A method, apparatus, device and storage medium for determining an image

Abstract

Embodiments of the present application disclose a method, device and equipment for determining a portrait, and a storage medium. The method comprises: obtaining demand information of a target operator and characteristic data of a cross-network user; determining a target service index corresponding to the demand information according to a preset corresponding relationship between demand information and service indexes; extracting target characteristic data corresponding to the target service index from the characteristic data; inputting the target characteristic data corresponding to the target service index into a portrait model to determine a portrait of the cross-network user; and wherein the portrait model is trained by using the target service index, sample portraits of a plurality of users of the target operator, and characteristic data corresponding to the target service index of each user. The method provided by the embodiments can improve the accuracy of the portrait of the cross-network user.

Description

Technical Field

[0001] This application belongs to the field of user profiling technology, and in particular relates to a method, apparatus, device and storage medium for determining a profile. Background Technology

[0002] User profiles reveal a user's business needs. For example, if a user's profile identifies them as a high-value user, then that user requires high-value services. If operators develop high-value services for users with high-value profiles, it will improve user satisfaction. To improve user satisfaction, operators need to develop services based on user needs; therefore, determining user profiles before developing services is essential.

[0003] For users on other networks of an operator, the profile of these users is generally estimated by taking into account factors such as the number of permanent residents in a certain area, the local context, and the measured signals from other frequencies.

[0004] Because the indicators used to determine the profiles of users on different networks are complex and some indicators interfere with the profiles, the accuracy of the determined profiles of users on different networks is poor. Summary of the Invention

[0005] This application provides a method, apparatus, device, and storage medium for determining user profiles, which can improve the accuracy of user profiles on different networks.

[0006] In a first aspect, embodiments of this application provide a method for determining an image, the method comprising:

[0007] Obtain the target operator's demand information and the characteristic data of users on other networks;

[0008] Based on the pre-defined correspondence between demand information and business indicators, determine the target business indicators corresponding to the demand information;

[0009] Extract target feature data corresponding to the target business metrics from the feature data;

[0010] Input the target feature data corresponding to the target business indicators into the profiling model to determine the profile of users from other networks;

[0011] The profiling model is trained using target business metrics, sample profiles of each user from multiple users of the target operator, and feature data corresponding to the target business metrics of each user.

[0012] In one possible implementation, the profiling model includes a first neural network and a second neural network; the first neural network includes multiple first sub-neural networks, each of which is of a different type; target feature data corresponding to target business indicators are input into the profiling model to determine the profile of users from other networks, including:

[0013] The target feature data is input into multiple first sub-neural networks respectively to obtain the portrait probability corresponding to the portrait feature output by each first sub-neural network;

[0014] The profile probability corresponding to the profile feature output by each first sub-neural network is input into the second neural network to obtain the profile of the user from the other network.

[0015] In one possible implementation, the target business metric includes multiple influencing factors, and the target feature data corresponding to the target business metric includes data corresponding to each of the multiple influencing factors; the method also includes:

[0016] Obtain the coefficients corresponding to each first sub-neural network;

[0017] Based on the probabilities and coefficients of each first sub-neural network, the first neural network with the largest first influence weight on the image is determined among multiple first sub-neural networks, and the first sub-neural network with the largest first influence weight on the image is taken as the target neural network.

[0018] Obtain the weights corresponding to each influencing factor from the target neural network;

[0019] Based on the weight of each influencing factor and the data corresponding to each influencing factor, at least one influencing factor that meets the preset conditions is determined.

[0020] In one possible implementation, based on the probabilities and coefficients of each first sub-neural network, the first sub-neural network with the largest first influence weight on the image is determined from among multiple first sub-neural networks, including:

[0021] Based on the probabilities and coefficients of each first sub-neural network, determine the first influence weight of each first sub-neural network on the image.

[0022] Based on the first influence weight of each first sub-neural network on the image, determine the first sub-neural network with the largest first influence weight among multiple first sub-neural networks.

[0023] In one possible implementation, based on the weight of each influencing factor and the data corresponding to each influencing factor, at least one influencing factor that satisfies preset conditions is determined, including:

[0024] Based on the weight of each influencing factor and the data corresponding to each influencing factor, determine the second influence weight of each influencing factor on the probability of the profile output by the target neural network.

[0025] Based on the second influence weight of each influencing factor on the profile probability output by the target neural network, at least one influencing factor that meets the preset conditions is determined.

[0026] In one possible implementation, before inputting the target feature data corresponding to the target business indicators into the profiling model to obtain the profile of users from other networks, the method further includes:

[0027] Obtain sample profiles for each user among multiple users of the target operator, as well as feature data corresponding to the target business metrics for each user;

[0028] A profile model is trained using target business metrics, sample profiles of each user from multiple users of the target operator, and feature data corresponding to the target business metrics of each user.

[0029] Secondly, embodiments of this application provide an apparatus for determining an image, the apparatus comprising:

[0030] The acquisition module is used to acquire the target operator's demand information and the characteristic data of users from other networks.

