Abnormal operation identification methods and related equipment
By standardizing and extracting features from the operational data of the target object, and using machine learning and neural network models, abnormal operations in internet advertising are identified, solving the problem of low identification accuracy in existing technologies and achieving more efficient identification and type discrimination of abnormal operations.
Patent Information
- Application Number
- CN202111086570.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-09-16
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2041-09-16
AI Technical Summary
In the current technology, the accuracy of abnormal operation identification in the Internet advertising industry is low, making it difficult to effectively identify fraudulent behavior and resulting in damage to advertisers' paid benefits.
By acquiring the operation data of the target object within a preset time period, performing data standardization and feature extraction, constructing operation sequence features using machine learning technology, and combining them with a neural network model to identify abnormal operations.
It improves the accuracy of abnormal operation identification, enabling more precise identification and differentiation of different types of abnormal operations, thus protecting the interests of advertisers and the advertising ecosystem.
Smart Images

Figure CN115829648B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of computer technology, specifically to an abnormal operation identification method and related equipment. Background Technology
[0002] With the rapid development of internet technology, information promotion via the internet has gradually become a new form of information dissemination. Internet advertising, as a typical technology for information promotion on the internet, has become an important channel for users to obtain information and for advertisers to enhance brand awareness and promote their products.
[0003] In the internet advertising industry, traffic owners generate revenue by driving desired conversions for advertisers through user clicks, impressions, downloads, and activations. However, some traffic owners resort to fraudulent methods to increase ad clicks and thus generate fake ad traffic in order to boost their earnings. This fraudulent traffic is worthless to advertisers, and since advertisers pay based on user responses to ads, this would only increase their costs. Therefore, advertisers want their paid ad responses to be genuine and valid, not fraudulent. Identifying abnormal ad activity remains a crucial issue in the internet advertising industry.
[0004] Current technologies typically identify abnormal activity by using outcome-based advertising metrics (such as clicks, click-through rate, and conversion rate) at the user or traffic owner level. However, this method can only identify users with extremely abnormal click volumes, resulting in low accuracy in identifying abnormal activity. Summary of the Invention
[0005] This application provides an abnormal operation identification method and related equipment. The related equipment may include an abnormal operation identification device, electronic equipment, computer-readable storage medium, and computer program product, which can improve the accuracy of abnormal operation identification.
[0006] This application provides an abnormal operation identification method, including:
[0007] Acquire the target object's operation data on promotional content within a preset time period, the operation data including operation records in at least one dimension;
[0008] The operation records for each dimension are standardized to obtain an operation data sequence, which includes the standardized operation records for each dimension.
[0009] Feature extraction is performed on the standard processed operation records for each dimension to obtain the operation sequence features of the target object in relation to the promotional content;
[0010] Based on the operation sequence characteristics, abnormal operations are identified in the target object.
[0011] Accordingly, embodiments of this application provide an abnormal operation identification device, including:
[0012] The acquisition unit is used to acquire the operation data of the target object on the promotional content within a preset time period, and the operation data includes operation records of at least one dimension.
[0013] A standardization unit is used to perform data standardization processing on the operation records of each dimension to obtain an operation data sequence, wherein the operation data sequence includes the standardized operation records of each dimension.
[0014] An extraction unit is used to extract features from the standard processed operation records in each dimension to obtain the operation sequence features of the target object in relation to the promotion content.
[0015] The identification unit is used to identify abnormal operations of the target object based on the operation sequence characteristics.
[0016] Optionally, in some embodiments of this application, the operation record includes at least one operation feature record, and the operation feature record includes at least one operation feature;
[0017] The standardized unit may include identification sub-units, merging sub-units, and constructing sub-units, as follows:
[0018] The identification subunit is used to identify the feature type of the corresponding operational feature in each dimension for each dimension;
[0019] The merging subunit is used to merge the operation feature records in each dimension according to the feature type of the operation feature, so as to obtain the standard processed operation record in each dimension.
[0020] Construct sub-units to build operation data sequences based on the standard-processed operation records in each dimension.
[0021] Optionally, in some embodiments of this application, the merging subunit may be specifically used to perform statistical analysis on operation feature records with the same operation features in each dimension according to the feature type of the operation feature, and obtain statistical analysis results; and to merge operation feature records with the same operation features in each dimension according to the statistical analysis results, and obtain standard processed operation records in each dimension.
[0022] Optionally, in some embodiments of this application, the step "based on the feature type of the operation feature, perform statistical analysis on the operation feature records with the same operation feature in each dimension to obtain statistical analysis results" may include:
[0023] Based on the feature type of the operation feature, frequency analysis is performed on the operation feature records with the same operation feature in each dimension to obtain frequency information.
[0024] Optionally, in some embodiments of this application, the step "based on the feature type of the operation feature, perform frequency analysis on the operation feature records with the same operation feature in each dimension to obtain frequency information" may include:
[0025] Based on the feature type of the operation feature, the frequency of operation feature records with the same operation feature in each dimension is calculated to obtain the operation frequency;
[0026] The frequency of the operation is compared with a preset frequency threshold to obtain the frequency information after comparison.
[0027] Optionally, in some embodiments of this application, the extraction unit may include an extraction subunit and a determination subunit, as follows:
[0028] The extraction subunit is used to extract features from the standard processed operation records in each dimension to obtain the target feature information of the standard processed operation records in each dimension.
[0029] A subunit is defined to determine the operation sequence characteristics of the target object for the promotional content based on the target feature information of the operation record after standard processing on each dimension.
