Image label determination method and device, computer device, and storage medium

Abstract

The application relates to a portrait label determination method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: identifying a target object based on target business data and target demand information; determining at least one initial associated feature of the target object corresponding to the target demand information; screening the at least one initial associated feature based on a preset complete condition to obtain at least one first associated feature; calculating the correlation degree of each first associated feature based on a preset correlation degree algorithm, and screening each first associated feature to obtain a target associated feature satisfying a preset correlation condition, and taking the target associated feature as the portrait label of the target object. By using the method, the data categories with weak relevance can be automatically screened out, the data collection efficiency can be improved, the difficulty of obtaining the portrait label of the target object is further reduced, and the success rate and accuracy of user portrait modeling are improved.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to a method, apparatus, computer equipment, storage medium, and computer program product for determining image tags. Background Technology

[0002] With the rapid development of the communications industry, mobile communication technology has been widely applied. Mobile communication networks contain user service data, which can be used to create user profiles and models. Based on the obtained user profile tags, the network can specifically meet the service needs of particular users, thereby improving user experience and enhancing the overall efficiency of network resource utilization.

[0003] In related technologies, the design of user profile tags is generally carried out through human intervention, supplemented by big data statistical methods to perform mathematical statistics on user attributes and user behavior to obtain potential tag options. Then, human experience is used to design the tag adoption and classification, resulting in low accuracy of the obtained user profile tags. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, apparatus, computer device, computer-readable storage medium, and computer program product for determining portrait labels that can improve classification accuracy, in order to address the aforementioned technical problems.

[0005] Firstly, this application provides a method for determining portrait tags. The method includes:

[0006] Identify target objects based on target business data and target demand information;

[0007] Determine at least one initial associated feature of the target object corresponding to the target requirement information;

[0008] Based on preset completeness conditions, at least one initial associated feature is filtered to obtain at least one first associated feature;

[0009] Based on a preset correlation algorithm, the correlation degree of each first correlation feature is calculated. Based on the correlation degree, each first correlation feature is filtered to obtain target correlation features that meet preset correlation conditions. The target correlation features are used as the portrait tags of the target object.

[0010] In one embodiment, the step of filtering the at least one initial association feature based on a preset completeness condition to obtain at least one first association feature includes:

[0011] Determine the complete temporal results for each of the initial association features;

[0012] The initial association features that meet the preset completeness conditions for time-complete results are identified as the first association features.

[0013] In one embodiment, the step of filtering each of the first associated features based on the degree of association to obtain target associated features that meet preset association conditions includes:

[0014] The first associated feature whose correlation degree is greater than or equal to the preset correlation degree threshold is taken as the target associated feature that satisfies the preset correlation condition.

[0015] In one embodiment, the step of filtering each of the first associated features based on the degree of association to obtain target associated features that meet preset association conditions includes:

[0016] Generate a sequence of associated features according to the descending order of the degree of association of each of the first associated features;

[0017] The first number of first associated features in the associated feature sequence that meet the preset association conditions are taken as target associated features.

[0018] In one embodiment, the method further includes:

[0019] Tests are performed based on the business data and the target association features of the target object to obtain business test results. Based on the business test results, the profile tags of the target object are adjusted to obtain the adjusted profile tags of the target object.

[0020] In one embodiment, adjusting the profile label of the target object based on the business test results to obtain the adjusted profile label of the target object includes:

[0021] If the business test results do not meet the preset test conditions, return to the step of identifying the target object based on the target business data and target requirement information.

[0022] In one embodiment, adjusting the profile label of the target object based on the business test results to obtain the adjusted profile label of the target object includes:

[0023] Based on the difference between the business test results and the standard results, the profile tags of the target object are adjusted to obtain the adjusted profile tags of the target object.

[0024] In one embodiment, the method further includes:

[0025] If an update is detected in the business data corresponding to the target object, the steps of determining the target object and target requirements based on the updated business data are performed to obtain the updated target association features.

[0026] In one embodiment, the initial association features include behavioral features and attribute features, wherein the behavioral features include behavioral type features and behavioral duration features.

[0027] Secondly, this application also provides a portrait tag determining device. The device includes:

[0028] The first determination module is used to identify target objects based on target business data and target requirement information;

[0029] The second determining module is used to determine at least one initial association feature of the target object corresponding to the target requirement information;

[0030] The filtering module is used to filter the at least one initial associated feature based on a preset completeness condition to obtain at least one first associated feature;

[0031] The first calculation module is used to calculate the correlation degree of each first correlation feature based on a preset correlation degree algorithm, filter each first correlation feature based on the correlation degree to obtain target correlation features that meet preset correlation conditions, and use the target correlation features as the portrait tags of the target object.