[0031] The determination module is used to determine the target business indicators corresponding to the pre-defined correspondence between demand information and business indicators.

[0032] The extraction module is used to extract target feature data corresponding to the target business indicators from the feature data;

[0033] The determination module is also used to input the target feature data corresponding to the target business indicators into the profiling model to determine the profile of users from other networks;

[0034] The profiling model is trained using target business metrics, sample profiles of each user from multiple users of the target operator, and feature data corresponding to the target business metrics of each user.

[0035] In one possible implementation, the profiling model includes a first neural network and a second neural network; the first neural network includes multiple first sub-neural networks, and each of the multiple first sub-neural networks is of a different type.

[0036] The module is specifically used for:

[0037] The target feature data is input into multiple first sub-neural networks respectively to obtain the portrait probability corresponding to the portrait feature output by each first sub-neural network;

[0038] The profile probability corresponding to the profile feature output by each first sub-neural network is input into the second neural network to obtain the profile of the user from the other network.

[0039] In one possible implementation, the target business indicator includes multiple influencing factors, and the target feature data corresponding to the target business indicator includes the data corresponding to each of the multiple influencing factors.

[0040] The acquisition module is also used to acquire the coefficients corresponding to each first sub-neural network;

[0041] The determination module is also used to determine the first neural network with the largest first influence weight on the first image among multiple first neural networks based on the image probability and coefficient of each first sub-neural network, and to take the first sub-neural network with the largest first influence weight as the target neural network.

[0042] The acquisition module is also used to obtain the weights corresponding to each influencing factor from the target neural network;

[0043] The determination module is also used to determine at least one influencing factor that meets preset conditions based on the weight of each influencing factor and the data corresponding to each influencing factor.

[0044] In one possible implementation, a module is defined, specifically for:

[0045] Based on the probabilities and coefficients of each first sub-neural network, determine the first influence weight of each first sub-neural network on the image.

[0046] Based on the first influence weight of each first sub-neural network on the image, determine the first sub-neural network with the largest first influence weight among multiple first sub-neural networks.

[0047] In one possible implementation, a module is defined, specifically for:

[0048] Based on the weight of each influencing factor and the data corresponding to each influencing factor, determine the second influence weight of each influencing factor on the probability of the profile output by the target neural network.

[0049] Based on the second influence weight of each influencing factor on the profile probability output by the target neural network, and the data corresponding to each influencing factor, at least one influencing factor that meets the preset conditions is determined.

[0050] In one possible implementation, the acquisition module is also used to acquire sample profiles of each user among multiple users of the target operator, as well as feature data corresponding to the target business indicators of each user;

[0051] The device also includes a training module for training a profile model using target business metrics, sample profiles of each user among multiple users of the target operator, and feature data corresponding to the target business metrics of each user.

[0052] Thirdly, embodiments of this application provide an electronic device, the device including: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, it implements the method as described in the first aspect or any possible implementation of the first aspect.

[0053] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer program instructions that, when executed by a processor, implement the method as described in the first aspect or any possible implementation thereof.

[0054] Fifthly, embodiments of this application provide a computer program product in which instructions, when executed by a processor of an electronic device, cause the electronic device to perform a method as described in the first aspect or any possible implementation thereof.

[0055] This application provides a method, apparatus, device, and storage medium for determining user profiles. When a target operator needs profiles of users on other networks, the method involves acquiring the target operator's demand information and the characteristic data of the users on other networks; determining the target business indicators corresponding to the demand information based on a preset correspondence between the demand information and business indicators; extracting target feature data corresponding to the target business indicators from the feature data; and inputting the target feature data corresponding to the target business indicators into a profile model to determine the profiles of the users on other networks. Since the target business indicators are selected based on the demand information, and the profile model for determining the profile is trained using the target business indicators, sample profiles of users from the target operator, and the feature data corresponding to the user's target business indicators, even if the acquired feature data of the users on other networks cannot directly reflect the attribute characteristics of the users on other networks, the profile determined using the profile model and the target feature data corresponding to the target business indicators of the users on other networks can accurately reflect the attribute characteristics of the users on other networks required by the target operator, thereby improving the accuracy of the profiles of users on other networks. Attached Figure Description

[0056] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1 This is a schematic flowchart of a method for determining an image provided in an embodiment of this application;

[0058] Figure 2 This is a schematic diagram of a device for determining an image provided in an embodiment of this application;

[0059] Figure 3 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0060] The features and exemplary embodiments of various aspects of this application will now be described in detail. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only configured to explain this application and are not configured to limit this application. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples of this application.

[0061] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes the element.