[0030] Optionally, in some embodiments of this application, the extraction subunit may be specifically used to predict reference feature information corresponding to the standard processed operation record in each reference dimension based on the standard processed operation record in each dimension, wherein the standard processed operation record in the reference dimension is the standard processed operation record in other dimensions besides the dimension in the operation data sequence; and the reference feature information corresponding to the standard processed operation record in each reference dimension is fused to obtain the target feature information of the standard processed operation record in the dimension.
[0031] Optionally, in some embodiments of this application, the determining subunit may be specifically used to determine the weight corresponding to the standard processed operation record in each dimension; based on the weight, the target feature information of the standard processed operation record in each dimension is fused to obtain the operation sequence feature of the target object for the promotion content.
[0032] Optionally, in some embodiments of this application, the identification unit includes a prediction subunit and an anomaly identification subunit, as follows:
[0033] The prediction subunit is used to predict the probability that the target object's operation on the promotional content belongs to each preset abnormal operation type based on the operation sequence characteristics.
[0034] An anomaly identification subunit is used to determine the target anomaly operation type of the target object for the promotional content from various preset anomaly operation types based on the probability.
[0035] An electronic device provided in this application includes a processor and a memory. The memory stores multiple instructions, and the processor loads the instructions to execute the steps in the abnormal operation identification method provided in this application.
[0036] This application also provides a computer-readable storage medium storing a computer program thereon, wherein the computer program, when executed by a processor, implements the steps in the abnormal operation identification method provided in this application.
[0037] Furthermore, embodiments of this application also provide a computer program product, including a computer program or instructions, which, when executed by a processor, implement the steps in the abnormal operation identification method provided in embodiments of this application.
[0038] This application provides an abnormal operation identification method and related equipment. It can acquire operation data of a target object regarding promotional content within a preset time period. The operation data includes operation records in at least one dimension. Data standardization processing is performed on the operation records in each dimension to obtain an operation data sequence, which includes standardized operation records for each dimension. Feature extraction is performed on the standardized operation records for each dimension to obtain the operation sequence features of the target object regarding the promotional content. Based on the operation sequence features, abnormal operations of the target object are identified. This application improves the accuracy of abnormal operation identification by processing the operation records of the target object regarding promotional content in at least one dimension. Attached Figure Description
[0039] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0040] Figure 1a This is a schematic diagram of a scenario for the abnormal operation identification method provided in the embodiments of this application;
[0041] Figure 1b This is a flowchart of the abnormal operation identification method provided in the embodiments of this application;
[0042] Figure 1c This is a schematic diagram of the page for the abnormal operation identification method provided in the embodiments of this application;
[0043] Figure 1d This is an explanatory diagram of the abnormal operation identification method provided in the embodiments of this application;
[0044] Figure 1e This is another illustrative diagram of the abnormal operation identification method provided in the embodiments of this application;
[0045] Figure 1f This is a model architecture diagram of the abnormal operation identification method provided in the embodiments of this application;
[0046] Figure 1g This is another illustrative diagram of the abnormal operation identification method provided in the embodiments of this application;
[0047] Figure 1h This is another scenario illustration of the abnormal operation identification method provided in the embodiments of this application;
[0048] Figure 1i This is another flowchart of the abnormal operation identification method provided in the embodiments of this application;
[0049] Figure 2 This is another flowchart of the abnormal operation identification method provided in the embodiments of this application;
[0050] Figure 3 This is a schematic diagram of the abnormal operation identification device provided in the embodiments of this application;
[0051] Figure 4 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation
[0052] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0053] This application provides an abnormal operation identification method and related equipment. The related equipment may include an abnormal operation identification device, an electronic device, a computer-readable storage medium, and a computer program product. Specifically, the abnormal operation identification device may be integrated into an electronic device, which may be a terminal or a server, etc.
[0054] It is understood that the abnormal operation identification method of this embodiment can be executed on a terminal, on a server, or jointly by a terminal and a server. The above examples should not be construed as limiting this application.
[0055] like Figure 1a As shown, an abnormal operation identification method jointly executed by a terminal and a server is used as an example. The abnormal operation identification system provided in this application includes a terminal 10 and a server 11, etc.; the terminal 10 and the server 11 are connected through a network, such as through a wired or wireless network, etc., wherein the abnormal operation identification device can be integrated into the server.
[0056] Server 11 can be used to: acquire operation data of a target object on promotional content within a preset time period, the operation data including operation records in at least one dimension; perform data standardization processing on the operation records in each dimension to obtain an operation data sequence, the operation data sequence including standardized operation records in each dimension; extract features from the standardized operation records in each dimension to obtain operation sequence features of the target object on the promotional content; and identify abnormal operations of the target object based on the operation sequence features. Server 11 can be a single server, a server cluster consisting of multiple servers, or a cloud server.
[0057] Terminal 10 can be used to receive abnormal operation identification results for a target object sent by server 11. Terminal 10 may include mobile phones, smart TVs, tablets, laptops, or personal computers (PCs). A client may also be set on terminal 10, which may be an application client or a browser client, etc.
[0058] The steps for identifying abnormal operations by the server 11 can also be performed by the terminal 10.
[0059] The abnormal operation identification method provided in this application relates to machine learning in the field of artificial intelligence. This application embodiment can identify whether the target object's operations on promotional content are abnormal by processing operation records of the target object in at least one dimension, thereby improving the accuracy of abnormal operation identification.