[0032] In one embodiment, the filtering module is specifically used for:

[0033] Determine the complete temporal results for each of the initial association features;

[0034] The initial association features that meet the preset completeness conditions for time-complete results are identified as the first association features.

[0035] In one embodiment, the first computing module is specifically used for:

[0036] The first associated feature whose correlation degree is greater than or equal to the preset correlation degree threshold is taken as the target associated feature that satisfies the preset correlation condition.

[0037] In one embodiment, the first computing module is specifically used for:

[0038] Generate a sequence of associated features according to the descending order of the degree of association of each of the first associated features;

[0039] The first number of first associated features in the associated feature sequence that meet the preset association conditions are taken as target associated features.

[0040] In one embodiment, the device further includes:

[0041] The testing module is used to perform tests based on the business data and the target association features of the target object, obtain business test results, and adjust the profile tags of the target object based on the business test results to obtain the adjusted profile tags of the target object.

[0042] In one embodiment, the test module is specifically used for:

[0043] If the business test results do not meet the preset test conditions, return to the step of identifying the target object based on the target business data and target requirement information.

[0044] In one embodiment, the test module is specifically used for:

[0045] Based on the difference between the business test results and the standard results, the profile tags of the target object are adjusted to obtain the adjusted profile tags of the target object.

[0046] In one embodiment, the device further includes:

[0047] The update module is used to, when an update is detected in the business data corresponding to the target object, perform the step of determining the target object and target requirements based on the updated business data, and obtain the updated target association features.

[0048] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0049] Identify target objects based on target business data and target demand information;

[0050] Determine at least one initial associated feature of the target object corresponding to the target requirement information;

[0051] Based on preset completeness conditions, at least one initial associated feature is filtered to obtain at least one first associated feature;

[0052] Based on a preset correlation algorithm, the correlation degree of each first correlation feature is calculated. Based on the correlation degree, each first correlation feature is filtered to obtain target correlation features that meet preset correlation conditions. The target correlation features are used as the portrait tags of the target object.

[0053] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:

[0054] Identify target objects based on target business data and target demand information;

[0055] Determine at least one initial associated feature of the target object corresponding to the target requirement information;

[0056] Based on preset completeness conditions, at least one initial associated feature is filtered to obtain at least one first associated feature;

[0057] Based on a preset correlation algorithm, the correlation degree of each first correlation feature is calculated. Based on the correlation degree, each first correlation feature is filtered to obtain target correlation features that meet preset correlation conditions. The target correlation features are used as the portrait tags of the target object.

[0058] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:

[0059] Identify target objects based on target business data and target demand information;

[0060] Determine at least one initial associated feature of the target object corresponding to the target requirement information;

[0061] Based on preset completeness conditions, at least one initial associated feature is filtered to obtain at least one first associated feature;

[0062] Based on a preset correlation algorithm, the correlation degree of each first correlation feature is calculated. Based on the correlation degree, each first correlation feature is filtered to obtain target correlation features that meet preset correlation conditions. The target correlation features are used as the portrait tags of the target object.

[0063] The aforementioned method, apparatus, computer equipment, storage medium, and computer program product for determining user profile tags include: identifying a target object based on target business data and target demand information; determining at least one initial associated feature of the target object corresponding to the target demand information; filtering the at least one initial associated feature based on preset completeness conditions to obtain at least one first associated feature; calculating the correlation degree of each first associated feature based on a preset correlation degree algorithm; filtering each first associated feature based on the correlation degree to obtain target associated features that satisfy preset correlation conditions; and using the target associated features as the user profile tags for the target object. By employing this method, data categories with weak relevance can be automatically filtered out, improving data collection efficiency, achieving data dimensionality reduction, and further reducing the difficulty of obtaining user profile tags for the target object, thereby improving the success rate and accuracy of user profile modeling. Attached Figure Description

[0064] Figure 1 This is a flowchart illustrating a method for determining image tags in one embodiment;

[0065] Figure 2 This is a flowchart illustrating the step of determining the first associated feature in one embodiment;

[0066] Figure 3 This is a flowchart illustrating the steps for determining target association features in one embodiment;

[0067] Figure 4a This is a flowchart illustrating the steps of filtering associated features using a random forest in one embodiment.

[0068] Figure 4b This is a flowchart illustrating the steps of filtering associated features using a random forest in one embodiment.

[0069] Figure 5 This is a structural block diagram of an image label determining device in one embodiment;

[0070] Figure 6 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0071] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0072] In one embodiment, such as Figure 1As shown, a method for determining profile tags is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, and to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. The aforementioned terminals can be, but are not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc. Portable wearable devices can include smartwatches, smart bracelets, head-mounted devices, etc. The server can be a standalone server or a server cluster composed of multiple servers. In this embodiment, the method for determining profile tags includes the following steps:

[0073] Step 102: Identify the target object based on the target business data and target requirement information.