[0062] User profiles reveal a user's business needs. For example, if a user's profile identifies them as a high-value user, then that user requires high-value services. If operators develop high-value services for users with high-value profiles, user satisfaction will improve. To improve user satisfaction, operators need to develop services tailored to user needs; therefore, determining user profiles before developing services is essential. For users on other networks, user profiles are typically estimated using factors such as the resident population in a region, the regional context, and measured inter-frequency signals. However, because the indicators used to determine inter-network user profiles are complex, and some indicators can interfere with the profile, the accuracy of the determined profiles is relatively poor.

[0063] This application provides a method, apparatus, device, and storage medium for determining user profiles. When a target operator needs profiles of users on other networks, the method involves acquiring the target operator's demand information and the characteristic data of the users on other networks; determining the target business indicators corresponding to the demand information based on a preset correspondence between the demand information and business indicators; extracting target feature data corresponding to the target business indicators from the feature data; and inputting the target feature data corresponding to the target business indicators into a profile model to determine the profiles of the users on other networks. Since the target business indicators are selected based on the demand information, and the profile model for determining the profile is trained using the target business indicators, sample profiles of users of the target operator, and the feature data corresponding to the target business indicators of the users, even if the acquired feature data of the users on other networks cannot directly reflect the attribute characteristics of the users on other networks, the profile determined by the profile model and the target feature data corresponding to the target business indicators of the users on other networks can accurately reflect the attribute characteristics of the users on other networks required by the target operator, thereby improving the accuracy of the profiles of users on other networks.

[0064] The method provided in this application is executed by a terminal with data acquisition and data processing functions, such as a server or computer.

[0065] The following will combine Figure 1 This application provides a detailed description of a method for determining an image, based on an embodiment of the present application.

[0066] like Figure 1 As shown, the method may include the following steps:

[0067] S110: Obtain the target operator's demand information and the characteristic data of users on other networks.

[0068] When defining the profile of users on other networks for a target operator, first obtain the target operator's demand information and the characteristic data of users on other networks.

[0069] The target operator's demand information can be either users from other networks who are profiled as high-value users, or users from other networks who are profiled as having poor user experience.

[0070] In one example, characteristic data of users from other networks is obtained from an internet company. The characteristic data of users from other networks may include characteristic data corresponding to at least one of the following indicators:

[0071] User-occupied cell identifier, user-occupied cell area code identifier, set of neighboring cell identifiers, operator identifier, abbreviation of physical cell identifier, timestamp, reported original longitude, reported original latitude, altitude, indoor / outdoor, whether it is mobile wireless broadband (Wireless-Fidelity, Wi-Fi), received signal strength indication, dynamic network type, reference signal received power, signal-to-interference-plus-noise ratio, reference signal received quality, cell identifier, base station identifier, set of neighboring reference signal received strengths, terminal temporary network identifier, user terminal brand, user terminal model, connected Wi-Fi name, connected Wi-Fi media access control (MAC) address, Wi-Fi signal strength, broadband operator, application (app) package name, app category, 5G synchronization signal strength, 5G Channel State Information (CSI) signal strength, 5G network configuration, 5G signal-to-interference-plus-noise ratio.

[0072] S120: Based on the pre-defined correspondence between demand information and business indicators, determine the target business indicators corresponding to the demand information.

[0073] The correspondence between demand information and business indicators is pre-defined. The pre-defined correspondence between demand information and business indicators is then searched to obtain the target business indicators corresponding to the demand information in the correspondence.

[0074] In one example, the demand information is for users from other networks whose profiles are high-value users. The corresponding target business metrics may include at least one of the following:

[0075] User attribute metrics, user business metrics, and user perception metrics.

[0076] The user attribute metrics are updated monthly, and these metrics may include user terminal brand, user terminal model, operator identifier, broadband operator, etc. User service metrics are updated weekly, and these metrics may include the daily average number of streaming media sampling points, the daily average number of World Wide Web (Web) sampling points, the daily average number of instant messaging sampling points, the daily average number of peer-to-peer (P2P) sampling points, the daily average proportion of Wi-Fi sampling points, and the daily average total number of sampling points. User perception metrics are also updated weekly, and these metrics may include the daily average signal strength, the daily average signal-to-interference-plus-noise ratio (SINR), the daily average proportion of weak coverage sampling points, and the daily average number of consecutive poor signal quality occurrences, etc.

[0077] S130, extract target feature data corresponding to the target business indicators from the feature data.

[0078] Extract feature data corresponding to the target business metrics from the feature data of users on other networks, and use the feature data corresponding to the target business metrics as the target feature data.

[0079] S140: Input the target feature data corresponding to the target business indicators into the profiling model to determine the profile of users on other networks.

[0080] Input the target feature data corresponding to the target business indicators into the profile model, and the profile model outputs the profile of users from other networks.

[0081] In one example, the target operator's required information could be users from other networks whose profiles are high-value users. In this case, the output profiles of the users from other networks would be either high-value users or non-high-value users. The profiles and identifiers of the users from other networks whose profiles are high-value users would be provided to the target operator.

[0082] In some embodiments, target feature data corresponding to the target business indicators in the previous period is extracted once in each period, and the target feature data corresponding to the target business indicators in the previous period is input into the profiling model to obtain the profile of the cross-network user determined in the current period.