[0060] Artificial intelligence (AI) is the theory, methods, technology, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a comprehensive technology within computer science that attempts to understand the essence of intelligence and produce a new kind of intelligent machine that can react in a way similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to have perception, reasoning, and decision-making capabilities. AI technology is a comprehensive discipline involving a wide range of fields, encompassing both hardware and software technologies. AI software technology mainly includes computer vision, speech processing, natural language processing, as well as machine learning / deep learning, autonomous driving, and intelligent transportation.
[0061] Machine learning (ML) is a multidisciplinary field involving probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory, among others. It specifically studies how computers can simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to endow computers with intelligence; its applications span all areas of artificial intelligence. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and instruction-based learning.
[0062] The following sections provide detailed descriptions of each example. It should be noted that the order in which the embodiments are described is not intended to limit the preferred order of the embodiments.
[0063] This embodiment will be described from the perspective of an abnormal operation identification device, which can be integrated into an electronic device, such as a server or a terminal.
[0064] The abnormal operation identification method of this application embodiment can be applied to various abnormal operation identification scenarios, such as the scenario of abnormal operation detection in advertising.
[0065] like Figure 1b As shown, the specific process of this abnormal operation identification method can be as follows:
[0066] 101. Obtain the target object's operation data on the promotional content within a preset time period, wherein the operation data includes operation records of at least one dimension.
[0067] Specifically, the target object is the user who needs to be identified for abnormal operations. The abnormal operation identification method provided in this application can determine whether the target object's operation on the promotional content is abnormal.
[0068] The promotional content can be pushed to users via the Internet in various formats (such as text, images, audio, video, etc.). Specifically, the promotional content can be advertisements, public service announcements, etc.
[0069] Operational data can refer to actions performed on promoted content, specifically data generated by advertising operations. Advertising operations can include the entire advertising-related data and actions, such as entering a traffic source, exposing an ad, clicking an ad, opening an ad (landing page), opening an external link, and taking a screenshot. Operational data can include information about the target audience's operational scenarios, types, and times related to the promoted content. Operational data can contain at least one dimension of operation records, and each dimension's operation records can include at least one operation feature record. An operation feature record can specifically be a record of relevant operation information corresponding to an action at a specific time. For example, if a user clicked on promoted content through a WeChat official account at 6 PM yesterday, this can be considered an operation feature record, containing the operation scenario: clicked through a WeChat official account, and the operation time: 6 PM yesterday. For each dimension's operation records, the various operation feature records within that dimension have a certain correlation. For example, operation feature records with the same operation time, the same operation scenario, and the same operation type can be considered as operation records belonging to the same dimension.
[0070] The preset time period can be set according to the actual situation, and this embodiment does not impose any restrictions on it. For example, the preset time period can be the most recent week or the most recent month.
[0071] In a scenario, such as Figure 1c As shown, WeChat Official Account and Mini Program developers can apply to open advertising slots and become traffic owners. When users click on ads in WeChat Official Account articles or Mini Programs, traffic owners can receive a share of the revenue. Generally, the higher the ad exposure and click volume under a traffic owner's account, the more revenue they receive. Therefore, some traffic owners resort to cheating to create invalid exposures to increase advertising revenue, but this results in very poor advertising performance, harming the interests of advertisers (ad placement parties). Cheating seriously affects advertising effectiveness and damages the overall advertising ecosystem. Therefore, it is necessary to identify abnormal advertising operations. This application embodiment can identify whether there are abnormal operations by processing the target object's operation records for at least one dimension of the promotional content, thereby improving the accuracy of abnormal operation identification.
[0072] 102. Perform data standardization processing on the operation records of each dimension to obtain an operation data sequence, wherein the operation data sequence includes the standardized operation records of each dimension.
[0073] Specifically, data standardization processing can involve statistical analysis of operation records for each dimension based on the operation features contained in the operation feature records. This embodiment does not impose any restrictions on this and can be set according to the actual situation.
[0074] Optionally, in this embodiment, the operation record includes at least one operation feature record, and the operation feature record includes at least one operation feature;
[0075] The step "standardize the operation records for each dimension to obtain the operation data sequence" may include:
[0076] For each dimension, identify the feature type of the corresponding operational feature in each dimension;
[0077] Based on the feature type of the operation feature, the operation feature records in each dimension are merged to obtain the standard processed operation record in each dimension;
[0078] Based on the standard processed operation records in each dimension, construct the operation data sequence.
[0079] Specifically, the operation feature record can be a record of relevant operation information corresponding to an operation action at a certain time. Each operation feature record may include one or more operation features, which may specifically include operation scenario, operation type, operation time, etc., and this embodiment does not limit this. Identifying the feature type of the operation feature corresponding to each dimension can specifically involve identifying whether the operation feature corresponding to each dimension belongs to the operation scenario, the operation type, or the operation time, etc.
[0080] Optionally, in this embodiment, the step "merging the operation feature records in each dimension according to the feature type of the operation feature to obtain the standard processed operation record in each dimension" may include:
[0081] Based on the feature type of the operation feature, statistical analysis is performed on the operation feature records with the same operation feature in each dimension to obtain the statistical analysis results;
[0082] Based on the statistical analysis results, operation feature records with the same operation characteristics in each dimension are merged to obtain the standard processed operation records in each dimension.