[0074] The target business data can be acquired, and the target objects corresponding to the target business data can be users corresponding to different businesses. For example, it can be mobile communication service subscribers and mobile communication terminals (4G terminals, 5G terminals, CPEs, etc.) in user profile modeling for mobile communication service evaluation. It can also be further divided into users of specific service brands, such as 5G Enjoy Package users, BBK Package users, etc. The target demand information can be the demand for user profile tags for different services. For example, it can include user profile demand for the service market, or it can also include demand for...

[0075] In implementation, the terminal can determine the target object corresponding to both the target business data and the target requirement information based on the acquired target business data and the acquired target requirement information. In other words, the terminal can perform subject identification on the target business data based on the target requirement information to obtain the target object that is oriented towards the target requirement information and corresponds to the target business data.

[0076] In one example, optionally, the target object can be a person or object capable of performing actions. One possible implementation is as follows: the target requirement information can be an evaluation requirement for mobile communication services, and the target object can be a mobile communication service subscriber or a mobile communication terminal. The mobile communication terminal can include 4G (the 4th generation mobile communication technology) terminals, 5G (the 5th generation mobile communication technology) terminals, and CPE (Customer Premise Equipment), etc.; the target object can also be users classified according to different services in mobile communication, such as 5G package users, 4G package users, etc.; the target object can also include the user's mobile trajectory, such as network-side behavioral indicators such as indoor time and outdoor time.

[0077] Step 104: Determine at least one initial associated feature of the target object corresponding to the target requirement information.

[0078] The initial association features include the target object's behavioral and attribute features. Behavioral features include behavioral type and behavioral duration features. Attribute features can be relatively stable characteristics of the target object, such as the target object's identification information, terminal model information (e.g., user's mobile phone number, user's terminal model, frequently used applications, and active services). In one example, where the target requirement is for network performance maintenance and optimization, attribute features could include the user's frequent network presence, 4G dwell time, 5G dwell time, etc.

[0079] Behavioral characteristics can represent the mobile communication service-related behaviors that a target object will engage in based on target business data. These can include resource data interaction behaviors and interactive behaviors. Resource data interaction behaviors can include resource data interaction characteristics, resource data interaction content, and resource data interaction methods. Resource data interaction characteristics can include the interaction values, number of interactions, and interaction time frequency corresponding to the resource data. Resource data interaction content can include interaction content corresponding to different types of services, such as subscription-based reading, video-based interaction, membership-based interaction, and game-based interaction. Resource data interaction methods include the interaction methods provided by payment institutions for different resource data.

[0080] One possible implementation could include: the target object and its corresponding behavioral characteristics could be established by creating a relational database of target objects and behavioral characteristics that includes attribute characteristics. For example, the target object could be a 5G package user. This target object generates resource data interaction behaviors corresponding to video-type interactive services through 4G, and generates resource data interaction behaviors for reading-type subscriptions through online payment institutions through 5G, etc.

[0081] In practice, the terminal can process the target business data to obtain multiple initial associated features of the target object that meet the target requirements.

[0082] Step 106: Filter at least one initial associated feature based on preset complete conditions to obtain at least one first associated feature.

[0083] Among them, the preset reliability conditions represent the reliability of the target business data. The preset reliability conditions may include completeness conditions, which are used to screen the completeness of each initial associated feature in the time dimension. The preset reliability conditions may also include reasonableness conditions, which are used to judge the reasonableness of the data.

[0084] In implementation, the terminal can filter multiple initial association features corresponding to the target business data based on pre-configured reliability conditions, remove initial association features that do not meet the preset reliability conditions, and use the retained initial association features as the first association features of the target object.

[0085] Step 108: Based on the preset correlation degree algorithm, calculate the correlation degree of each first correlation feature, filter each first correlation feature based on the correlation degree, obtain the target correlation feature that meets the preset correlation condition, and use the target correlation feature as the portrait label of the target object.

[0086] The preset correlation degree algorithm can be a random forest algorithm, and the correlation degree can characterize the importance of the first correlation feature. The preset correlation condition is a condition used to filter each first correlation feature based on the correlation degree of each first correlation feature. The content of the preset correlation condition can be that the correlation degree of the correlation feature is greater than the preset correlation degree threshold, or that the correlation feature is one of the top-number correlation features in the correlation feature sequence, etc. This disclosure does not specifically limit the specific value of the preset correlation degree threshold and the specific value of the number of targets. Those skilled in the art can determine them specifically based on the actual application scenario.

