feature after the expansion adjustment) and the P(F' | S=0) (i.e., the sixth probability information after the adjustment corresponding to the third service feature after the expansion adjustment) can be based on the principle shown in the above formula, that is, the dimension of the part before the adjustment is directly reused, and the part added by the dimension adjustment can be determined by counting in the third service feature after the adjustment, and then the update of the preset recognition model is completed, so that the prediction of the first probability information and the second probability information can be based on the updated preset recognition model in the future.

[0173] It can be understood that when the device information data dimension is adjusted but the service product to be recommended is not changed in the embodiment, the preset recognition model can be updated by the above-mentioned manner, without the need to repeatedly obtain new service information to be trained, thereby improving the optimization efficiency of the preset recognition model.

[0174] In one example, on the basis of the above-mentioned example, before step S208, the following steps can also be included: if it is determined that the data dimension in the service information to be processed is greater than the data dimension in the service information to be trained obtained in the preset model establishment process, and the service product to be recommended is changed, then the third service feature corresponding to the service information to be trained is adjusted according to the data dimension in the service information to be processed, to obtain the third service feature after the adjustment; the second usage information corresponding to the service information to be trained is obtained, wherein the second usage information is used to represent whether the service product to be recommended after the change is used; the third probability information, the fourth probability information after the adjustment, the fifth probability information after the adjustment, the sixth probability information after the adjustment are determined according to the second usage information and the service feature information after the adjustment of the service information to be trained; and the preset recognition model is updated based on the third probability information after the adjustment, the fourth probability information after the adjustment, the fifth probability information after the adjustment, the sixth probability information after the adjustment and the Bayes formula.

[0175] For example, in the embodiment, when the service product to be recommended and the data dimension in the service information to be processed are changed, the second usage information corresponding to the service information to be trained needs to be reacquired, wherein the second usage information is used to represent whether the service product to be recommended after the change is used.

[0176] Moreover, the third service feature can be adjusted according to the determination of the data dimension in the service information to be processed, and specifically, the service data dimension in the third service feature can be adjusted, the dimension of the generation time of the service data in the third service feature can be adjusted, and the dimension of the device information of the service data in the third service feature can be adjusted, and then the third service feature after the adjustment is obtained.

[0177] Afterwards, the adjusted third service feature is determined according to the adjusted third service feature and the second use information, the adjusted fifth probability information and the adjusted sixth probability information corresponding to the adjusted third service feature are determined, and the adjusted third probability information and the adjusted fourth probability information are determined. The determination of the adjusted third probability information, the adjusted fourth probability information, the adjusted fifth probability information, and the adjusted sixth probability information can be determined according to the above-mentioned preset identification model training process, and will not be repeated here.

[0178] In addition, when determining the adjusted fifth probability information, the probability information obtained in the preset identification model training process can also be considered, that is, in the dimension of service data that has not changed, in the dimension of service data generation time that has not changed, and in the dimension of service data device information that has not changed, the above three unchanged data can directly reuse the previous calculation result, and the changed one can be recalculated. After multiplying the above results, the adjusted fifth probability information to be obtained can be obtained.

[0179] Further, the adjusted third probability information, the adjusted fourth probability information, the adjusted fifth probability information, the adjusted sixth probability information, and the Bayesian formula are updated to update the preset identification model, so as to determine whether to recommend the current changed service product to the user based on the updated preset identification model.

[0180] It can be understood that when the service product to be recommended needs to be changed, the embodiment provides an optimization method of the preset identification model, which can be adjusted based on the preset identification model corresponding to the remaining service product to be recommended to obtain the identification model corresponding to the service product to be recommended to be recommended. The unchanged part can directly reuse the result of the original model, and the time consumed in the model training process can be reduced.

[0181] S209, if the first probability information is greater than the second probability information, the service product to be recommended is recommended to the user.

[0182] Exemplarily, the specific principle of step S209 can be referred to step S104, and will not be repeated here.

[0183] In this embodiment, by constructing the first vector set and the second vector set in the information group respectively, and then in subsequent clustering, the first clustering feature between the service data and the generation time of the service data, and the second clustering feature between the service data and the device information of the service data can be determined, realizing feature mining in multiple dimensions, so that the obtained first service feature can better reflect the use habit of the service product currently used by the user, and then the accuracy of the service product to be recommended pushed to the user is improved.

