A method and system for generating user data profile based on traffic

Abstract

The present invention discloses a method and system for generating a user data portrait based on traffic, and relates to the field of network security technology. It is mainly used to solve the problem of low accuracy of existing user data portrait analysis. It mainly includes obtaining user basic data and user traffic data, and dividing the user traffic data according to the feature classification corresponding to the user basic data to obtain feature traffic data; parsing user behavior characteristics through user basic data, and matching the feature traffic data with the preset behavior traffic threshold corresponding to the user behavior characteristics; if the feature traffic data is greater than or equal to the preset behavior traffic threshold, retraining the user portrait recognition model according to the difference between the feature traffic data and the preset behavior traffic threshold; recognizing the user basic data based on the trained user portrait recognition model, generating a user data portrait recognition result, and generating a visual user portrait according to the user data portrait recognition result.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method and system for generating a user data profile based on traffic. Background Art

[0002] User data profiling is a deep analysis based on a large amount of user information. It uses data mining and machine learning technology to abstract user behaviors, interests, needs and other characteristics into a visual model. This kind of profiling helps companies understand users more accurately and provide personalized products and services, thereby optimizing user experience and improving marketing efficiency.

[0003] At present, the existing commonly used user data portrait generation usually adopts the method of classifying user data for feature extraction, that is, directly classifying the collected user data through the classification model that has been learned according to different feature samples. However, due to the uncertainty of user features, if the samples are not updated for newly emerging features, the classification model cannot accurately classify them, which greatly reduces the accuracy and effectiveness of user data portrait analysis. Summary of the invention

[0004] In view of this, the present invention provides a method and system for generating a user data portrait based on traffic, the main purpose of which is to solve the problem of low accuracy and effectiveness of existing portrait analysis based on user traffic data.

[0005] According to one aspect of the present invention, a method for generating a user data profile based on traffic flow is provided, comprising:

[0006] Obtaining user basic data and user flow data, and dividing the user flow data according to feature categories corresponding to the user basic data to obtain feature flow data;

[0007] Analyzing user behavior characteristics through the user basic data, and matching the characteristic traffic data with a preset behavior traffic threshold corresponding to the user behavior characteristics;

[0008] If the characteristic traffic data is greater than or equal to the preset behavior traffic threshold, the user classification label weight is updated according to the difference between the characteristic traffic data and the preset behavior traffic threshold to retrain the user portrait recognition model;

[0009] The user basic data is identified based on the trained user portrait recognition model to generate a user data portrait recognition result, and a visual user portrait is generated according to the user data portrait recognition result.

[0010] Furthermore, the user traffic data is divided according to the feature classification corresponding to the user basic data to obtain the feature traffic data, including:

[0011] Parsing the data tags of the basic user data, and retrieving feature classifications from a feature classification library according to the data tags, wherein the feature classification library stores different feature classifications corresponding to different data tags, and at least one data tag matches one feature classification;

[0012] Analyzing traffic statistical features of the user traffic data, the traffic statistical features including at least one of time statistical features, network statistical features, application statistical features, and device statistical features;

[0013] The flow statistics features are overlapped and calculated according to the feature classification to determine feature flow data.

[0014] Furthermore, performing overlapping calculation on the traffic statistical features according to the feature classification to determine the feature traffic data includes:

[0015] Selecting at least one target flow value matching the feature classification from at least one of the time statistical feature, the network statistical feature, the application statistical feature, and the device statistical feature;

[0016] The target traffic values ​​are added in a weighted summation method to obtain the characteristic traffic data, and the weight coefficient in the weighted summation method is determined based on time characteristics, network characteristics, application characteristics, and device characteristics.

[0017] Furthermore, the user behavior characteristics include access behavior data, download behavior data, transaction behavior data, and input behavior data, and the matching of the characteristic traffic data with a preset behavior traffic threshold corresponding to the user behavior characteristics includes:

[0018] Determine a preset access flow threshold, a preset download flow threshold, a preset transaction flow threshold, and a preset entry flow threshold corresponding to the access behavior data, download behavior data, transaction behavior data, and entry behavior data;

[0019] Compare the characteristic traffic data with the preset access traffic threshold, the preset download traffic threshold, the preset transaction traffic threshold, and the preset input traffic threshold respectively;

[0020] Wherein, the determining of the preset access flow threshold, the preset download flow threshold, the preset transaction flow threshold, and the preset input flow threshold corresponding to the access behavior data, the download behavior data, the transaction behavior data, and the input behavior data comprises:

[0021] Determine a preset access flow threshold matching the access behavior data according to the time node, login address information, and webpage business information;

[0022] Determine a preset download flow threshold matching the download behavior data according to web page service information, download object information, and network connection information;

[0023] Determine a preset transaction flow threshold that matches the transaction behavior data according to the transaction object information, transaction party information, and transaction volume information;

[0024] A preset input flow threshold matching the input behavior data is determined according to the webpage business information, the input object information, and the network connection information.

[0025] Furthermore, after the user classification label weight is updated according to the difference between the characteristic traffic data and the preset behavior traffic threshold value to retrain the user portrait recognition model, the method further includes:

[0026] Obtain updated user data portrait sample data, and train the user portrait recognition model according to the updated user classification label weights and the updated user data portrait sample data;

[0027] When, during the process of updating and training the user portrait recognition model, the classification samples corresponding to the user classification label weights are greater than the preset number of classifications, the classification categories are added to the classification samples according to the newly added classification labels, and the newly added classification labels are determined by extracting keywords from the classification samples.