[0087] In implementation, for multiple primary association features of a target object corresponding to the target business data, the terminal can calculate the association degree of each primary association feature using a preset association degree algorithm. Based on the calculated association degrees of each primary association feature, the terminal sorts and filters the multiple primary association features, ensuring that the primary association features with higher importance are selected. That is, the primary association features that meet the preset association conditions are used as the target association features of the target object. Based on this, the terminal can use the selected target association features as profile tags for the target object.

[0088] The aforementioned method for determining user profile tags includes: identifying target objects based on target business data and target demand information; determining at least one initial related feature of the target object corresponding to the target demand information; filtering the at least one initial related feature based on preset completeness conditions to obtain at least one first related feature; calculating the correlation degree of each first related feature based on a preset correlation degree algorithm; filtering each first related feature based on the correlation degree to obtain target related features that meet preset correlation conditions; and using the target related features as the user profile tags for the target object. By adopting this method, data categories with weak correlation can be automatically filtered out, improving the efficiency of data collection, achieving data dimensionality reduction, and further reducing the difficulty of obtaining user profile tags for target objects, thereby improving the success rate and accuracy of user profile modeling.

[0089] In one embodiment, such as Figure 2 As shown, the specific processing steps of step 106, "screening at least one initial associated feature based on preset complete conditions to obtain at least one first associated feature," include:

[0090] Step 202: Determine the complete temporal results for each initial association feature.

[0091] Among them, the time integrity result represents the completeness of the business data corresponding to the initial association feature in the time dimension, which may include complete state and incomplete state.

[0092] In practice, for each initial association feature, the terminal can determine the complete state of the target business data corresponding to that initial association feature in the time dimension and obtain the complete time result of that initial association feature.

[0093] Step 204: The initial association features that meet the preset completeness conditions for time-based complete results are identified as the first association features.

[0094] The preset completeness conditions may include: determining the initial association features that are in a complete state in the time dimension as initial association features that satisfy the preset completeness conditions. The preset completeness conditions may be pre-configured.

[0095] In implementation, for each initial association feature, after obtaining the complete temporal result of that initial association feature, the terminal can determine whether the initial association feature corresponding to the complete temporal result meets the preset completeness condition based on the completeness status in the complete temporal result. Based on this, the terminal can determine the initial association feature corresponding to the complete temporal result with a complete status as the initial association feature that meets the preset completeness condition, and the terminal can determine the initial association feature that meets the preset completeness condition as the first association feature.

[0096] Optionally, the terminal can determine the completeness of the target service data corresponding to each initial association feature; the target service data can be service data obtained through market statistics, as well as service data obtained through wireless network base stations and core networks. The terminal can determine the time series corresponding to the above service data. If the time series of the service data corresponding to the initial association feature is complete, then the initial association feature is determined to meet the preset completeness condition, and the initial association feature is determined as the first association feature.

[0097] In this embodiment, by judging the time integrity of the target business data corresponding to the initial association feature, the integrity and reliability of the target business data can be guaranteed, as well as the reliability of the association feature determined based on the target business data.

[0098] In one embodiment, the specific processing steps of step 108, "screening each first correlation feature based on correlation degree to obtain target correlation features that meet preset correlation conditions," include:

[0099] The first associated feature whose correlation degree is greater than or equal to the preset correlation degree threshold is taken as the target associated feature that satisfies the preset correlation condition.

[0100] The correlation degree of the first correlation feature can be calculated by the terminal using a random forest algorithm. The correlation degree represents the importance of the first correlation feature among multiple correlation features corresponding to the target business data. The preset correlation degree threshold can be a threshold used to filter multiple first correlation features. The specific value of the preset correlation degree threshold can be determined by those skilled in the art based on the actual application scenario. This disclosure does not specifically limit the specific value of the preset correlation degree threshold.

[0101] In implementation, the terminal can filter each first associated feature based on the degree of association of each first associated feature. For example, it can compare the degree of association of each first associated feature with a preset degree of association threshold, remove the first associated features with a degree of association less than the preset degree of association threshold, and retain the first associated features with a degree of association greater than or equal to the preset degree of association threshold. That is, the first associated features with a degree of association greater than or equal to the preset degree of association threshold are determined as associated features that meet the preset association conditions, which are the target associated features.

[0102] In this embodiment, a preset correlation threshold is used to determine whether each first correlation feature meets the preset correlation conditions. Based on the importance of each correlation feature, data categories with weak correlation are eliminated, thereby reducing the dimensionality of business data and improving the success rate and accuracy of extracting user profile tags.