[0184] Figure 4 A structural schematic diagram of a service product recommendation device is provided for the embodiments of the present application, as shown in the figure, the device comprises: Figure 4

[0185] The first acquisition unit 401 is configured to acquire service information to be processed of a user, wherein the service information comprises at least one information group, and the information group has a service category identifier; the information group comprises service data, generation time of the service data, and device information of the service data.

[0186] The processing unit 402 is configured to perform feature mining processing on the information group to obtain a first service feature of the information group.

[0187] The first determination unit 403 is configured to determine a second service feature corresponding to the service information.

[0188] The second determination unit 404 is configured to determine first probability information and second probability information according to the second service feature and a preset identification model, wherein the first probability information is used to indicate a probability value of needing to use a service product to be recommended; the second probability information is used to indicate a probability value of not needing to use the service product to be recommended; and the preset identification model is used to determine the probability of needing to use the service product to be recommended and the probability of not needing to use the service product to be recommended.

[0189] The pushing unit 405 is configured to recommend the service product to be recommended to the user if it is determined that the first probability information is greater than the second probability information.

[0190] The device provided in the embodiments is used to realize the technical solutions provided by the above method, and has similar implementation principles and technical effects, which will not be described herein.

[0191] Figure 5 A structural schematic diagram of another service product recommendation device is provided for the embodiments of the present application, and on the basis of the above embodiments, the processing unit 402 in this embodiment comprises:

[0192] ​The first determining module 4021 is configured to determine a first vector set based on the information group service data and the generation time of the service data, where the first vector set includes at least one first feature vector, the first feature vector includes the service data and the generation time of the service data, and the generation time of the service data included in different first feature vectors is different.

[0193] The first processing module 4022 is configured to perform clustering processing on the first vector set to obtain a first clustering result, where the first clustering result includes at least one first cluster and a first clustering feature corresponding to the first cluster, and the first cluster includes at least one first feature vector.

[0194] The second determining module 4023 is configured to determine a second vector set based on the information group service data and the device information of the service data, where the second vector set includes at least one second feature vector, the second feature vector includes the service data and the device information of the service data, and the generation time of the service data included in different second feature vectors is different.

[0195] The second processing module 4024 is configured to perform clustering processing on the second vector set to obtain a second clustering result, where the second clustering result includes at least one second cluster and a second clustering feature corresponding to the second cluster, and the second cluster includes at least one second feature vector.

[0196] The third determining module 4025 is configured to determine that the first clustering feature and the second clustering feature are first service features of the information group.

[0197] In a possible implementation, the first processing module 4022 is specifically configured to perform filtering processing on the first vector set, and determine K first center vectors obtained after the filtering processing, where K is a positive integer, each first center vector corresponds to a first cluster, for each first feature vector, determine first difference value results of the first feature vector and the K first center vectors in the first vector set, determine a first cluster corresponding to a first center vector with a minimum first difference value result as a first cluster to which the first feature vector belongs, for each first cluster, determine a first difference value sum corresponding to the first feature vector in the first cluster, where the first difference value sum is a sum of difference values of the first feature vector and each first feature vector in the first cluster, determine a second difference value sum of the first center vector corresponding to the first cluster and each first feature vector in the first cluster, determine a first clustering feature corresponding to the first cluster based on the first difference value sum and the second difference value sum, and determine the first cluster and the first clustering feature of the first cluster as the first clustering result.

[0198] In a possible implementation, the first processing module 4022 is specifically configured to: if the first sum of differences is less than the second sum of differences, determine that the first feature vector corresponding to the first sum of differences is the first clustering feature corresponding to the first class cluster; and if the first sum of differences is greater than or equal to the second sum of differences, determine that the first center vector corresponding to the second sum of differences is the first clustering feature corresponding to the first class cluster.

[0199] In a possible implementation, the preset identification model is established based on the third service feature of each piece of service information to be trained and the first use information corresponding to each piece of service information to be trained; the third service feature of the service information to be trained is determined based on the fourth service feature of the information group contained in the service information to be trained; the fourth service feature of the information group is obtained based on feature mining processing on the information group; and the first use information is used to represent whether the user has used the service product to be recommended under the service information to be trained.