[0028] Furthermore, the method further comprises:

[0029] Performing semantic division on the text words in the classification sample to obtain classification keywords;

[0030] Matching and filtering all the portrait keywords in the updated user data portrait sample data with the classification keywords;

[0031] The classification keywords that do not match the portrait keywords are clustered and newly added classification labels are determined.

[0032] Furthermore, the method further comprises:

[0033] If the characteristic traffic data is less than the preset behavior traffic threshold, the user basic data is identified based on the user portrait recognition model that has completed model training to generate a user data portrait recognition result;

[0034] Generating a visualized user portrait according to the user data portrait recognition result includes:

[0035] Different user classification results in the user data portrait recognition result are rendered, and the user classification results corresponding to the user classification label weights are marked to generate a visual user portrait to display the traffic characteristics as visual labels.

[0036] According to another aspect of the present invention, a system for generating a user data profile based on traffic is provided, comprising:

[0037] A division module is used to obtain user basic data and user flow data, and divide the user flow data according to the feature classification corresponding to the user basic data to obtain feature flow data;

[0038] A matching module, used to analyze user behavior characteristics through the user basic data, and match the characteristic traffic data with a preset behavior traffic threshold corresponding to the user behavior characteristics;

[0039] An updating module, configured to update the user classification label weight according to the difference between the characteristic traffic data and the preset behavior traffic threshold if the characteristic traffic data is greater than or equal to the preset behavior traffic threshold, so as to retrain the user portrait recognition model;

[0040] The recognition module is used to recognize the user basic data based on the trained user portrait recognition model, generate a user data portrait recognition result, and generate a visual user portrait according to the user data portrait recognition result.

[0041] Furthermore, the division module includes:

[0042] A first parsing unit, configured to parse the data tags of the basic user data, and retrieve feature classifications from a feature classification library according to the data tags, wherein the feature classification library stores different feature classifications corresponding to different data tags, and at least one data tag matches one feature classification;

[0043] A second parsing unit, configured to parse traffic statistical features of the user traffic data, wherein the traffic statistical features include at least one of a time statistical feature, a network statistical feature, an application statistical feature, and a device statistical feature;

[0044] A calculation unit is used to perform overlapping calculations on the traffic statistical features according to the feature classification to determine characteristic traffic data.

[0045] Furthermore, in a specific application scenario, the calculation unit is specifically used to select at least one target traffic value that matches the feature classification from at least one of the time statistical features, the network statistical features, the application statistical features, and the device statistical features; the target traffic values ​​are added in a weighted summation method to obtain the characteristic traffic data, and the weight coefficient in the weighted summation method is determined based on the time features, network features, application features, and device features.

[0046] Furthermore, the matching module includes:

[0047] A first determining unit, used to determine a preset access flow threshold, a preset download flow threshold, a preset transaction flow threshold, and a preset entry flow threshold corresponding to the access behavior data, the download behavior data, the transaction behavior data, and the entry behavior data;

[0048] A comparison unit is used to compare the characteristic traffic data with a preset access traffic threshold, a preset download traffic threshold, a preset transaction traffic threshold, and a preset input traffic threshold respectively; wherein the determining of the preset access traffic threshold, the preset download traffic threshold, the preset transaction traffic threshold, and the preset input traffic threshold corresponding to the access behavior data, the download behavior data, the transaction behavior data, and the input behavior data comprises:

[0049] A second determination unit is used to determine a preset access flow threshold matching the access behavior data according to the time node, login address information, and webpage service information;

[0050] A third determining unit, configured to determine a preset download flow threshold matching the download behavior data according to webpage service information, download object information, and network connection information;

[0051] A fourth determining unit, configured to determine a preset transaction flow threshold matching the transaction behavior data according to the transaction object information, the transaction party information, and the transaction volume information;

[0052] The fifth determining unit is used to determine a preset input flow threshold matching the input behavior data according to the webpage business information, the input object information, and the network connection information.

[0053] Furthermore, the system further comprises:

[0054] An acquisition module, used to acquire updated user data portrait sample data, and train a user portrait recognition model according to the updated user classification label weights and the updated user data portrait sample data;

[0055] The classification module is used to add classification categories to the classification samples according to the newly added classification labels when the classification samples corresponding to the user classification label weights are greater than the preset number of classifications during the update training of the user portrait recognition model. The newly added classification labels are determined by extracting keywords from the classification samples.

[0056] Furthermore, the system further comprises:

[0057] A semantic segmentation module, used to perform semantic segmentation on the text words in the classification sample to obtain classification keywords;

[0058] A filtering module, used for matching and filtering all portrait keywords in the updated user data portrait sample data with the classification keywords;

[0059] The determination module is used to cluster the classification keywords that do not match the portrait keywords and determine new classification labels.

[0060] Furthermore, the system further comprises:

[0061] A generation module, configured to identify the user basic data based on the user portrait recognition model that has completed model training, and generate a user data portrait recognition result if the characteristic traffic data is less than the preset behavior traffic threshold;

[0062] The recognition module further includes a rendering unit;

[0063] The rendering unit is used to render different user classification results in the user data portrait recognition result, and mark the user classification results corresponding to the user classification label weights, generate a visual user portrait, and display the traffic characteristics as visual labels.

[0064] According to another aspect of the present invention, a storage medium is provided, wherein at least one executable instruction is stored in the storage medium, and the executable instruction enables a processor to perform operations corresponding to the above-mentioned method for generating a user data profile based on traffic.

[0065] According to another aspect of the present invention, there is provided a terminal, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus;

[0066] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the above-mentioned traffic-based user data portrait generation method.