[0103] In one embodiment, such as Figure 3 As shown, the specific processing steps of step 108, "screening each first correlation feature based on correlation degree to obtain target correlation features that meet preset correlation conditions," include:

[0104] Step 302: Generate a sequence of associated features according to the order of their degree of association from largest to smallest.

[0105] The correlation degree of the first correlation feature can be calculated by the terminal using a random forest algorithm. The correlation degree represents the importance of the first correlation feature among multiple correlation features corresponding to the target business data. The correlation feature sequence may include multiple first correlation features arranged in descending order.

[0106] In implementation, for each of the first associated features determined in the above embodiments, the terminal can arrange them according to the degree of association of each first associated feature, and generate an associated feature sequence in descending order of the degree of association of each first associated feature.

[0107] Step 304: Select the first number of first associated features in the associated feature sequence as target associated features that satisfy the preset associated conditions.

[0108] The target number is used to filter multiple first associated features contained in the associated feature sequence. The target number can be pre-configured, and the specific value of the target number can be determined by those skilled in the art based on the actual application scenario. This disclosure does not specifically limit the specific value of the target number.

[0109] In practice, the terminal can filter multiple first associated features contained in the associated feature sequence, and select the first target number of first associated features contained in the associated feature sequence as associated features that meet the preset associated conditions, that is, determine them as target associated features.

[0110] In one example, the number of targets can be 10. In this way, after the terminal obtains a sequence of associated features containing multiple first associated features, it can determine the first associated features that are in the top ten in order as the target associated features that meet the preset associated conditions.

[0111] In this embodiment, a sequence of associated features can be generated in descending order of the degree of association of each first associated feature. The first associated features can be filtered according to the sequence of associated features to remove data categories with weak correlation, thereby reducing the dimensionality of business data and improving the success rate and accuracy of extracting user profile tags.

[0112] In one embodiment, the image label determination method further includes:

[0113] Tests are conducted based on business data and the target object's target association characteristics to obtain business test results. Based on these results, the target object's profile tags are adjusted to obtain the adjusted target object's profile tags.

[0114] Among them, business data can be business data within a historical time period obtained by the terminal, and business test results can be test results obtained by testing the target association features of the target object. The business test results represent the classification accuracy of the target association features.

[0115] In implementation, the terminal can verify the validity of the target association features obtained from the target objects, that is, determine the classification accuracy of the obtained target association features. Specifically, this can be done by testing the target objects corresponding to the target association features based on business data within a historical time period, and obtaining business test results based on the behavioral characteristics of the target objects. Based on this, if the business test results do not meet the preset test conditions, the terminal can adjust the profile tags obtained in the above embodiments to obtain multiple profile tags corresponding to the adjusted target objects.

[0116] In this embodiment, by testing and adjusting the image tags, image tags that do not meet the preset test conditions can be adjusted in a timely manner, thereby further improving the accuracy of the image tags.

[0117] In one embodiment, the specific processing steps for the step "adjusting the target object's profile tags based on business test results to obtain the adjusted target object's profile tags" include:

[0118] If the business test results do not meet the preset test conditions, return to the step of identifying the target object based on the target business data and target requirement information.

[0119] Among them, the preset test conditions are the conditions for judging the business test results. The content of the preset test conditions can be that the difference between the business test results and the standard results is less than or equal to a preset threshold, etc.

[0120] In implementation, the terminal can determine whether the business test results meet preset test conditions. This involves calculating the difference between the business test result and the standard result, and comparing this difference with a preset threshold. If the calculated difference is less than or equal to the preset threshold, the business test result is considered to meet the preset test conditions. If the calculated difference is greater than the preset threshold, the business test result is considered to not meet the preset test conditions. Based on this, the terminal can return to the step of identifying the target object based on the target business data and target requirement information, and redetermine the target association features corresponding to the target business data.

[0121] In one example, the terminal can verify the accuracy of profile tags using historical data, such as the classification accuracy of the profile tags, or verify the validity of the profile tags using dynamic data (historical data). The main methods may include configuring the network for the target object corresponding to the target profile tag, monitoring changes in the behavioral and attribute characteristics of the target object after network configuration, and determining whether the profile tag meets the expected results, i.e., whether it meets the preset test conditions based on the obtained changes in behavioral and attribute characteristics.

[0122] In this embodiment, by adjusting the image tags that do not meet the preset test conditions, the accuracy of the image tags can be improved in a timely manner.

[0123] In one embodiment, the specific processing steps for the step "adjusting the target object's profile tags based on business test results to obtain the adjusted target object's profile tags" include:

[0124] Based on the difference between the business test results and the standard results, the profile tags of the target object are adjusted to obtain the adjusted profile tags of the target object.