[0200] In a possible implementation, the preset identification model is established based on a Bayesian formula; the third probability information, the fourth probability information, the fifth probability information, and the sixth probability information included in the preset model are determined based on the third service feature of each piece of service information to be trained and the first use information corresponding to each piece of service information to be trained; the third probability information is used to represent the probability of occurrence of the first use information representing use of the service product to be recommended in each piece of first use information; the fourth probability information is used to represent the probability of occurrence of the first use information representing non-use of the service product to be recommended in each piece of first use information; the fifth probability information is used to represent the probability of occurrence of the third service feature in the case of use of the service product to be recommended; and the sixth probability information is used to represent the probability of occurrence of the third service feature in the case of non-use of the service product to be recommended.

[0201] In a possible implementation, the apparatus further includes:

[0202] The third determination unit 406 is configured to: if it is determined that the data dimension of the device information in the service information to be processed is greater than the data dimension of the device information in the service information to be trained obtained in the preset model establishment process, and the service product to be recommended is not changed, adjust the third service feature corresponding to the service information to be trained according to the data dimension of the device information in the service information to be processed, to obtain an adjusted third service feature.

[0203] The first adjustment unit 407 is configured to: based on the adjusted third service feature, adjust the fifth probability information and the sixth probability information of the third service feature corresponding to the service information to be trained, to obtain adjusted fifth probability information and adjusted sixth probability information.

[0204] The first updating unit 408 is configured to update the preset identification model based on the third probability information, the fourth probability information, the adjusted fifth probability information, the adjusted sixth probability information, and the Bayes formula.

[0205] In a possible implementation, the apparatus further includes:

[0206] The fourth determining unit 409 is configured to, if it is determined that the data dimension in the service information to be processed is greater than the data dimension in the service information to be trained obtained in the preset model establishment process, and the service product to be recommended changes, adjust the third service feature corresponding to the service information to be trained according to the data dimension in the service information to be processed, to obtain an adjusted third service feature.

[0207] The second obtaining unit 410 is configured to obtain second usage information corresponding to the service information to be trained, where the second usage information is used to represent whether the service product to be recommended after the change is used.

[0208] The fifth determining unit 411 is configured to determine the third probability information, the adjusted fourth probability information, the adjusted fifth probability information, and the adjusted sixth probability information according to the second usage information and the adjusted service feature information to be trained.

[0209] The second updating unit 412 is configured to update the preset identification model based on the adjusted third probability information, the adjusted fourth probability information, the adjusted fifth probability information, the adjusted sixth probability information, and the Bayes formula.

[0210] In a possible implementation, the information group is obtained by performing cyclic shift processing on an initial information group; the initial information group includes initial service data, a generation time of the initial service data, and device information of the initial service data.

[0211] The apparatus provided in the embodiment is used to implement the technical solutions provided by the above method, and has similar implementation principles and technical effects, which will not be described herein.

[0212] The present application provides an electronic device, comprising: a memory, a processor;

[0213] The memory; the memory for storing the processor executable instructions;

[0214] The processor is configured to execute the method according to the executable instructions.

[0215] Figure 6 A structural schematic diagram of an electronic device provided in the embodiment of the present application is shown in FIG. 1. Figure 6 As shown in FIG. 1, the electronic device includes:

[0216] The electronic device further includes a processor 291 and a memory 292. The electronic device can further include a communication interface 293 and a bus 294. The processor 291, the memory 292, and the communication interface 293 can communicate with each other through the bus 294. The communication interface 293 can be used for information transmission. The processor 291 can invoke the logical instructions in the memory 292 to execute the method of the above-described embodiments.

[0217] In addition, the logical instructions in the memory 292 described above can be implemented in the form of a software functional unit and sold or used as an independent product, which can be stored in a computer readable storage medium.

[0218] The memory 292, as a computer readable storage medium, can be used to store software programs, computer executable programs, such as program instructions / modules corresponding to the method in the embodiments of the present application. The processor 291 executes the functions and data processing by running the software programs, instructions and modules stored in the memory 292, that is, implements the method in the above-described method embodiments.

[0219] The memory 292 can include a program storage area and a data storage area. The program storage area can store an operating system and at least one application required by a function. The data storage area can store data created during use of the terminal device, and the like. In addition, the memory 292 can include a high-speed random access memory, and can further include a nonvolatile memory.

[0220] The present application provides a computer readable storage medium, and the computer readable storage medium stores computer execution instructions. The computer execution instructions are executed by a processor to implement any one of the methods.

[0221] The present application provides a computer program product, and the computer program product includes a computer program. The computer program is executed by a processor to implement any one of the methods.