[0067] By means of the above technical solution, the technical solution provided by the embodiment of the present invention has at least the following advantages:

[0068] The present invention provides a method and system for generating a user data portrait based on traffic. The embodiment of the present invention obtains user basic data and user traffic data, and divides the user traffic data according to the feature classification corresponding to the user basic data to obtain feature traffic data; analyzes user behavior characteristics through the user basic data, and matches the feature traffic data with a preset behavior traffic threshold corresponding to the user behavior characteristics; if the feature traffic data is greater than or equal to the preset behavior traffic threshold, updates the user classification label weight according to the difference between the feature traffic data and the preset behavior traffic threshold to retrain a user portrait recognition model; recognizes the user basic data based on the trained user portrait recognition model to generate a user data portrait recognition result, and generates a visualized user portrait according to the user data portrait recognition result, so as to recognize the mismatch between the user basic data and the user portrait recognition model, and timely update and train the user portrait recognition model, so that the user portrait recognition model has a higher matching degree with the user basic data, thereby improving the accuracy and effectiveness of user portrait recognition.

[0069] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0071] Figure 1 A flow chart of a method for generating a user data profile based on traffic flow is shown in an embodiment of the present invention;

[0072] Figure 2 Another flow chart of a method for generating a user data profile based on traffic flow provided by an embodiment of the present invention is shown;

[0073] Figure 3 A block diagram of a system for generating a user data profile based on traffic flow is shown in an embodiment of the present invention;

[0074] Figure 4 A schematic diagram of the structure of a terminal provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0075] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0076] The embodiment of the present invention provides a method for generating a user data profile based on traffic flow, such as Figure 1 As shown, the method includes:

[0077] 101. Obtain user basic data and user flow data, and classify the user flow data according to feature categories corresponding to the user basic data to obtain feature flow data.

[0078] In an embodiment of the present invention, in order to extract a user data portrait, it is necessary to determine a feature classification based on the user characteristics reflected by the user basic data, and to divide the user traffic data based on the feature classification to obtain a corresponding relationship between traffic consumption and the features. The user basic data includes the user's account information, gender, age, occupation, permanent residence, device information that generates traffic consumption, and application information bound to the user account in a historical time period. In order to decompose the user traffic data into feature dimensions, at least one feature that can be represented by the user basic data is determined, and the user's feature classification is determined based on this feature, and then the user traffic data is divided into traffic data corresponding to at least one feature, i.e., feature traffic data, according to the feature classification.

[0079] It should be noted that user basic data can reflect user habits, user preferences, user identity and other characteristics. These can be used to determine a rough feature classification, and user traffic data can be divided according to the feature classification. The user traffic consumption can be associated with the user's habits and preferences, providing data support for subsequent traffic-based analysis of user data behavior.

[0080] 102. Analyze user behavior characteristics through the user basic data, and match the characteristic traffic data with a preset behavior traffic threshold corresponding to the user behavior characteristics.

[0081] In an embodiment of the present invention, corresponding preset behavior traffic thresholds are pre-configured for different user behavior characteristics. This preset behavior traffic threshold is determined based on traffic usage statistics of different user behavior characteristics, that is, the traffic usage corresponding to a user with a certain user behavior characteristic will be less than the preset behavior traffic threshold. If it is not less than, it indicates that the user behavior characteristic recognition result is not accurate enough to accurately describe the user's behavior. In order to determine the adaptability of the current version of the user portrait recognition model to the current user, the user basic data is identified and processed based on the current version of the user portrait recognition model to obtain the user behavior characteristics, and the characteristic traffic data is matched with the preset behavior traffic threshold corresponding to the user behavior characteristics.

[0082] It should be noted that user behavior characteristics are not static and may change significantly due to changes in user preferences, changes in the nature of work, or changes in user leisure and entertainment habits and working methods brought about by the promotion of emerging applications. If the user portrait recognition model is not updated for a long time, the model is likely to be unsuitable for users, but frequent updates of model parameters will also cause unnecessary resource consumption. By configuring a preset behavior traffic threshold to determine the applicability of user behavior characteristics from the dimension of traffic usage, the necessity of model updates can be accurately identified, thereby accurately determining the timing of model updates.

[0083] 103. If the characteristic traffic data is greater than or equal to the preset behavior traffic threshold, the user classification label weight is updated according to the difference between the characteristic traffic data and the preset behavior traffic threshold to retrain the user portrait recognition model.

[0084] In the embodiment of the present invention, the basic model of the user portrait recognition model is a binary tree model. The binary tree is a tree structure in which each node has at most two child nodes. How the two child nodes of each node grow and whether to stop growing need to be decided based on the user classification label weight. The user classification label weight is updated, and the current version of the user portrait recognition model is trained according to the updated user classification label weight to complete the update of the model. The difference between the characteristic traffic data and the preset behavior traffic threshold can characterize the degree of accuracy deviation of the recognition result of the user portrait recognition model. The larger the difference, the larger the deviation value. Therefore, updating the user classification label weight according to the difference can achieve accurate adjustment of the model user classification label weight, greatly shorten the model training time, and thus improve the model training efficiency. In addition, updating and training the split nodes of the binary tree based on the characteristic traffic data of the current user can improve the classification accuracy of the model, that is, improve the granularity of the user portrait recognition result.

[0085] 104. Identify the user basic data based on the trained user portrait recognition model, generate a user data portrait recognition result, and generate a visualized user portrait according to the user data portrait recognition result.