[0083] The method provided in this application provides a method that periodically outputs profiles of users on other networks. It can update the profiles of users on other networks as their feature data changes, thereby improving the real-time performance and accuracy of the profiles of determined users on other networks.

[0084] The profiling model is trained using target business metrics, sample profiles of each user from multiple users of the target operator, and feature data corresponding to the target business metrics of each user.

[0085] The method provided in this application, when a target operator needs a profile of users on other networks, acquires the target operator's demand information and the characteristic data of the users on other networks; determines the target business indicators corresponding to the demand information based on the preset correspondence between the demand information and business indicators; extracts target feature data corresponding to the target business indicators from the feature data; and inputs the target feature data corresponding to the target business indicators into a profile model to determine the profile of the users on other networks. Since the target business indicators are selected based on the demand information, and the profile model for determining the profile is trained using the target business indicators, sample profiles of users from the target operator, and the feature data corresponding to the user's target business indicators, even if the acquired feature data of the users on other networks cannot directly reflect the attribute characteristics of the users on other networks, the profile determined using the profile model and the target feature data corresponding to the target business indicators of the users on other networks can accurately reflect the attribute characteristics of the users on other networks required by the target operator, thereby improving the accuracy of the profile of users on other networks.

[0086] In some embodiments, the portrait model includes a first neural network and a second neural network; the first neural network includes a plurality of first sub-neural networks, and each of the plurality of first sub-neural networks is of a different type.

[0087] S140: Input the target feature data corresponding to the target business indicators into the profiling model to determine the profile of users from other networks. This may include the following steps:

[0088] First, the target feature data is input into multiple first sub-neural networks to obtain the portrait probability corresponding to the portrait feature output by each first sub-neural network.

[0089] The target feature data is input into multiple first sub-neural networks, and each first sub-neural network determines the profile probability corresponding to the profile feature.

[0090] Profile features can represent the probability that a user from another network meets the required information, and profile probability can represent the probability value that a user from another network meets the required information.

[0091] In one example, the demand information is a user from another network whose profile is a high-value user. Then the profile feature is the probability that the user from another network is a high-value user, and the profile probability is the probability value that the user from another network is a high-value user. For example, the first sub-neural network outputs the probability that user A from another network is a high-value user as 0.9.

[0092] In one example, the first sub-neural network is a random forest model, an XgBoost model, or a support vector machine model, and the first neural network includes at least two of the random forest model, XgBoost model, or support vector machine model.

[0093] Then, the profile probability corresponding to the profile feature output by each first sub-neural network is input into the second neural network to obtain the profile of the user from the other network.

[0094] The probabilities corresponding to the profile features output by each of the first sub-neural networks are input into the second neural network, and the second neural network outputs the profile of the user from the other network.

[0095] In one example, the second neural network is a logistic regression model. The logistic regression model outputs the profile of the user from the other network based on the profile probabilities corresponding to the profile features output by each of the first sub-neural networks. For example, the output profile of user A from the other network is a high-value user.

[0096] The method provided in this application uses a first neural network and a second neural network to determine the image. Compared with the image determined by a single neural network, the image determined by a multi-layer neural network is more accurate, thereby improving the accuracy of the determined image.

[0097] In some embodiments, the target business metric includes multiple influencing factors, and the target feature data corresponding to the target business metric includes data corresponding to each of the multiple influencing factors; after determining the profile of users from other networks, the method may further include the following steps:

[0098] First, obtain the coefficients corresponding to each first sub-neural network.

[0099] During the training and adjustment of the first and second neural networks, the coefficients corresponding to each first sub-neural network after each training and adjustment are recorded. After obtaining the profile model, the coefficients corresponding to each first sub-neural network after the last training and adjustment are extracted from the recorded coefficients.

[0100] In one example, the second neural network is a logistic regression model, which can be represented by the following formula:

[0101]

[0102] Where y = 1 represents the profile of a user from another network that meets the required information. For example, if the required information is a user from another network whose profile is a high-value user, then y = 1 represents the profile of a high-value user from the profile of a user from another network. x represents the profile probability corresponding to the profile feature input to the second neural network, and θ represents the set of coefficients corresponding to multiple first sub-neural networks. When the first neural network includes three types of first sub-neural networks: random forest model, XgBoost model, and support vector machine model, θ = (θ1, θ2, θ3), where the random forest model corresponds to coefficient θ1, the XgBoost model corresponds to coefficient θ2, and the support vector machine model corresponds to coefficient θ3.

[0103] Secondly, based on the probabilities and coefficients of each first sub-neural network, the first neural network with the largest first influence weight on the image is determined among multiple first sub-neural networks, and the first sub-neural network with the largest first influence weight on the image is taken as the target neural network.

[0104] Calculate the first influence weight of each first sub-neuron in the image based on the image probability and coefficient of each first sub-neuron. Compare the first influence weights of each first sub-neuron in the image and select the first sub-neuron with the largest first influence weight as the target neural network.