[0083] Specifically, statistical analysis is performed on the operation feature records with the same operation features in each dimension. This can involve counting the number of operation feature records with the same operation features in each dimension. The statistical analysis results can include the frequency of occurrence of operation feature records with the same operation features in each dimension, which is denoted as operation frequency.
[0084] The merged standard operation records can include operation frequency characteristics, specifically the number of operation feature records merged in that dimension. It can be understood that the merged standard operation records can include all operation features from the original operation feature records. After merging, standard operation records are obtained for each dimension. These records can be sorted chronologically, and then an operation data sequence is constructed based on the sorted records. The resulting operation data sequence includes the chronologically ordered standard operation records for each dimension.
[0085] Optionally, the step "merging the operation feature records in each dimension according to the feature type of the operation feature to obtain the standard processed operation record in each dimension" can specifically be:
[0086] Operation feature records with the same operation scenario, operation type, and operation time within the same time period are merged to obtain merged operation feature records, which are the standard processed operation records in this dimension.
[0087] Among them, the operation scenario can be an operation scenario for the promoted content, such as an advertising operation scenario; each advertising operation will have corresponding scenario data. For example, under the public account, there are operation scenarios such as entering advertising content through a subscription account or entering advertising content through a group chat. Under the mini program, there are operation scenarios such as opening advertising content through recently used apps, invoking advertising content through an external application (app, Application), and opening advertising content by scanning a QR code.
[0088] Specifically, the operation type can be an advertising operation type, which can include opening a traffic owner, exposing an ad, clicking an ad, taking a screenshot of an ad, or accessing an external link.
[0089] Optionally, in this embodiment, the step "based on the feature type of the operation feature, perform statistical analysis on the operation feature records with the same operation feature in each dimension to obtain the statistical analysis results" may include:
[0090] Based on the feature type of the operation feature, frequency analysis is performed on the operation feature records with the same operation feature in each dimension to obtain frequency information.
[0091] The frequency information can specifically be the operation frequency described in the above embodiments. In some embodiments, since there are many frequency values, the calculated operation frequency can be segmented. Specifically, based on the preset frequency threshold corresponding to the preset frequency band, it is determined which preset frequency band the currently calculated operation frequency belongs to, and thus the preset frequency band is used as the target frequency information.
[0092] Optionally, in this embodiment, the step "based on the feature type of the operation feature, perform frequency analysis on the operation feature records with the same operation feature in each dimension to obtain frequency information" may include:
[0093] Based on the feature type of the operation feature, the frequency of operation feature records with the same operation feature in each dimension is calculated to obtain the operation frequency;
[0094] The frequency of the operation is compared with a preset frequency threshold to obtain the frequency information after comparison.
[0095] The preset frequency threshold can be set according to actual conditions, and this embodiment does not impose any restrictions on it. Specifically, the preset frequency threshold can be an endpoint value of a preset frequency band. For example, the preset frequency threshold can be set to (0, 1, 3, 5, 10, 20, 50, 100, 500, 1000). By comparing it with the preset frequency threshold, it can be determined which frequency band the operation frequency belongs to, and the target frequency information, i.e., the frequency information after comparison, can be obtained.
[0096] In this embodiment, the anomaly of a user's current advertising operation can be determined based on the differences between normal and abnormal advertising operations (i.e., fraudulent advertising behavior). The advertising operations of normal users and abnormal users have distinct operational characteristics, such as:
[0097] In terms of user experience, abnormal users typically access advertising content through methods such as scanning a QR code or being launched by an external app, while normal users typically access it through methods such as subscribed accounts or recently used apps.
[0098] Regarding the type of operation, abnormal users typically take screenshots of advertisements (abnormal users retain evidence of clicking on advertisements in exchange for third-party rewards) and access suspicious external links; while normal users usually do not have these suspicious operations.
[0099] Regarding the frequency of operations, normal users operate ads less frequently, while abnormal users typically have high-frequency exposure and clicks.
[0100] Regarding the timing of actions, normal users' ad actions are usually scattered, while abnormal users' ad actions tend to be more concentrated.
[0101] Abnormal users often engage in unusual advertising activities driven by profit. These activities accumulate into action sequences, which differ from those of normal users. Therefore, action feature records can be constructed to represent advertising actions and aggregated into user action data sequences for the identification and detection of advertising anomalies.
[0102] Specifically, ad anomalies refer to instances where users, with malicious intent, manipulate ad impressions, clicks, or performance to artificially inflate ad impressions, clicks, or performance. Such malicious actions, not driven by the user's true intentions, are termed ad anomalies.
[0103] In one specific embodiment, user operation data for promotional content over a period of time (e.g., the last 7 days) can first be extracted from the advertising operation database. This operation data includes operation records of at least one dimension, and each dimension of operation record includes at least one operation feature record. Specifically, the operation data can also be regarded as a sequence of operations for promotional content at the user level, and the operation feature records in the operation sequence can be sorted by time.
[0104] For example, a user's sequence of actions towards promoted content within a certain time period can be as follows:
[0105] (Open Traffic Ownership, Scenario 1, 6th hour)
[0106] (Open Traffic Ownership, Scenario 2, 6th hour)
[0107] (Exposure, Scene 3, 7th Hour)
[0108] (Exposure, Scene 3, 7th Hour)
[0109] (Click, Scene 3, 7th hour)
[0110] The operation sequence contains five operation feature records, each containing operation features—operation type, operation scenario, and operation time. After obtaining the user's operation data (i.e., the operation sequence), the user's operation sequence can be standardized to construct an operation data sequence.