[0086] In an embodiment of the present invention, after the user portrait recognition model is updated and trained based on the user classification label weight, the trained user portrait recognition model is used to re-recognize the user basic data to obtain the user data portrait recognition result. Among them, the user data portrait recognition result may include network usage habits, points of interest, social interaction characteristics, and the user's personality characteristics, values, life attitudes and emotional tendencies. In order to further improve the viewing convenience of the user data portrait recognition results, the user data portrait recognition results are converted into a visualized user portrait through a visualization generation tool. Among them, the visualization generation tool can be HubSpotBuyerPersona, Tableau, etc., and the example of the present invention is not specifically limited.

[0087] It should be noted that since the user portrait recognition model is trained based on the user classification label weights after updating the difference between the characteristic traffic data and the preset behavior traffic threshold, it has a higher degree of match with the user and can better capture the characteristics of the user's basic data, thereby obtaining more accurate and fine-grained user data portrait recognition results, thereby greatly improving the accuracy of user data portrait recognition.

[0088] In an embodiment of the present invention, for further explanation and limitation, as Figure 2 As shown, in step 101, the user traffic data is divided according to the feature classification corresponding to the user basic data, and the characteristic traffic data obtained includes:

[0089] 201. Parse the data tags of the basic user data, and retrieve feature classifications from a feature classification library according to the data tags;

[0090] 202. Analyze traffic statistics characteristics of the user traffic data;

[0091] 203: Perform overlapping calculation on the traffic statistical features according to the feature classification to determine feature traffic data.

[0092] In an embodiment of the present invention, the information in the user basic data is parsed into data tags. For example, if the user basic data includes bound social programs A, B, and C, the data tag is a social data tag; if it includes bound video playback programs E, D, and G, the data tag is an entertainment playback data tag; if it includes game software K and H, the data tag is a game data tag. The determination of the social software feature classification is based on the mapping relationship between the feature classification and the data tag pre-constructed in the feature classification library. Among them, the feature classification library stores different feature classifications corresponding to different data tags, and at least one data tag matches a feature classification, that is, based on the user basic data, multiple data tags can be determined, and some of these data tags can independently represent a data classification of the user, and some require a combination of multiple data tags to jointly represent a data classification of the user. For example, the entertainment playback data tag, the male data tag, and the game data tag can determine that the data classification is a game feature classification; the entertainment playback data tag and the female data tag can determine that the data classification is a drama chasing feature classification; the social data tag can determine the social feature classification. After determining the feature classification, the traffic statistical features of the user traffic data are further analyzed, wherein the data traffic statistical features include at least one of time statistical features, network statistical features, application statistical features, and device statistical features. Among them, the time statistical features are used to characterize the time period distribution features in the user traffic data. For example, the use of user traffic data is concentrated from 12:00 noon to 13:00 pm and from 17:00 pm to 22:00 pm, then the time statistical features are non-working hours. The network statistical features are used to characterize the network usage types corresponding to different feature classifications (different operator networks, public networks or home networks, mobile data networks or broadband networks, etc.). For example, if the user account has logged in through the public network and the mobile data network, the network statistical features include the public network and mobile data. The application statistical features are used to characterize the applications involved in the traffic consumption of the user traffic data. For example, the traffic consumption in the user traffic data is mainly concentrated in social programs A, B, and C, then the application statistical features include social programs A, B, and C. The device statistical features are used for the traffic consumption of different devices in the user traffic data, for example, the traffic used by the user account through a laptop and the traffic used through a smartphone, etc. Then, the feature classification and traffic statistical features are matched, and the feature traffic data is determined based on the overlapping parts of the features.

[0093] In one embodiment of the present invention, for further explanation and limitation, the step of performing overlapping calculation on the traffic statistical features according to the feature classification to determine the feature traffic data includes:

[0094] Selecting at least one target flow value matching the feature classification from at least one of the time statistical feature, the network statistical feature, the application statistical feature, and the device statistical feature;

[0095] The target flow values ​​are added in a weighted summation manner to obtain the characteristic flow data.

[0096] In an embodiment of the present invention, one or more of the time statistical features, network statistical features, application statistical features, and device statistical features are matched with the feature classification to determine at least one target traffic value. The target traffic value is the traffic value generated by the overlapped portion of the time statistical features, network statistical features, application statistical features, and device statistical features with the feature classification. For example, the application statistical features include social software A, B, and C, and the feature classification of the current user basic data is the social feature classification. The two overlap in social software A, B, and C, then the traffic usage values ​​of social software A, B, and C are all target traffic values. There are multiple target traffic values ​​obtained, and multiple statistical features can be matched. After determining the target traffic value, the target traffic is multiplied by the weight coefficient of the corresponding statistical feature, and then different target traffics are added to obtain feature traffic data. The weight coefficient in the weight summation method is determined based on the time feature, network feature, application feature, and device feature, that is, the weight coefficient corresponding to the feature can be customized according to the importance of the time feature, network feature, application feature, and device feature in the specific application scenario. For example, in the scenario of capturing the user's dependence on application behavior, the importance of time features and application features is higher than that of network features and device features; in the scenario of capturing the user's application usage environment, the importance of network features and device features is higher than that of time features and application features.

[0097] It should be noted that by determining the feature classification through user basic data and matching the traffic statistical features and feature classification with feature overlap, the feature classification can be described from the traffic dimension, thereby providing accurate data basis for the subsequent effective identification of the user portrait recognition model.