[0125] In implementation, the terminal can calculate the difference between the business test results and the pre-configured standard results, and adjust the profile labels of the target objects based on the calculated difference, thus obtaining the adjusted profile labels.

[0126] In this embodiment, adjusting the profile tags of the target object based on the degree of difference between the business test results and the standard results can ensure the effectiveness of the adjustment and further improve the accuracy of the profile tags.

[0127] In one embodiment, the image label determination method further includes:

[0128] If an update to the business data corresponding to the target object is detected, the steps of identifying the target object based on the target business data and target requirement information are performed to obtain the updated target association features.

[0129] In practice, the terminal can monitor the target business data corresponding to the target object. If the terminal detects that the business data corresponding to the target object has been updated, it can re-execute the steps described in the above embodiment to identify the target object based on the target business data and target requirement information, and determine the updated target association features.

[0130] In this embodiment, the business data corresponding to the target object can be continuously monitored. When the business data is updated, the profile tags can be updated in a timely manner to achieve adaptive adjustment of profile tags based on changes in statistical data and ensure the accuracy of profile tags.

[0131] The following describes in detail the specific execution process of the above-mentioned method for determining image tags, with reference to a specific embodiment:

[0132] The method for determining user profile tags provided in this embodiment relates to the field of big data. With the rapid development of the communications industry, mobile communication technology is widely used, and multi-layered and multi-dimensional user service data can be collected during mobile communication network operation. After sensitive information processing, this data can be used on the network side to perform profile modeling based on user service behavior. Based on user profile tags, the network side can clearly define the type, time, and frequency of user-initiated services, as well as user behavioral preferences. The network side can predictively and specifically meet the service needs of specific users through reasonable cell load balancing and resource scheduling algorithm optimization, thereby improving user experience and overall network resource utilization efficiency.

[0133] In related technologies, the main reliance is on big data technology for massive data collection and statistics. The classification of profile features is mainly based on statistical features, and the determination of key profile tags is based on human experience, resulting in low accuracy in profile tag determination.

[0134] This embodiment provides a method for determining user profile labels, which can effectively eliminate inefficient data information in the data collected from the network, reduce the dimensionality of the input data for user profile modeling, further reduce the complexity of user profile model training, and improve the success rate and accuracy of user profile modeling.

[0135] S1, Determine user profile requirements, i.e., determine target requirement information. Specifically, the terminal can determine the target requirement information for the user profile. For mobile communication networks, user profile requirements for the business market and user profile requirements for operation and maintenance are different. For example, the user profile tag requirements for the business market focus on resource data interaction and business type profile tags, while the user profile tag requirements for operation and maintenance focus on the overall quality of network operation.

[0136] S2, determine user profile tags. Specifically, the terminal divides the data and time relationship, analyzes the connection between the data and the corresponding subject, verifies the rationality of the logical relationship model, and thus determines the rationality of each tag of the user profile.

[0137] S3, user profile tag development, specifically, allows for the labeling of user profiles according to the design scheme.

[0138] S4. Verify user profiles. Specifically, the terminal can verify the accuracy of profile tags by utilizing historical data, and can verify the classification accuracy corresponding to the profile tags. This involves influencing network configuration and service guidance for users with a specific tag, and collecting data on subsequent behavior and attribute changes of these users to verify whether the tag achieves the expected results.

[0139] S5 applies user profiling. Specifically, based on the classification results of user profile tags, and according to the user profile's needs and objectives, it takes corresponding business or network-side actions, such as load balancing and resource scheduling optimization measures.

[0140] Specifically, the process of determining profile labels may include data segmentation and data association modeling.

[0141] like Figure 4aThe diagram illustrates a static data partitioning flowchart. The terminal can identify the target entity (target object), which is a definite person or thing capable of taking action. The terminal acquires target attribute values. For a given target entity, it determines its attributes, such as the user's mobile phone number, terminal model, active apps, and active services. The integrity of attribute values ​​is checked, and usable data attributes are filtered based on the integrity results. Specifically, the data can include market statistics and data obtained through wireless network base stations and the core network. If there are instances where a large amount of data is missing for a certain attribute during certain periods, that attribute is removed to ensure the usability and reliability of the obtained attributes. A random forest algorithm is used to remove attributes with low correlation.

[0142] like Figure 4b The diagram shown is a flowchart of another possible embodiment, including: identifying behavior types. Specifically, behavior characteristics can represent the mobile communication service-related behavior characteristics that the target object will perform based on the target service data, such as resource data interaction behavior and interactive behavior. Resource data interaction behavior can include resource data interaction characteristics, resource data interaction content, resource data interaction methods, etc. The resource data interaction characteristics can include the interaction value, number of interactions, and interaction time frequency corresponding to the resource data, etc. The resource data interaction content can include interaction content corresponding to different types of services, such as reading type subscriptions, video type interactions, membership type interactions, and game type interactions, etc. The resource data interaction methods include the interaction methods provided by payment institutions for different resource data.