[0105] The target neural network with the largest first influence weight on the image represents the first sub-neural network with the greatest influence on the image.

[0106] Next, the weights corresponding to each influencing factor are obtained from the target neural network.

[0107] The influencing factors represent the metrics within the target business metrics, and the weight of each metric affects the probability of the profile corresponding to the profile features output by the target neural network. To identify the factors that have a significant impact on the profile, the weights corresponding to each influencing factor are extracted from the target neural network.

[0108] Finally, based on the weight of each influencing factor and the data corresponding to each influencing factor, at least one influencing factor that meets the preset conditions is determined.

[0109] The preset conditions can include the top N in terms of influence, where N is a positive integer.

[0110] Based on the weight of each influencing factor and the corresponding data, the degree of influence of each influencing factor is determined. The degree of influence of each influencing factor is then sorted from largest to smallest to obtain the top N influencing factors.

[0111] The method provided in this application first selects the target neural network with the greatest impact on the portrait based on the coefficients and portrait probability of the first sub-neural network; then, based on the weight of each influencing factor in the target neural network and the data corresponding to each influencing factor, it selects the influencing factors with a greater impact on the portrait probability from multiple influencing factors, that is, the influencing factors with a greater impact on the portrait, providing a basis for optimizing the portrait model. After optimizing the portrait model based on the influencing factors with a greater impact on the portrait, the accuracy of the portrait model is further improved.

[0112] In some embodiments, determining the first sub-neuron with the largest first influence weight on the image among multiple first sub-neurons based on the image probability and coefficient of each first sub-neuron may include the following steps:

[0113] First, based on the probabilities and coefficients of each first sub-neural network, the first influence weight of each first sub-neural network on the image is determined.

[0114] For each first sub-neural network, calculate the product of the first sub-neural network's image probability and coefficient, and use the product as the first influence weight of the first sub-neural network on the image.

[0115] Then, based on the first influence weight of each first sub-neural network on the image, the first sub-neural network with the largest first influence weight among multiple first sub-neural networks is determined.

[0116] Compare the first influence weights of multiple first sub-neural networks on the image, and determine the first sub-neural network with the largest first influence weight among multiple first sub-neural networks.

[0117] The first influence weight represents the degree of influence of the first sub-neural network on the profile.

[0118] The method provided in this application embodiment obtains the first sub-neural network with the greatest influence on the portrait among multiple first sub-neural networks, providing a basis for optimizing the portrait model. After optimizing the portrait model based on the first sub-neural network with a greater influence on the portrait, the accuracy of the portrait model is further improved.

[0119] In some embodiments, determining at least one influencing factor that meets preset conditions based on the weight of each influencing factor and the data corresponding to each influencing factor may include the following steps:

[0120] First, based on the weight of each influencing factor and the data corresponding to each influencing factor, the second influence weight of each influencing factor on the probability of the profile output by the target neural network is determined.

[0121] For each influencing factor, calculate the absolute value of the product of the influencing factor's weight and the corresponding data. Use the absolute value of the product as the second influence weight of the influencing factor on the profile probability output by the target neural network.

[0122] The second influence weight represents the degree of influence of the influencing factors on the probability of the profile output by the target neural network.

[0123] Then, based on the second influence weight of each influencing factor on the probabilities of the target neural network output, at least one influencing factor that meets the preset conditions is determined.

[0124] The preset conditions can include the top N in terms of influence, where N is a positive integer.

[0125] The second influence weight of each influencing factor is sorted from largest to smallest to obtain the top N influencing factors.

[0126] The method provided in this application embodiment obtains the influencing factors that have a significant impact on the probabilities of the target neural network's portrait, that is, the influencing factors that have a significant impact on the portrait, providing a basis for optimizing the portrait model. After optimizing the portrait model based on the influencing factors that have a significant impact on the portrait, the accuracy of the portrait model is further improved.

[0127] In some embodiments, before S140: inputting the target feature data corresponding to the target business indicator into the profiling model to obtain the profile of the user from another network, the method may further include:

[0128] First, obtain sample profiles of each user from multiple users of the target operator, as well as feature data corresponding to the target business metrics of each user.

[0129] First, determine the sample profile of each user based on the business data of each user among multiple users of the target operator, and then extract the feature data corresponding to the target business indicators of each user.

[0130] Among them, business data refers to data that directly reflects the attribute characteristics of users corresponding to their needs.

[0131] In one example, when the target operator's demand information is for users on other networks whose profile is that of high-value users, the business data includes user package information that directly reflects the user's value.

[0132] In some embodiments, the business data of the same user is matched with the feature data corresponding to the target business indicators, and a sample profile of each user is determined based on the business data of each user among multiple users of the target operator. In this way, the feature data corresponding to the target business indicators of a user is matched with the sample profile, and the feature data corresponding to the target business indicators of a user and the sample profile are used as a sample.