[0098] In one embodiment of the present invention, for further explanation and limitation, the step of matching the characteristic traffic data with a preset behavior traffic threshold corresponding to the user behavior feature includes:

[0099] Determine a preset access flow threshold, a preset download flow threshold, a preset transaction flow threshold, and a preset entry flow threshold corresponding to the access behavior data, download behavior data, transaction behavior data, and entry behavior data;

[0100] Compare the characteristic traffic data with the preset access traffic threshold, the preset download traffic threshold, the preset transaction traffic threshold, and the preset input traffic threshold respectively;

[0101] Wherein, the determining of the preset access flow threshold, the preset download flow threshold, the preset transaction flow threshold, and the preset input flow threshold corresponding to the access behavior data, the download behavior data, the transaction behavior data, and the input behavior data comprises:

[0102] Determine a preset access flow threshold matching the access behavior data according to the time node, login address information, and webpage business information;

[0103] Determine a preset download flow threshold matching the download behavior data according to web page service information, download object information, and network connection information;

[0104] Determine a preset transaction flow threshold that matches the transaction behavior data according to the transaction object information, transaction party information, and transaction volume information;

[0105] A preset input flow threshold matching the input behavior data is determined according to the webpage business information, the input object information, and the network connection information.

[0106] In an embodiment of the present invention, the user behavior characteristics include access behavior data, download behavior data, transaction behavior data, and input behavior data. The preset access flow threshold, the preset download flow threshold, the preset transaction flow threshold, and the preset input flow threshold are used to limit the corresponding behavior data in the above user behavior characteristics. Among them, the preset access flow threshold is determined according to the time node, the login address information, and the web page business information. For example, the time node is a work node, the login address information is a company address, if the web page business information is entertainment, then the preset access flow threshold is low; if the web page business information is a business handling website, then the preset access flow threshold is high. The preset download flow threshold is determined according to the web page business information, the download object information, and the network connection information. The network connection information can be whether the connected network is an office network or a home network. For example, when the network connection information is an office network, if the web page business information and the download object information are entertainment-related, then the preset download flow threshold needs to be configured lower; if the web page business information and the download object information are office business-related, then the preset download flow threshold can be configured higher. The preset transaction flow threshold is determined according to the transaction object information, the transaction party information, and the transaction volume information. For example, each transaction product (object) corresponds to a transaction volume statistic, and the transaction party corresponds to a trust level. The transaction volume statistic (peak or average) of the transaction object can be used as the basic data, and the trust level of the transaction party can be used as a coefficient to correct the basic data to determine the preset transaction traffic threshold. The preset input traffic threshold is the traffic threshold for users to upload data to a page or application, and is determined according to the web page business information, input object information, and network connection information. For example, when the network connection is an office network, if the web page business information is entertainment and the input object is video, the preset input traffic threshold needs to be configured lower. If the web page business information is government affairs and the input object is a file, the preset input traffic threshold can be configured higher.

[0107] It should be noted that by configuring the preset behavior traffic threshold based on multi-dimensional information such as different web page business information, network connection information, etc., the preset behavior traffic threshold can be made more flexible to meet the needs of different application scenarios. When compared with the characteristic traffic data, the accuracy of the recognition results of the user portrait recognition model can be more accurately identified, providing more accurate data basis for subsequent model updates.

[0108] In one embodiment of the present invention, for further explanation and limitation, after the step of updating the user classification label weight according to the difference between the characteristic traffic data and the preset behavior traffic threshold to retrain the user portrait recognition model, the method further includes:

[0109] Obtain updated user data portrait sample data, and train the user portrait recognition model according to the updated user classification label weights and the updated user data portrait sample data;

[0110] When, during the process of updating and training the user portrait recognition model, the classification samples corresponding to the user classification label weights are greater than the preset number of classifications, the classification categories are added to the classification samples according to the newly added classification labels.

[0111] In an embodiment of the present invention, the preset number of classifications is used to characterize the threshold of the number of samples in each classification. When the number of samples under a certain classification exceeds the threshold, it indicates that the current classification needs to be further refined. In the process of updating and training the user portrait recognition model, different classification samples are obtained, the number of classification samples under each classification is counted, and this number is compared with the preset number of classifications. If the number of classification samples under the current classification is less than or equal to the preset number of classifications, it indicates that the update training of this model only needs to adjust the content of the classification category, but does not need to update the number of classification categories; if the number of classification samples under the current classification is greater than the preset number of classifications, it indicates that the classification accuracy of the classification samples output by the updated user portrait recognition model needs to be higher, and the granularity needs to be further refined. It is necessary to increase the classification category of the classification sample according to the newly added classification label. For example, if the classification sample is a frequently used social chat software, and the number of samples that meet this classification category is greater than the preset number of classifications, the samples of the frequently used social chat software can be further divided into frequently used social chat software during working hours and frequently used social chat software during non-working hours according to time. For another example, the classification sample is the use of application video playback software, which can be further divided into watching two-dimensional animation, watching short videos, watching live broadcasts, etc. through video playback software.

[0112] It should be noted that the newly added classification labels are determined by extracting keywords from the classification samples. For example, keywords are extracted from the classification samples, the keywords are clustered, and the clustering results are used as newly added classification labels. By adding classification labels, the update training of the user portrait recognition model can not only update the model parameters, but also improve the classification recognition granularity, thereby improving the classification recognition accuracy, thereby meeting more sophisticated user portrait recognition needs.

[0113] In an embodiment of the present invention, for further explanation and limitation, the method further includes:

[0114] Performing semantic division on the text words in the classification sample to obtain classification keywords;

[0115] Matching and filtering all the portrait keywords in the updated user data portrait sample data with the classification keywords;

[0116] The classification keywords that do not match the portrait keywords are clustered and newly added classification labels are determined.