[0143] Establish a database linking actors to their behavioral records. For example, a 5G plan user might have made video-related payments while using 4G, and then made web payments and read ebooks while using 5G.

[0144] Adding time-based statistical markers and quantities can improve the user profile tagging system. Specifically, after adding time-based statistics, the information is improved as follows: Since a 5G Enjoy Package user signed up 5 months ago, the user's total 4G usage time accounted for 60% of the total time spent on 4G, and made 28 video payment service consumption behaviors, with a total duration of 40 hours and an average single service duration of 34 minutes. The total 5G usage time accounted for 40% of the total time spent on 5G, and made web payment, e-book reading and other behaviors under 5G, with a total duration of 10 hours and an average single service duration of 13 minutes, and so on.

[0145] The random forest algorithm is used to remove behaviors with low correlation.

[0146] The method for determining user profile tags provided in this embodiment can also be implemented using the following device, which includes a big data collector, a data cleaner, and a data classifier. The big data collector is used to collect direct data from various dimensions of statistics and acquisition. The data cleaner is used to perform numerical range verification, reasonableness verification, and data integrity verification of the collected data, removing useless and erroneous data. The data classifier is the core of the entire processing, responsible for establishing user profile tags based on reliable data sources and identifying the validity of the tags.

[0147] Specifically, removing behaviors with low correlation by using the random forest algorithm can include:

[0148] Let VIM represent variable importance score and Gini represent the Gini index. Assuming there are m features X1, X2, X3, ..., Xc, calculate the Gini index score for each feature Xj, which is the average change in node split impurity of the j-th feature across all decision trees in the random forest. The average change in node split impurity of the j-th feature across all decision trees in the random forest (which can be denoted as Gini index) is calculated using the following formula. m )

[0149]

[0150] Where K represents the number of categories, p mk This represents the proportion of class k in node m, that is, the probability that two samples randomly drawn from node m will have different class labels.

[0151] Calculate feature X using the following formula. j The importance of node m, i.e., the change in the Gini exponent before and after node m branches (which can be denoted as...). ):

[0152]

[0153] Among them, Gini l and Gini r These represent the Gini indices of the two new nodes after the branching.

[0154] If feature X j If a node in the decision tree appears in set M, then X can be calculated using the following formula. j The importance of the i-th tree (which can be denoted as ) ):

[0155]

[0156] In a random forest with n trees, then:

[0157]

[0158] In other words, the importance of the j-th feature (denoted as VIM) can be calculated using the following formula. j The importance score, or correlation score, is calculated by normalizing all the obtained importance scores to obtain the importance of the feature variable.

[0159]

[0160] The user profile labeling method provided in this embodiment can calculate and rank the importance of relevant features for different application needs, thereby selectively extracting important feature information under different application needs. This reduces inefficient data information in network-collected data, achieves dimensionality reduction of input data for user profile modeling, improves the success rate and accuracy of user profile modeling, and enables precise user profile modeling. It can promptly discover important objective features hidden in the data, ensures the extensibility of the classified profile labels, enriches the profile results, and can be effectively applied to network resource scheduling optimization, network operation, and production and marketing profit-enhancing activities.

[0161] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0162] Based on the same inventive concept, this application also provides a portrait tag determining device for implementing the portrait tag determining method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more portrait tag determining device embodiments provided below can be found in the limitations of the portrait tag determining method described above, and will not be repeated here.

[0163] In one embodiment, such as Figure 5 As shown, a portrait tag determining device 500 is provided, comprising:

[0164] The first determination module 502 is used to identify the target object based on the target business data and target requirement information.

[0165] The second determining module 504 is used to determine at least one initial associated feature of the target object corresponding to the target requirement information.

[0166] The filtering module 506 is used to filter at least one initial associated feature based on preset complete conditions to obtain at least one first associated feature.

[0167] The first calculation module 508 is used to calculate the correlation degree of each first correlation feature based on a preset correlation degree algorithm, filter each first correlation feature based on the correlation degree, obtain target correlation features that meet the preset correlation conditions, and use the target correlation features as the portrait tags of the target object.

[0168] In one embodiment, the filtering module is specifically used for:

[0169] Determine the complete temporal results for each of the initial association features;

[0170] The initial association features that meet the preset completeness conditions for time-complete results are identified as the first association features.

[0171] In one embodiment, the first computing module is specifically used for:

[0172] The first associated feature whose correlation degree is greater than or equal to the preset correlation degree threshold is taken as the target associated feature that satisfies the preset correlation condition.