[0111] For the original operation feature records (i.e., operation feature records before data standardization), operation feature records with the same operation type, the same operation scenario, and the same operation data at the same time can be regarded as operation records in the same dimension. The number of operation feature records in the operation records in this dimension is counted, that is, the frequency of occurrence. Then, the operation feature records in the operation records in this dimension are merged to obtain the standardized operation records in this dimension, and then the operation data sequence is constructed, as shown below:
[0112] (Open traffic monetization, scenario 1, number 1, 6th hour)
[0113] (Open traffic monetization, scenario 2, number 1, 6th hour)
[0114] (Exposure, Scene 3, Number of exposures, 7th hour)
[0115] (Click, Scene 3, 1 time, 7th hour)
[0116] Among them, the third and fourth operation feature records in the original operation sequence have the same operation type (both are "exposure"), the same operation scenario (both are "scenario 3"), and the same operation time (both are "7th hour"). They are operation records in the same dimension and can be merged to obtain the standard processed operation record in this dimension: (exposure, scenario 3, number of times 2, 7th hour).
[0117] A single standard processed operation record can be represented in the form of "operation type_operation scenario_operation frequency_operation time".
[0118] Thus, through the above processing, an operation data sequence will be obtained for each user. This operation data sequence includes standard post-processing operation records in at least one dimension. Specifically, the operation data sequence can be denoted as (operation 1, operation 2, ..., operation n), where operation k represents the standard post-processing operation record in each dimension, and the value of k ranges from 1 to n.
[0119] There is a clear distinction between the action data sequences of normal users and users with different abnormal types regarding promotional content. For example... Figure 1d and Figure 1e The images show visualizations of the action data sequences of normal users and various abnormal user types regarding promotional content. Specifically, Figure 1d (a) in the table represents the operation data sequence of a normal user. Figure 1d (b) in the table represents the sequence of operation data of the machine-cheating user. Figure 1e (c) in the table represents the operation data sequence of the user who is AFK (away from keyboard) cheating. Figure 1eIn the diagram, (d) represents the operation data sequence of a user engaging in online money-making fraud. Each operation in the diagram represents a user's operation data sequence. Different colors (grayscale) represent different operation types or scenarios, the length of the grayscale bar represents the operation frequency (the longer the length, the higher the frequency), and the order represents the chronological order of the operations. It can be observed that the operation data sequences of normal users are relatively rich and lack obvious patterns; while the operation data sequences of abnormal users show obvious clustering and regularity, and different types of abnormal users exhibit different patterns. This embodiment can utilize the characteristic differences between the operation data sequences of normal and abnormal users to identify whether a target object has abnormal operations, and can also determine the type of abnormality.
[0120] 103. Extract features from the standard processed operation records for each dimension to obtain the operation sequence features of the target object for the promotion content.
[0121] In this process, feature extraction is performed on the standard processed operation records for each dimension, which may involve convolution, pooling, or other similar processing.
[0122] Specifically, features can be extracted from the standard processed operation records in each dimension using a neural network model. This neural network model can be a Skip-gram model (Continuous Skip-gram Model), a Convolutional Neural Network (CNN), or a Residual Network (ResNet), etc. However, it should be understood that the neural network in this embodiment is not limited to the types listed above.
[0123] Among them, the Skip-gram model is widely used in natural language processing. It can consist of three layers: an input layer, a projection layer, and an output layer; it can be used to process words given the current word W. t Under the premise of predicting the current word W t Context W t-2 W t-1 W t+1 W t+2 This embodiment treats the standardized processed operation record in each dimension as a "word," and a sequence of operation data as a "sentence." For example... Figure 1f The diagram illustrates the specific architecture of the Skip-gram model. Its main idea is to predict surrounding advertising actions (standardized operation records in other dimensions) based on the current advertising action (i.e., the standard processed operation record in the current dimension). Specifically, the operation data sequence is denoted as W.N*M The operation data sequence includes standard processed operation records in N dimensions, each with a dimension of M. The standard processed operation record in the current dimension can be denoted as w. i The operation data sequence is input into the input layer, and features are extracted through the projection layer, such as the standard processed operation record w in the current dimension. i It can predict the reference feature information corresponding to the standard processed operation record in other dimensions, and then obtain the target feature information W of the standard processed operation record in the current dimension. M ′ *K This yields target feature information W, recorded from standard processed operational features across N dimensions. M ′ *K The characteristics of the operation sequence constituted.
[0124] Optionally, in this embodiment, the step "extracting features from the standard processed operation records for each dimension to obtain the operation sequence features of the target object for the promotional content" may include:
[0125] Feature extraction is performed on the standard processed operation records in each dimension to obtain the target feature information of the standard processed operation records in each dimension;
[0126] Based on the target feature information of the standard processed operation records in each dimension, the operation sequence features of the target object for the promotion content are determined.
[0127] In some embodiments, the target feature information of the standard processed operation records in each dimension can be concatenated to obtain operation sequence features. The concatenation order can be determined based on the operation time in the standard processed operation records of each dimension. In other embodiments, operation sequence features can also be obtained by taking the maximum value or average value of the standard processed operation records in each dimension.
[0128] Optionally, in this embodiment, the step "extracting features from the standard processed operation records in each dimension to obtain the target feature information of the standard processed operation records in each dimension" may include:
[0129] For each dimension of the standard processed operation record, based on the standard processed operation record of the dimension, predict the reference feature information corresponding to the standard processed operation record of each reference dimension. The standard processed operation record of the reference dimension is the standard processed operation record of other dimensions in the operation data sequence besides the dimension.