[0117] In the embodiment of the present invention, semantic segmentation and keyword extraction are performed on the text words in the classification sample to obtain classification keywords. All the image keywords in the updated user data portrait sample data are retrieved. All the image keywords are used to characterize the existing sample classification categories. The classification keywords that do not match any image keywords are the keywords that need to be refined. Therefore, these keywords are further clustered to obtain new classification labels.

[0118] In an embodiment of the present invention, for further explanation and limitation, the method further includes:

[0119] If the characteristic traffic data is less than the preset behavior traffic threshold, the user basic data is identified based on the user portrait recognition model that has completed model training to generate a user data portrait recognition result;

[0120] Generating a visualized user portrait according to the user data portrait recognition result includes:

[0121] Different user classification results in the user data portrait recognition result are rendered, and the user classification results corresponding to the user classification label weights are marked to generate a visual user portrait to display the traffic characteristics as visual labels.

[0122] In an embodiment of the present invention, if the characteristic traffic data is less than the preset behavior traffic threshold, indicating that the current version of the user portrait recognition model does not need to be updated, the user basic data is continuously recognized based on the current user portrait recognition model to obtain the user data portrait recognition result. In order to display the user portrait more clearly and concisely, the user classification results corresponding to the user classification label weights are marked, and the traffic characteristics are displayed as visual labels. For example, according to the user classification results, users of different classifications are respectively displayed with a user portrait, and each traffic feature is displayed around the corresponding user portrait in the form of a bubble chart around the user portrait.

[0123] The present invention provides a method for generating a user data portrait based on traffic. The embodiment of the present invention obtains user basic data and user traffic data, and divides the user traffic data according to the feature classification corresponding to the user basic data to obtain feature traffic data; analyzes user behavior characteristics through the user basic data, and matches the feature traffic data with a preset behavior traffic threshold corresponding to the user behavior characteristics; if the feature traffic data is greater than or equal to the preset behavior traffic threshold, updates the user classification label weight according to the difference between the feature traffic data and the preset behavior traffic threshold to retrain a user portrait recognition model; recognizes the user basic data based on the trained user portrait recognition model to generate a user data portrait recognition result, and generates a visualized user portrait according to the user data portrait recognition result, so as to recognize the mismatch between the user basic data and the user portrait recognition model, and timely update and train the user portrait recognition model, so that the user portrait recognition model has a higher matching degree with the user basic data, thereby improving the accuracy and effectiveness of user portrait recognition.

[0124] Furthermore, as a response to the above Figure 1 The embodiment of the present invention provides a system for generating user data profiles based on traffic flow, such as Figure 3 As shown, the system includes:

[0125] The division module 31 is used to obtain user basic data and user flow data, and divide the user flow data according to the feature classification corresponding to the user basic data to obtain feature flow data;

[0126] A matching module 32, configured to analyze user behavior characteristics through the user basic data, and match the characteristic flow data with a preset behavior flow threshold corresponding to the user behavior characteristics;

[0127] An updating module 33, configured to update the user classification label weight according to the difference between the characteristic traffic data and the preset behavior traffic threshold if the characteristic traffic data is greater than or equal to the preset behavior traffic threshold, so as to retrain the user portrait recognition model;

[0128] The recognition module 34 is used to recognize the user basic data based on the trained user portrait recognition model, generate a user data portrait recognition result, and generate a visual user portrait according to the user data portrait recognition result.

[0129] Furthermore, the division module 31 includes:

[0130] A first parsing unit, configured to parse the data tags of the basic user data, and retrieve feature classifications from a feature classification library according to the data tags, wherein the feature classification library stores different feature classifications corresponding to different data tags, and at least one data tag matches one feature classification;

[0131] A second parsing unit, configured to parse traffic statistical features of the user traffic data, wherein the traffic statistical features include at least one of a time statistical feature, a network statistical feature, an application statistical feature, and a device statistical feature;

[0132] A calculation unit is used to perform overlapping calculations on the traffic statistical features according to the feature classification to determine characteristic traffic data.

[0133] Furthermore, in a specific application scenario, the calculation unit is specifically used to select at least one target traffic value that matches the feature classification from at least one of the time statistical features, the network statistical features, the application statistical features, and the device statistical features; the target traffic values ​​are added in a weighted summation method to obtain the characteristic traffic data, and the weight coefficient in the weighted summation method is determined based on the time features, network features, application features, and device features.

[0134] Furthermore, the matching module 32 includes:

[0135] A first determining unit, used to determine a preset access flow threshold, a preset download flow threshold, a preset transaction flow threshold, and a preset entry flow threshold corresponding to the access behavior data, the download behavior data, the transaction behavior data, and the entry behavior data;

[0136] A comparison unit is used to compare the characteristic traffic data with a preset access traffic threshold, a preset download traffic threshold, a preset transaction traffic threshold, and a preset input traffic threshold respectively; wherein the determining of the preset access traffic threshold, the preset download traffic threshold, the preset transaction traffic threshold, and the preset input traffic threshold corresponding to the access behavior data, the download behavior data, the transaction behavior data, and the input behavior data comprises:

[0137] A second determination unit is used to determine a preset access flow threshold matching the access behavior data according to the time node, login address information, and webpage service information;

[0138] A third determining unit, configured to determine a preset download flow threshold matching the download behavior data according to webpage service information, download object information, and network connection information;

[0139] A fourth determining unit, configured to determine a preset transaction flow threshold matching the transaction behavior data according to the transaction object information, the transaction party information, and the transaction volume information;

[0140] The fifth determining unit is used to determine a preset input flow threshold matching the input behavior data according to the webpage business information, the input object information, and the network connection information.