[0133] In some embodiments, the multiple users include users who meet the demand information and users who do not meet the demand information. The sample profiles of users who meet the demand information and the feature data corresponding to the target business indicators are used as positive samples, and the sample profiles of users who do not meet the demand information and the feature data corresponding to the target business indicators are used as negative samples.

[0134] In one example, all positive samples and a portion of negative samples are extracted to train the profile model. To obtain a better model, the balance between positive and negative samples needs to be optimized.

[0135] The optimization objective is to make the proportion of positive samples P = m / (m+n), where m represents the number of positive samples and n represents the number of negative samples extracted.

[0136] If P < Q, since the data of positive samples has been fully extracted, only the upsampling technique can be used to process the positive samples to synthesize some positive samples. The number of synthesized samples is ceiling(Q * n / (1 - Q)) – m, where ceiling is rounding up, and Q is the preset threshold.

[0137] The strategy for synthesizing positive samples is to randomly select a sample b from the nearest neighbors of each minority-class sample a, and then randomly select a point on the line connecting a and b as the newly synthesized minority-class sample.

[0138] If P > Q, then negative samples are extracted from the unextracted negative samples, and the number of extracted negative samples is ceiling((m – Q * m) / Q).

[0139] Then, using the target business metrics, the sample portraits of each user among multiple users of the target operator, and the feature data corresponding to the target business metrics of each user, a portrait model is trained.

[0140] Input the target business metrics and the feature data corresponding to the target business metrics of each user into the portrait model to be trained to obtain a first portrait. Compare the first portrait of each user with the sample portrait, and calculate the accuracy rate of the portrait model to be trained. When the accuracy rate does not meet the preset accuracy rate, adjust the portrait model to be trained; when the accuracy rate meets the preset accuracy rate, obtain the portrait model.

[0141] The method provided by the embodiments of this application uses the sample portraits of the users of the target operator and the feature data corresponding to the target business metrics of the users to train and obtain a portrait model. Even if the feature data of off-net users obtained cannot directly reflect the attribute characteristics of off-net users, the portrait determined by using the portrait model and the target feature data corresponding to the target business metrics of off-net users can accurately reflect the attribute characteristics of off-net users required by the target operator, achieving the improvement of the accuracy of the portraits of off-net users.

[0142] The embodiments of this application also provide a device for determining portraits, as Figure 2 shown. The device 200 may include: an acquisition module 210, a determination module 220, and an extraction module 230.

[0143] The acquisition module 210 is used to acquire the demand information of the target operator and the feature data of off-net users.

[0144] The determination module 220 is used to determine the target business metrics corresponding to the demand information according to the corresponding relationship between the preset demand information and business metrics.

[0145] The extraction module 230 is used to extract the target feature data corresponding to the target business metrics from the feature data.

[0146] The determination module 220 is also used to input the target feature data corresponding to the target business indicators into the profile model to determine the profile of users on other networks.

[0147] The profiling model is trained using target business metrics, sample profiles of each user from multiple users of the target operator, and feature data corresponding to the target business metrics of each user.

[0148] The apparatus provided in this application, when a target operator needs a profile of users on other networks, acquires the target operator's demand information and the characteristic data of the users on other networks; determines the target business indicators corresponding to the demand information based on a preset correspondence between the demand information and business indicators; extracts target feature data corresponding to the target business indicators from the feature data; and inputs the target feature data corresponding to the target business indicators into a profile model to determine the profile of the users on other networks. Since the target business indicators are selected based on the demand information, and the profile model for determining the profile is trained using the target business indicators, sample profiles of users of the target operator, and the feature data corresponding to the user's target business indicators, even if the acquired feature data of the users on other networks cannot directly reflect the attribute characteristics of the users on other networks, the profile determined using the profile model and the target feature data corresponding to the target business indicators of the users on other networks can accurately reflect the attribute characteristics of the users on other networks required by the target operator, thereby improving the accuracy of the profile of users on other networks.

[0149] In some embodiments, the profiling model includes a first neural network and a second neural network; the first neural network includes a plurality of first sub-neural networks, and each of the plurality of first sub-neural networks is of a different type.

[0150] Module 220 can be specifically used for:

[0151] The target feature data is input into multiple first sub-neural networks respectively to obtain the portrait probability corresponding to the portrait feature output by each first sub-neural network;

[0152] The profile probability corresponding to the profile feature output by each first sub-neural network is input into the second neural network to obtain the profile of the user from the other network.

[0153] The apparatus provided in this application uses a first neural network and a second neural network to determine the image. Compared with the image determined by a single neural network, the image determined by a multi-layer neural network is more accurate, thereby improving the accuracy of the determined image.

[0154] In some embodiments, the target business metric includes multiple influencing factors, and the target feature data corresponding to the target business metric includes data corresponding to each of the multiple influencing factors.