[0130] The reference feature information corresponding to the standard processed operation record on each reference dimension is fused to obtain the target feature information of the standard processed operation record on the dimension.
[0131] There are various ways to fuse the reference feature information, and this embodiment does not limit this one. For example, the fusion method can be weighted fusion or splicing processing, etc.
[0132] Optionally, in this embodiment, the step "determining the operation sequence characteristics of the target object for the promotional content based on the target feature information of the standard-processed operation records in each dimension" may include:
[0133] Determine the weights corresponding to the standard post-processing operation records for each dimension;
[0134] Based on the weights, the target feature information of the standard processed operation records in each dimension is fused to obtain the operation sequence features of the target object for the promotional content.
[0135] 104. Based on the operation sequence characteristics, identify abnormal operations on the target object.
[0136] Optionally, in this embodiment, the step "extracting features from the standard processed operation records for each dimension to obtain the operation sequence features of the target object for the promotional content" may include:
[0137] By using the operation recognition model, feature extraction is performed on the standard processed operation records in each dimension to obtain the operation sequence features of the target object in response to the promotion content.
[0138] The step "identifying abnormal operations of the target object based on the operation sequence characteristics" may include:
[0139] By using an operation recognition model, abnormal operations are identified in the target object based on the characteristics of the operation sequence.
[0140] The operation recognition model can include neural network models built based on neural network algorithms and tree models built based on decision tree algorithms, etc., and this embodiment does not limit this. The type of neural network is not limited; for example, it can be a convolutional neural network, a residual network (ResNet), or a densely connected convolutional network (DenseNet), etc. The tree model can be an X-gboost model, etc. X-gboost (eXtreme Gradient Boosting) is an optimized distributed gradient boosting library, derived from the gradient boosting framework, but more efficient because the X-gboost algorithm can perform parallel computation, approximate tree construction, effectively handle sparse data, and optimize memory usage.
[0141] It should be noted that the operation recognition model can be trained by other devices and then provided to the abnormal operation recognition device, or it can be trained by the abnormal operation recognition device itself.
[0142] If the abnormal operation identification device is trained automatically, then before the step "using the operation identification model to extract features from the standard processed operation records in each dimension to obtain the operation sequence features of the target object for the promotional content", the abnormal operation identification device may further include:
[0143] Acquire training data, which includes sample operation data of sample objects for promotional content and expected anomaly types corresponding to the sample operation data. The sample operation data includes operation records after standard processing of samples in at least one dimension.
[0144] By using a preset operation recognition model, feature extraction is performed on the operation records after standard processing of samples in each dimension to obtain the sample operation sequence features of the sample object in relation to the promotion content.
[0145] Based on the characteristics of the sample operation sequence, the actual anomaly type of the sample object is identified;
[0146] Based on the actual and expected anomaly types of the sample objects, the parameters of the preset operation recognition model are adjusted to obtain the trained operation recognition model.
[0147] Specifically, the training process involves adjusting the parameters of a pre-defined operation recognition model using a backpropagation algorithm. Based on the actual and expected anomaly types of the sample objects, the parameters are optimized to ensure that the predicted actual anomaly type matches the expected anomaly type, resulting in a well-trained operation recognition model. Specifically, the probability that the predicted operation of the sample object on the promotional content belongs to the expected anomaly type can be higher than a preset value, which can be set according to actual conditions. After training, the operation recognition model can be applied online to identify abnormal operations in real time, enabling more precise anomaly detection and mitigation, thus protecting advertisers' interests and the advertising ecosystem.
[0148] Optionally, in this embodiment, the step "identifying abnormal operations of the target object based on the operation sequence characteristics" may include:
[0149] Based on the operation sequence characteristics, predict the probability that the target object's operation on the promotional content belongs to each preset abnormal operation type;
[0150] Based on the probability, the target abnormal operation type of the target object for the promotional content is determined from each preset abnormal operation type.
[0151] Specifically, a classifier can be used to predict the probability that the target object's actions towards the promotional content belong to various preset abnormal operation types. This classifier can be a Support Vector Machine (SVM), a Recurrent Neural Network (RNN), a Fully Connected Deep Neural Network (DNN), etc., and this embodiment does not impose any limitations on it.
[0152] In some embodiments, the preset abnormal operation type with the highest probability can be determined as the target abnormal operation type of the target object for the promotional content. In other embodiments, the preset abnormal operation type with a probability greater than a preset value can be used as the target abnormal operation type of the target object for the promotional content, and this preset value can be set according to the actual situation.
[0153] This application designs a method for learning dense feature vectors of advertising operations, which obtains the data sequence of the object's operations in response to the advertisement. This data can be input into a supervised model (such as the operation recognition model in the above embodiment) for abnormal operation recognition. It can not only determine whether the object has abnormal advertising operations, but also identify the type of advertising abnormality.
[0154] Optionally, in this embodiment, before the step of "identifying abnormal operations of the target object based on the operation sequence features", the operation sequence features can be reduced in dimensionality. Specifically, the operation sequence features are 64-dimensional, and a two-dimensional representation can be obtained through feature dimensionality reduction. This feature dimensionality reduction algorithm can be t-sne (t-DistributedStochastic Neighbor Embedding) algorithm, etc., and this embodiment does not limit it.