[0141] Furthermore, the system further comprises:

[0142] An acquisition module, used to acquire updated user data portrait sample data, and train a user portrait recognition model according to the updated user classification label weights and the updated user data portrait sample data;

[0143] The classification module is used to add classification categories to the classification samples according to the newly added classification labels when the classification samples corresponding to the user classification label weights are greater than the preset number of classifications during the update training of the user portrait recognition model. The newly added classification labels are determined by extracting keywords from the classification samples.

[0144] Furthermore, the system further comprises:

[0145] A semantic segmentation module, used to perform semantic segmentation on the text words in the classification sample to obtain classification keywords;

[0146] A filtering module, used for matching and filtering all portrait keywords in the updated user data portrait sample data with the classification keywords;

[0147] The determination module is used to cluster the classification keywords that do not match the portrait keywords and determine new classification labels.

[0148] Furthermore, the system further comprises:

[0149] A generation module, configured to identify the user basic data based on the user portrait recognition model that has completed model training, and generate a user data portrait recognition result if the characteristic traffic data is less than the preset behavior traffic threshold;

[0150] The identification module 34 further includes a rendering unit;

[0151] The rendering unit is used to render different user classification results in the user data portrait recognition result, and mark the user classification results corresponding to the user classification label weights, generate a visual user portrait, and display the traffic characteristics as visual labels.

[0152] The present invention provides a traffic-based user data portrait generation system. The embodiment of the present invention obtains user basic data and user traffic data, and divides the user traffic data according to the feature classification corresponding to the user basic data to obtain feature traffic data; analyzes user behavior characteristics through the user basic data, and matches the feature traffic data with a preset behavior traffic threshold corresponding to the user behavior characteristics; if the feature traffic data is greater than or equal to the preset behavior traffic threshold, updates the user classification label weight according to the difference between the feature traffic data and the preset behavior traffic threshold to retrain the user portrait recognition model; recognizes the user basic data based on the trained user portrait recognition model to generate a user data portrait recognition result, and generates a visualized user portrait according to the user data portrait recognition result, so as to recognize the mismatch between the user basic data and the user portrait recognition model, and timely update and train the user portrait recognition model, so that the user portrait recognition model has a higher matching degree with the user basic data, thereby improving the accuracy and effectiveness of user portrait recognition.

[0153] According to one embodiment of the present invention, a storage medium is provided, wherein the storage medium stores at least one executable instruction, and the computer executable instruction can execute the traffic-based user data profile generation method in any of the above method embodiments.

[0154] Figure 4 A schematic diagram of the structure of a terminal provided according to an embodiment of the present invention is shown. The specific embodiment of the present invention does not limit the specific implementation of the terminal.

[0155] like Figure 4 As shown, the terminal may include: a processor (processor) 402 , a communication interface (Communications Interface) 404 , a memory (memory) 406 , and a communication bus 408 .

[0156] The processor 402 , the communication interface 404 , and the memory 406 communicate with each other via a communication bus 408 .

[0157] The communication interface 404 is used for network communication with other devices such as a client or other servers.

[0158] Processor 402 is used to execute program 410, and specifically can execute the relevant steps in the above-mentioned traffic-based user data portrait generation method embodiment.

[0159] Specifically, the program 410 may include program codes, which include computer operation instructions.

[0160] The processor 402 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiment of the present invention. The one or more processors included in the terminal may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0161] The memory 406 is used to store the program 410. The memory 406 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0162] The program 410 may be specifically configured to enable the processor 402 to perform the following operations:

[0163] Obtaining user basic data and user flow data, and dividing the user flow data according to feature categories corresponding to the user basic data to obtain feature flow data;

[0164] Analyzing user behavior characteristics through the user basic data, and matching the characteristic traffic data with a preset behavior traffic threshold corresponding to the user behavior characteristics;

[0165] If the characteristic traffic data is greater than or equal to the preset behavior traffic threshold, the user classification label weight is updated according to the difference between the characteristic traffic data and the preset behavior traffic threshold to retrain the user portrait recognition model;

[0166] The user basic data is identified based on the trained user portrait recognition model to generate a user data portrait recognition result, and a visual user portrait is generated according to the user data portrait recognition result.

[0167] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing system, they can be concentrated on a single computing system, or distributed on a network composed of multiple computing systems, and optionally, they can be implemented by a program code executable by a computing system, so that they can be stored in a storage system and executed by the computing system, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0168] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for generating a user data profile based on traffic, characterized in that: include: Obtaining user basic data and user flow data, and dividing the user flow data according to feature categories corresponding to the user basic data to obtain feature flow data; Analyzing user behavior characteristics through the user basic data, and matching the characteristic traffic data with a preset behavior traffic threshold corresponding to the user behavior characteristics, wherein the user behavior characteristics include access behavior data, download behavior data, transaction behavior data, and input behavior data; If the characteristic traffic data is greater than or equal to the preset behavior traffic threshold, the user classification label weight is updated according to the difference between the characteristic traffic data and the preset behavior traffic threshold to retrain the user portrait recognition model; Identify the user basic data based on the trained user portrait recognition model, generate a user data portrait recognition result, and generate a visual user portrait according to the user data portrait recognition result; The user traffic data is divided according to the feature classification corresponding to the user basic data to obtain the feature traffic data, including: Parsing the data tags of the basic user data, and retrieving feature classifications from a feature classification library according to the data tags, wherein the feature classification library stores different feature classifications corresponding to different data tags, and at least one data tag matches one feature classification; Analyzing traffic statistical features of the user traffic data, the traffic statistical features including at least one of time statistical features, network statistical features, application statistical features, and device statistical features; Perform overlapping calculation on the traffic statistical features according to the feature classification to determine feature traffic data; The matching of the characteristic traffic data with a preset behavior traffic threshold corresponding to the user behavior characteristic includes: Determine a preset access flow threshold, a preset download flow threshold, a preset transaction flow threshold, and a preset entry flow threshold corresponding to the access behavior data, download behavior data, transaction behavior data, and entry behavior data; Compare the characteristic traffic data with the preset access traffic threshold, the preset download traffic threshold, the preset transaction traffic threshold, and the preset input traffic threshold respectively; Wherein, the determining of the preset access flow threshold, the preset download flow threshold, the preset transaction flow threshold, and the preset input flow threshold corresponding to the access behavior data, the download behavior data, the transaction behavior data, and the input behavior data comprises: Determine a preset access flow threshold matching the access behavior data according to the time node, login address information, and webpage business information; Determine a preset download flow threshold matching the download behavior data according to web page service information, download object information, and network connection information; Determine a preset transaction flow threshold that matches the transaction behavior data according to the transaction object information, transaction party information, and transaction volume information; A preset input flow threshold matching the input behavior data is determined according to the webpage business information, the input object information, and the network connection information.