[0222] Other embodiments of the present application will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. The application is intended to cover any variations, uses, or adaptations of the application following, in general, the principles of the application and including such departures from the present disclosure as come within known or customary practice in the art to which the application pertains. The specification and examples are to be regarded as exemplary only, and the true scope and spirit of the application are indicated by the appended claims.

[0223] It is to be understood that the application is not limited to the precise construction already described above and shown in the drawings, and that various modifications and changes can be made by those skilled in the art without departing from the scope of the application. The scope of the application should only be limited by the claims appended hereto.

Claims

1. A method of recommending a service product, characterized by, The method comprises the following steps: obtaining service information to be processed of a user, wherein the service information comprises at least one information group obtained by cyclically shifting an initial information group, and the information group has a service category identifier; the information group comprises service data, generation time of the service data, and device information of the service data; performing feature mining processing on the information group to obtain first service features of the information group, and taking the first service features corresponding to each information group included in the service information to be processed as second service features corresponding to the service information to be processed; determining first probability information and second probability information according to the second service features and a preset identification model, wherein the first probability information is used to indicate a probability value of needing to use a service product to be recommended, the second probability information is used to indicate a probability value of not needing to use the service product to be recommended, the preset identification model is used to determine a probability of needing to use the service product to be recommended and a probability of not needing to use the service product to be recommended, and the preset identification model is established based on a Bayesian formula; if the first probability information is greater than the second probability information, recommending the service product to be recommended to the user; performing feature mining processing on the information group to obtain first service features of the information group, comprising: determining a first vector set based on the service data of the information group and the generation time of the service data, wherein the first vector set comprises at least one first feature vector, the first feature vector comprises the service data and the generation time of the service data, and the generation time of the service data included in different first feature vectors is different; performing clustering processing on the first vector set to obtain a first clustering result, wherein the first clustering result comprises at least one first cluster and a first clustering feature corresponding to the first cluster, and the first cluster comprises at least one first feature vector; determining a second vector set based on the service data of the information group and the device information of the service data, wherein the second vector set comprises at least one second feature vector, the second feature vector comprises the service data and the device information of the service data, and the generation time of the service data included in different second feature vectors is different; performing clustering processing on the second vector set to obtain a second clustering result, wherein the second clustering result comprises at least one second cluster and a second clustering feature corresponding to the second cluster, and the second cluster comprises at least one second feature vector; determining the first clustering feature and the second clustering feature as the first service features of the information group.

2. The method of claim 1, wherein, performing clustering processing on the first vector set to obtain a first clustering result, comprising: performing screening processing on the first vector set, and determining K first feature vectors obtained after screening as K first center vectors; K is a positive integer; wherein each first center vector corresponds to a first cluster. For each first feature vector, a first difference result of the first feature vector in the first vector set and the K first center vectors is determined respectively; a first center vector corresponding to a first difference result with a minimum value is determined as a first class cluster to which the first feature vector belongs. For each first class cluster, a first difference sum corresponding to the first feature vectors in the first class cluster is determined, wherein the first difference sum is a sum of differences between the first feature vectors and each first feature vector in the first class cluster; a second difference sum of the first center vector corresponding to the first class cluster and each first feature vector in the first class cluster is determined; and a first clustering feature corresponding to the first class cluster is determined based on the first difference sum and the second difference sum. The first class cluster and the first clustering feature of the first class cluster are determined as the first clustering result.

3. The method of claim 2, wherein, Determining the first clustering feature corresponding to the first class cluster based on the first difference sum and the second difference sum comprises: If the first difference sum is less than the second difference sum, the first feature vector corresponding to the first difference sum is determined as the first clustering feature corresponding to the first class cluster; If the first difference sum is greater than or equal to the second difference sum, the first center vector corresponding to the second difference sum is determined as the first clustering feature corresponding to the first class cluster.

4. The method of claim 1, wherein, The preset identification model is established based on third service features of each to-be-trained service information and first use information corresponding to each to-be-trained service information; the third service feature of the to-be-trained service information is determined based on a fourth service feature of an information group contained in the to-be-trained service information; the fourth service feature of the information group is obtained based on feature mining processing on the information group; and the first use information is used to represent whether the user has used the to-be-recommended service product under the to-be-trained service information.