[0155] Among them, t-sne is a dimensionality reduction technique used to represent high-dimensional datasets in a low-dimensional space of two or three dimensions, thereby making them visible.
[0156] like Figure 1g The figure shows the distribution of feature vectors corresponding to the operation sequence features of users with different abnormal types and normal users (non-abnormal types). It can be observed that the operation sequence features of normal users and abnormal users are significantly distinguishable, and the operation sequence features of abnormal users show obvious clustering. Specifically, malicious users who leave their accounts running, malicious users with Trojan viruses, and users who cheat in online earning schemes show significant clustering, while the distribution of operation sequence features of normal users (randomly selected from the general market) regarding promotional content is relatively dispersed (or, in other words, normal users' operations are more varied) and differ significantly from those of other abnormal users. Therefore, an operation recognition model can be used to identify abnormal operations of a target object based on its operation sequence features.
[0157] This embodiment can detect ad anomalies based on users' multi-dimensional operation records of ads. It comprehensively uses factors such as the user's current and historical ad operation type, operation scenario, operation frequency, and operation time to design a feature generation method for ad operations. This method can effectively distinguish between various types of abnormal and normal ad operations. It makes full use of multi-dimensional information at each stage of the ad and can determine whether there are abnormal ad operations at a fine-grained level (down to a single exposure click).
[0158] In a specific scenario, such as Figure 1hAs shown, when a user performs ad-related actions through the client (such as opening a traffic monetization tool, exposing an ad, clicking an ad, taking a screenshot, or opening an external link), the client reports the ad operation data (including but not limited to operation type, operation scenario, and operation time) to the access layer server. The access layer server can then store the user's ad operation data in the ad operation database. Subsequently, if it is necessary to identify abnormal ad operations by the target user, a real-time request can be sent to the real-time computing server to request the server to detect whether the target object has any abnormal ad operations. At this time, the real-time computing server can obtain the target object's historical and current ad operation data from the ad operation database, identify abnormal operations based on the target object's historical and current ad operation data, and return the judgment result.
[0159] In one specific embodiment, the anomaly detection process based on the operational data sequence of promotional content provided in this application is as follows: Figure 1i As shown below:
[0160] Step 1: Extract the target audience's action data on the promotional content within a preset time period. This action data includes action records in at least one dimension.
[0161] Step 2: Standardize the operation records in each dimension of the operation data. Specifically, each operation record in each dimension includes at least one operation feature record. Based on the feature type of the operation feature in the operation feature record, calculate the frequency of operation feature records with the same operation feature in each dimension to obtain frequency information. Based on the frequency information, merge the operation feature records with the same operation feature in each dimension to obtain the standardized operation record in each dimension, and then construct the operation data sequence.
[0162] Step 3: Using the continuous jump character model, extract features from the standard processed operation records (specifically, advertising operations) in each dimension to obtain the target feature information of the standard processed operation records (specifically, advertising operations) in each dimension.
[0163] Step 4: Perform calculations on the target feature information of the standard processed operation records (specifically, advertising operations) for each dimension (such as taking the maximum value or the average value) to obtain the operation sequence features;
[0164] Step 5: Input the operation sequence features into the supervised model to identify abnormal operations of the target object.
[0165] As can be seen from the above, this embodiment can acquire the operation data of the target object regarding the promotional content within a preset time period. The operation data includes operation records in at least one dimension. Data standardization processing is performed on the operation records in each dimension to obtain an operation data sequence, which includes standardized operation records in each dimension. Feature extraction is performed on the standardized operation records in each dimension to obtain the operation sequence features of the target object regarding the promotional content. Based on the operation sequence features, abnormal operation identification is performed on the target object. This embodiment can identify whether there are abnormal operations in the target object's operations regarding the promotional content by processing the operation records of the target object in at least one dimension, thus improving the accuracy of abnormal operation identification.
[0166] Based on the method described in the preceding embodiments, the following will provide a more detailed explanation by taking the specific integration of the abnormal operation identification device into a server as an example.
[0167] This application provides an abnormal operation identification method, such as... Figure 2 As shown, the specific process of this abnormal operation identification method can be as follows:
[0168] 201. The server obtains the operation data of the target object on the promotion content within a preset time period. The operation data includes operation records of at least one dimension. The operation records include at least one operation feature record. The operation feature record includes at least one operation feature.
[0169] The target object is the user who needs to be identified for abnormal operations. The abnormal operation identification method provided in this application can determine whether the target object's operation on the promotional content is abnormal.
[0170] Operational data can refer to actions performed on promoted content, specifically data generated by advertising operations. Advertising operations can include the entire advertising-related data and actions, such as entering a traffic source, exposing an ad, clicking an ad, opening an ad (landing page), opening an external link, and taking a screenshot. Operational data can include information about the target audience's operational scenarios, types, and times related to the promoted content. Operational data can contain at least one dimension of operation records, and each dimension's operation records can include at least one operation feature record. An operation feature record can specifically be a record of relevant operation information corresponding to an action at a specific time. For example, if a user clicked on promoted content through a WeChat official account at 6 PM yesterday, this can be considered an operation feature record, containing the operation scenario: clicked through a WeChat official account, and the operation time: 6 PM yesterday. For each dimension's operation records, the various operation feature records within that dimension have a certain correlation. For example, operation feature records with the same operation time, the same operation scenario, and the same operation type can be considered as operation records belonging to the same dimension.