2. The method according to claim 1, characterized in that The performing overlapping calculation on the traffic statistical features according to the feature classification to determine the feature traffic data comprises: Selecting at least one target flow value matching the feature classification from at least one of the time statistical feature, the network statistical feature, the application statistical feature, and the device statistical feature; The target traffic values ​​are added in a weighted summation method to obtain the characteristic traffic data, and the weight coefficient in the weighted summation method is determined based on time characteristics, network characteristics, application characteristics, and device characteristics.

3. The method according to claim 1, characterized in that After the user classification label weight is updated according to the difference between the characteristic traffic data and the preset behavior traffic threshold value to retrain the user portrait recognition model, the method further includes: Obtain updated user data portrait sample data, and train the user portrait recognition model according to the updated user classification label weights and the updated user data portrait sample data; When, during the process of updating and training the user portrait recognition model, the classification samples corresponding to the user classification label weights are greater than the preset number of classifications, the classification categories are added to the classification samples according to the newly added classification labels, and the newly added classification labels are determined by extracting keywords from the classification samples.

4. The method according to claim 3, characterized in that The method further comprises: Performing semantic division on the text words in the classification sample to obtain classification keywords; Matching and filtering all the portrait keywords in the updated user data portrait sample data with the classification keywords; The classification keywords that do not match the portrait keywords are clustered and newly added classification labels are determined.

5. The method according to any one of claims 1 to 4, characterized in that: The method further comprises: If the characteristic traffic data is less than the preset behavior traffic threshold, the user basic data is identified based on the user portrait recognition model that has completed model training to generate a user data portrait recognition result; Generating a visualized user portrait according to the user data portrait recognition result includes: Different user classification results in the user data portrait recognition result are rendered, and the user classification results corresponding to the user classification label weights are marked to generate a visual user portrait to display the traffic characteristics as visual labels.

6. A user data profile generation system based on traffic, characterized in that: include: A division module is used to obtain user basic data and user flow data, and divide the user flow data according to the feature classification corresponding to the user basic data to obtain feature flow data; A matching module, configured to analyze user behavior characteristics through the user basic data, and match the characteristic traffic data with a preset behavior traffic threshold corresponding to the user behavior characteristics, wherein the user behavior characteristics include access behavior data, download behavior data, transaction behavior data, and input behavior data; An updating module, configured to update the user classification label weight according to the difference between the characteristic traffic data and the preset behavior traffic threshold if the characteristic traffic data is greater than or equal to the preset behavior traffic threshold, so as to retrain the user portrait recognition model; An identification module, used to identify the user basic data based on the trained user portrait identification model, generate a user data portrait identification result, and generate a visual user portrait according to the user data portrait identification result; The division module comprises: A first parsing unit, configured to parse the data tags of the basic user data, and retrieve feature classifications from a feature classification library according to the data tags, wherein the feature classification library stores different feature classifications corresponding to different data tags, and at least one data tag matches one feature classification; A second parsing unit, configured to parse traffic statistical features of the user traffic data, wherein the traffic statistical features include at least one of a time statistical feature, a network statistical feature, an application statistical feature, and a device statistical feature; A calculation unit, configured to perform overlapping calculations on the traffic statistical features according to the feature classification to determine feature traffic data; The matching module comprises: A first determining unit, used to determine a preset access flow threshold, a preset download flow threshold, a preset transaction flow threshold, and a preset entry flow threshold corresponding to the access behavior data, the download behavior data, the transaction behavior data, and the entry behavior data; A comparison unit, used to compare the characteristic traffic data with a preset access traffic threshold, a preset download traffic threshold, a preset transaction traffic threshold, and a preset input traffic threshold respectively; A second determination unit is used to determine a preset access flow threshold matching the access behavior data according to the time node, login address information, and webpage service information; A third determining unit, configured to determine a preset download flow threshold matching the download behavior data according to webpage service information, download object information, and network connection information; A fourth determining unit, configured to determine a preset transaction flow threshold matching the transaction behavior data according to the transaction object information, the transaction party information, and the transaction volume information; The fifth determining unit is used to determine a preset input flow threshold matching the input behavior data according to the webpage business information, the input object information, and the network connection information.

7. A storage medium, wherein at least one executable instruction is stored in the storage medium, and the executable instruction enables a processor to perform operations corresponding to the traffic-based user data portrait generation method as described in any one of claims 1-5.

8. A terminal, comprising: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the traffic-based user data portrait generation method as described in any one of claims 1-5.

Images