5. The method of claim 4, wherein, The third probability information, the fourth probability information, the fifth probability information, and the sixth probability information included in the preset identification model are determined based on the third service features of each to-be-trained service information and the first use information corresponding to each to-be-trained service information; The third probability information is used to represent a probability of occurrence of the first use information representing use of the to-be-recommended service product in each first use information; the fourth probability information is used to represent a probability of occurrence of the first use information representing non-use of the to-be-recommended service product in each first use information; the fifth probability information is used to represent a probability of occurrence of the third service feature in the case of use of the to-be-recommended service product; and the sixth probability information is used to represent a probability of occurrence of the third service feature in the case of non-use of the to-be-recommended service product.

6. The method of claim 5, wherein, The method further comprises: If it is determined that the data dimension of the device information in the to-be-processed service information is greater than the data dimension of the device information in the to-be-trained service information obtained in the preset model establishment process, and the to-be-recommended service product does not change, then the third service feature corresponding to the to-be-trained service information is adjusted according to the data dimension of the device information in the to-be-processed service information, to obtain an adjusted third service feature; Based on the adjusted third service feature, the fifth probability information and the sixth probability information of the third service feature corresponding to the to-be-trained service information are adjusted, to obtain adjusted fifth probability information and adjusted sixth probability information; The preset identification model is updated based on the third probability information, the fourth probability information, the adjusted fifth probability information, the adjusted sixth probability information, and a Bayes formula.

7. The method of claim 5, wherein, The method further includes: If it is determined that the data dimension in the to-be-processed service information is greater than the data dimension in the to-be-trained service information obtained in the preset model establishment process, and the to-be-recommended service product changes, then the third service feature corresponding to the to-be-trained service information is adjusted according to the data dimension in the to-be-processed service information, to obtain an adjusted third service feature; Second usage information corresponding to the to-be-trained service information is obtained, where the second usage information is used to represent whether the changed to-be-recommended service product is used or not; The third probability information, the adjusted third probability information, the adjusted fourth probability information, the adjusted fifth probability information, and the adjusted sixth probability information are determined according to the second usage information and the adjusted to-be-trained service feature information; The preset identification model is updated based on the adjusted third probability information, the adjusted fourth probability information, the adjusted fifth probability information, the adjusted sixth probability information, and a Bayes formula.

8. The method according to any one of claims 1-7, characterized in that, The initial information group includes initial service data, a generation time of the initial service data, and device information of the initial service data.

9. A service product recommendation apparatus characterized by comprising: It includes: A first obtaining unit is configured to obtain to-be-processed service information of a user, where the service information includes at least one information group, the information group is obtained by performing cyclic shift processing on an initial information group, the information group has a service category identifier, and the information group includes service data, a generation time of the service data, and device information of the service data. A processing unit is configured to perform feature mining processing on the information group, to obtain first service features of the information group. A first determining unit is configured to take the first service features corresponding to each information group included in the to-be-processed service information as second service features corresponding to the to-be-processed service information. The second determining unit is configured to determine first probability information and second probability information according to the second service feature and a preset identification model, wherein the first probability information is used to indicate a probability value of needing to use a service product to be recommended; the second probability information is used to indicate a probability value of not needing to use the service product to be recommended; the preset identification model is used to determine a probability of needing to use the service product to be recommended and a probability of not needing to use the service product to be recommended, and the preset identification model is established based on a Bayesian formula; The pushing unit is configured to recommend the service product to be recommended to the user if it is determined that the first probability information is greater than the second probability information. The processing unit is specifically configured to determine a first vector set based on service data of the information group and generation time of the service data, wherein the first vector set includes at least one first feature vector, the first feature vector includes service data and generation time of the service data, and different first feature vectors include different generation time of service data. The first vector set is subjected to clustering processing to obtain a first clustering result, the first clustering result includes at least one first cluster and a first clustering feature corresponding to the first cluster in one-to-one manner, and the first cluster includes at least one first feature vector. A second vector set is determined based on service data of the information group and device information of the service data, wherein the second vector set includes at least one second feature vector, the second feature vector includes service data and device information of the service data, and different second feature vectors include different generation time of service data. The second vector set is subjected to clustering processing to obtain a second clustering result, the second clustering result includes at least one second cluster and a second clustering feature corresponding to the second cluster, and the second cluster includes at least one second feature vector. The first clustering feature and the second clustering feature are determined as the first service feature of the information group.

Images