[0155] The acquisition module 210 is also used to acquire the coefficients corresponding to each first sub-neural network;

[0156] The determination module 220 is also used to determine the first neural network with the largest first influence weight on the first image among multiple first sub-neural networks based on the image probability and coefficient of each first sub-neural network, and to take the first sub-neural network with the largest first influence weight as the target neural network.

[0157] The acquisition module 210 is also used to acquire the weights corresponding to each influencing factor from the target neural network;

[0158] The determination module 220 is also used to determine at least one influencing factor that meets the preset conditions based on the weight of each influencing factor and the data corresponding to each influencing factor.

[0159] The apparatus provided in this application first selects the target neural network that has the greatest impact on the portrait based on the coefficients and portrait probability of the first sub-neural network. Then, based on the weight of each influencing factor in the target neural network and the data corresponding to each influencing factor, it selects the influencing factor that has a greater impact on the portrait probability from multiple influencing factors, that is, the influencing factor that has a greater impact on the portrait, so as to provide a basis for optimizing the portrait model. After optimizing the portrait model based on the influencing factor that has a greater impact on the portrait, the accuracy of the portrait model is further improved.

[0160] In some embodiments, the determining module 220 may also be specifically used for:

[0161] Based on the probabilities and coefficients of each first sub-neural network, determine the first influence weight of each first sub-neural network on the image.

[0162] Based on the first influence weight of each first sub-neural network on the image, determine the first sub-neural network with the largest first influence weight among multiple first sub-neural networks.

[0163] The device provided in this application embodiment obtains the first sub-neural network with the greatest influence on the portrait among multiple first sub-neural networks, providing a basis for optimizing the portrait model. After optimizing the portrait model based on the first sub-neural network with a greater influence on the portrait, the accuracy of the portrait model is further improved.

[0164] In some embodiments, the determining module 220 may also be specifically used for:

[0165] Based on the weight of each influencing factor and the data corresponding to each influencing factor, determine the second influence weight of each influencing factor on the probability of the profile output by the target neural network.

[0166] Based on the second influence weight of each influencing factor on the profile probability output by the target neural network, and the data corresponding to each influencing factor, at least one influencing factor that meets the preset conditions is determined.

[0167] The apparatus provided in this application embodiment obtains the influencing factors that have a significant impact on the probabilities of the target neural network's portrait, that is, the influencing factors that have a significant impact on the portrait, providing a basis for optimizing the portrait model. After optimizing the portrait model based on the influencing factors that have a significant impact on the portrait, the accuracy of the portrait model is further improved.

[0168] In some embodiments, the acquisition module 210 is further configured to acquire a sample profile of each user among multiple users of the target operator, and feature data corresponding to the target business indicators of each user.

[0169] The device 200 may also include a training module 240.

[0170] Training module 240 is used to train a profile model using target business metrics, sample profiles of each user among multiple users of the target operator, and feature data corresponding to the target business metrics of each user.

[0171] The device provided in this application uses sample profiles of users from the target operator and feature data corresponding to the user's target business indicators to train a profile model. Even if the feature data of the acquired users from other networks cannot directly reflect the attribute characteristics of the users from other networks, the profile determined by using the profile model and the target feature data corresponding to the target business indicators of the users from other networks can accurately reflect the attribute characteristics of the users from other networks required by the target operator, thereby improving the accuracy of the profiles of users from other networks.

[0172] The apparatus for determining an image provided in this application embodiment performs... Figure 1 The steps in the method shown, and the technical effect of improving the accuracy of user profiles on other networks, will not be elaborated further here for the sake of brevity.

[0173] Figure 3 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application is shown.

[0174] An electronic device may include a processor 301 and a memory 302 storing computer program instructions.

[0175] Specifically, the processor 301 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0176] Memory 302 may include mass storage for data or instructions. For example, and not limitingly, memory 302 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 302 may include removable or non-removable (or fixed) media. Where appropriate, memory 302 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 302 is non-volatile solid-state memory.

[0177] Memory may include read-only memory (ROM), random access memory (RAM), disk storage media devices, optical storage media devices, flash memory devices, and electrical, optical, or other physical / tangible memory storage devices. Therefore, typically, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the methods according to one aspect of this disclosure.

[0178] The processor 301 reads and executes computer program instructions stored in the memory 302 to implement any of the methods for determining an image in the above embodiments.

[0179] In one example, the electronic device may also include a communication interface 303 and a bus 310. For example, Figure 3 As shown, the processor 301, memory 302, and communication interface 303 are connected through bus 310 and complete communication with each other.

[0180] The communication interface 303 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0181] Bus 310 includes hardware, software, or both, that couples components of an online data traffic metering device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 310 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, any suitable bus or interconnect is contemplated herein.

[0182] The electronic device can execute the method for determining the image in the embodiments of this application, thereby achieving the combination Figure 1 The method described is for determining the portrait.