[0173] In one embodiment, the first computing module is specifically used for:

[0174] Generate a sequence of associated features according to the descending order of the degree of association of each of the first associated features;

[0175] The first number of first associated features in the associated feature sequence that meet the preset association conditions are taken as target associated features.

[0176] In one embodiment, the device further includes:

[0177] The testing module is used to perform tests based on the business data and the target association features of the target object, obtain business test results, and adjust the profile tags of the target object based on the business test results to obtain the adjusted profile tags of the target object.

[0178] In one embodiment, the test module is specifically used for:

[0179] If the business test results do not meet the preset test conditions, return to the step of identifying the target object based on the target business data and target requirement information.

[0180] In one embodiment, the test module is specifically used for:

[0181] Based on the difference between the business test results and the standard results, the profile tags of the target object are adjusted to obtain the adjusted profile tags of the target object.

[0182] In one embodiment, the device further includes:

[0183] The update module is used to, when an update is detected in the business data corresponding to the target object, perform the step of determining the target object and target requirements based on the updated business data, and obtain the updated target association features.

[0184] The various modules in the aforementioned image tag determination device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0185] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 6 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores data related to image tags. The network interface communicates with external terminals via a network connection. When executed by the processor, the computer program implements an image tag determination method.

[0186] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0187] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0188] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0189] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0190] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0191] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0192] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0193] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. An image label determination method characterized by, The method comprises: identifying a target object based on target business data and target demand information; determining at least one initial association feature of the target object corresponding to the target demand information; the initial association feature comprises a behavior feature and an attribute feature, and the behavior feature comprises a behavior type feature and a behavior duration feature; respectively determining a time complete result of each initial association feature; determining an initial association feature with a time complete result meeting a preset complete condition as a first association feature; calculating an association degree of each first association feature based on a preset association degree algorithm, screening each first association feature based on the association degree, obtaining a target association feature meeting a preset association condition, and taking the target association feature as a portrait label of the target object; the preset association degree algorithm is a random forest algorithm, and the association degree represents the importance of the first association feature.

2. The method of claim 1, wherein, The screening of each first association feature based on the association degree to obtain a target association feature meeting a preset association condition comprises: taking a first association feature with an association degree greater than or equal to a preset association degree threshold as a target association feature meeting a preset association condition.

3. The method of claim 1, wherein, The screening of each first association feature based on the association degree to obtain a target association feature meeting a preset association condition comprises: generating an association feature sequence in descending order of the association degrees of the first association features; taking the first association features in the association feature sequence as target association features meeting a preset association condition.

4. The method of claim 1, wherein, The method further comprises: testing according to the business data and the target association feature of the target object to obtain a business test result, and adjusting the portrait label of the target object based on the business test result to obtain an adjusted portrait label of the target object.

5. The method of claim 4, wherein, The adjustment of the portrait label of the target object based on the business test result to obtain an adjusted portrait label of the target object comprises: in a case where the business test result does not meet a preset test condition, returning to execute the step of identifying a target object based on target business data and target demand information.

6. The method of claim 4, wherein, The adjustment of the portrait label of the target object based on the business test result to obtain an adjusted portrait label of the target object comprises: adjusting the portrait label of the target object based on a difference between the business test result and a standard result to obtain an adjusted portrait label of the target object.

7. The method of claim 1, wherein, The method further comprises: in a case where it is detected that the business data corresponding to the target object is updated, executing the steps of determining a target object and target demand based on business data based on the updated business data to obtain an updated target association feature.

8. An image label determination apparatus characterized by comprising: The device comprises: a first determination module configured to identify a target object based on target business data and target demand information; a second determination module configured to determine at least one initial association feature of the target object corresponding to the target demand information; the initial association feature comprises a behavior feature and an attribute feature, and the behavior feature comprises a behavior type feature and a behavior duration feature; The screening module is configured to determine time complete results of the initial correlation features respectively, and determine an initial correlation feature with a time complete result meeting a preset complete condition as a first correlation feature. The first calculation module is configured to calculate a correlation degree of each first correlation feature based on a preset correlation degree algorithm, and screen each first correlation feature based on the correlation degree to obtain a target correlation feature meeting a preset correlation condition, and take the target correlation feature as a portrait label of the target object, wherein the preset correlation degree algorithm is a random forest algorithm, and the correlation degree represents an importance of the first correlation feature. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. The processor executes the computer program to implement the steps of the method in any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method in any one of claims 1 to 7.

11. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method in any one of claims 1 to 7. The computer program is executed by the processor to implement the steps of the method in any one of claims 1 to 7.

Images