[0183] Furthermore, in conjunction with the methods for determining an image in the above embodiments, this application embodiment can provide a computer-readable storage medium for implementation. This computer-readable storage medium stores computer program instructions; when executed by a processor, these computer program instructions implement any of the methods for determining an image in the above embodiments.

[0184] Based on the methods for determining an image in the above embodiments, this application can provide a computer program product for implementation. When the instructions in the computer program product are executed by the processor of an electronic device, they implement any of the methods for determining an image in the above embodiments.

[0185] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0186] The functional blocks shown in the above-described structural diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0187] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0188] The above description is merely a specific implementation of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A method for determining a portrait, characterized in that, The method includes: Obtain the target operator's demand information and the characteristic data of users on other networks; Based on the pre-defined correspondence between demand information and business indicators, the target business indicators corresponding to the demand information are determined. Extract target feature data corresponding to the target business indicators from the feature data; Input the target feature data corresponding to the target business indicators into the profiling model to determine the profile of the user from the other network. The profiling model is trained using the target business metrics, sample profilings of each user among multiple users of the target operator, and feature data corresponding to the target business metrics of each user. The profiling model includes a first neural network and a second neural network; the first neural network includes multiple first sub-neural networks, each of which is of a different type; the step of inputting the target feature data corresponding to the target business indicator into the profiling model to determine the profile of the user from another network includes: The target feature data is input into the plurality of first sub-neural networks respectively to obtain the portrait probability corresponding to the portrait feature output by each first sub-neural network; The profile probability corresponding to the profile feature output by each of the first sub-neural networks is input into the second neural network to obtain the profile of the user from the other network.

2. The method according to claim 1, characterized in that, The target business indicator includes multiple influencing factors, and the target feature data corresponding to the target business indicator includes data corresponding to each of the multiple influencing factors; the method further includes: Obtain the coefficients corresponding to each of the first sub-neural networks; Based on the probabilities and coefficients of each first sub-neural network, the first neural network with the largest first influence weight on the image is determined among the plurality of first sub-neural networks, and the first sub-neural network with the largest first influence weight on the image is taken as the target neural network. Obtain the weight corresponding to each of the influencing factors from the target neural network; Based on the weight of each influencing factor and the data corresponding to each influencing factor, at least one influencing factor that meets the preset conditions is determined.

3. The method according to claim 2, characterized in that, The step of determining the first sub-neuron with the largest first influence weight on the image among the plurality of first sub-neurons based on the image probability and coefficient of each first sub-neuron includes: Based on the image probability and coefficient of each first sub-neural network, determine the first influence weight of each first sub-neural network on the image; Based on the first influence weight of each first sub-neural network on the image, determine the first sub-neural network with the largest first influence weight among the plurality of first sub-neural networks.

4. The method according to claim 2, characterized in that, The step of determining at least one influencing factor that satisfies preset conditions based on the weight of each influencing factor and the data corresponding to each influencing factor includes: Based on the weight of each influencing factor and the data corresponding to each influencing factor, a second influence weight is determined for each influencing factor on the probability of the profile output by the target neural network. Based on the second influence weight of each of the influencing factors on the profile probability output by the target neural network, at least one influencing factor that satisfies the preset condition is determined.

5. The method according to claim 1, characterized in that, Before inputting the target feature data corresponding to the target business indicator into the profiling model to obtain the profile of the user from the other network, the method further includes: Obtain a sample profile of each user among multiple users of the target operator, and feature data corresponding to the target business indicators of each user; The profiling model is trained using the target business metrics, sample profiles of each user among multiple users of the target operator, and feature data corresponding to the target business metrics of each user.

6. An apparatus for determining an image, characterized in that, The device includes: The acquisition module is used to acquire the target operator's demand information and the characteristic data of users from other networks. The determination module is used to determine the target business indicators corresponding to the pre-set demand information and business indicators based on the correspondence between the demand information and the business indicators. The extraction module is used to extract target feature data corresponding to the target business indicator from the feature data; The determining module is further configured to input the target feature data corresponding to the target business indicator into the profiling model to determine the profile of the user from the other network. The profiling model is trained using the target business metrics, sample profilings of each user among multiple users of the target operator, and feature data corresponding to the target business metrics of each user. The profiling model includes a first neural network and a second neural network; the first neural network includes multiple first sub-neural networks, each of which has a different type; the determining module is specifically used to: input the target feature data into the multiple first sub-neural networks respectively to obtain the profiling probability corresponding to the profiling feature output by each first sub-neural network; input the profiling probability corresponding to the profiling feature output by each first sub-neural network into the second neural network to obtain the profiling of the user from the other network.

7. An electronic device, characterized in that, The device includes: a processor and a memory storing computer program instructions; the processor, when executing the computer program instructions, implements the method for determining an image as described in any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions, which, when executed by a processor, implement the method for determining an image as described in any one of claims 1-5.

9. A computer program product, characterized in that, When the instructions in the computer program product are executed by the processor of the electronic device, the electronic device performs the method for determining a portrait as described in any one of claims 1-5.

Images