[0171] Specifically, the target audience can be traffic owners. In the internet advertising industry, traffic owners are the carriers that provide traffic, typically referring to media outlets, websites, or software. On a particular advertising platform, a traffic owner can be a business service account with a certain number of followers; this account can be a public account or a personal account. Traffic owners can participate in advertising revenue sharing. With the same ad exposure, the higher the click-through rate (CTR), the higher the profit they receive. Therefore, traffic owners have a strong incentive to artificially inflate their CTR. Traffic generated through these irregularities is worthless to advertisers. Advertisers are users or service providers who pay to place ads, and they want every paid ad click to be a valid click from a genuine user, not an abnormal click.
[0172] 202. For each dimension, the server identifies the feature type of the corresponding operational feature in each dimension.
[0173] Specifically, the operation feature record can be a record of relevant operation information corresponding to an operation action at a certain time. Each operation feature record may include one or more operation features, which may specifically include operation scenario, operation type, operation time, etc., and this embodiment does not limit this. Identifying the feature type of the operation feature corresponding to each dimension can specifically involve identifying whether the operation feature corresponding to each dimension belongs to the operation scenario, the operation type, or the operation time, etc.
[0174] 203. The server merges the operation feature records in each dimension according to the feature type of the operation feature to obtain an operation data sequence, which includes the standard processed operation records in each dimension.
[0175] Optionally, in this embodiment, the step "merging the operation feature records in each dimension according to the feature type of the operation feature to obtain the standard processed operation record in each dimension" may include:
[0176] Based on the feature type of the operation feature, statistical analysis is performed on the operation feature records with the same operation feature in each dimension to obtain the statistical analysis results;
[0177] Based on the statistical analysis results, operation feature records with the same operation characteristics in each dimension are merged to obtain the standard processed operation records in each dimension.
[0178] Specifically, statistical analysis is performed on the operation feature records with the same operation features in each dimension. This can involve counting the number of operation feature records with the same operation features in each dimension. The statistical analysis results can include the frequency of occurrence of operation feature records with the same operation features in each dimension, which is denoted as operation frequency.
[0179] Optionally, the step "merging the operation feature records in each dimension according to the feature type of the operation feature to obtain the standard processed operation record in each dimension" can specifically be:
[0180] Operation feature records with the same operation scenario, operation type, and operation time within the same time period are merged to obtain merged operation feature records, which are the standard processed operation records in this dimension.
[0181] 204. The server extracts features from the standard processed operation records for each dimension to obtain the operation sequence features of the target object in relation to the promotional content.
[0182] Optionally, in this embodiment, the step "extracting features from the standard processed operation records for each dimension to obtain the operation sequence features of the target object for the promotional content" may include:
[0183] Feature extraction is performed on the standard processed operation records in each dimension to obtain the target feature information of the standard processed operation records in each dimension;
[0184] Based on the target feature information of the standard processed operation records in each dimension, the operation sequence features of the target object for the promotion content are determined.
[0185] In some embodiments, the target feature information of the standard processed operation records in each dimension can be concatenated to obtain operation sequence features. The concatenation order can be determined based on the operation time in the standard processed operation records of each dimension. In other embodiments, operation sequence features can also be obtained by taking the maximum value or average value of the standard processed operation records in each dimension.
[0186] Optionally, in this embodiment, the step "extracting features from the standard processed operation records in each dimension to obtain the target feature information of the standard processed operation records in each dimension" may include:
[0187] For each dimension of the standard processed operation record, based on the standard processed operation record of the dimension, predict the reference feature information corresponding to the standard processed operation record of each reference dimension. The standard processed operation record of the reference dimension is the standard processed operation record of other dimensions in the operation data sequence besides the dimension.
[0188] The reference feature information corresponding to the standard processed operation record on each reference dimension is fused to obtain the target feature information of the standard processed operation record on the dimension.
[0189] There are various ways to fuse the reference feature information, and this embodiment does not limit this one. For example, the fusion method can be weighted fusion or splicing processing, etc.
[0190] Optionally, in this embodiment, the step "determining the operation sequence characteristics of the target object for the promotional content based on the target feature information of the standard-processed operation records in each dimension" may include:
[0191] Determine the weights corresponding to the standard post-processing operation records for each dimension;
[0192] Based on the weights, the target feature information of the standard processed operation records in each dimension is fused to obtain the operation sequence features of the target object for the promotional content.
[0193] 205. Based on the operation sequence characteristics, the server predicts the probability that the target object's operation on the promotional content belongs to each preset abnormal operation type.
[0194] Specifically, a classifier can be used to predict the probability that the target object's actions towards the promotional content belong to various preset abnormal operation types. This classifier can be a support vector machine, a recurrent neural network, a fully connected deep neural network, etc., and this embodiment does not impose any limitations on it.
[0195] 206. Based on the probability, the server determines the target abnormal operation type of the target object for the promotional content from various preset abnormal operation types.
[0196] In some embodiments, the preset abnormal operation type with the highest probability can be determined as the target abnormal operation type of the target object for the promotional content. In other embodiments, the preset abnormal operation type with a probability greater than a preset value can be used as the target abnormal operation type of the target object for the promotional content, and this preset value can be set according to the actual situation.
[0197] As can be seen from the above, this embodiment can obtain the operation data of the target object on the promotional content within a preset time period through a server. The operation data includes operation records of at least one dimension, and the operation records include at least one operation feature record, which includes at least one operation feature. For each dimension, the feature type of the corresponding operation feature on each dimension is identified. According to the feature type of the operation feature, the operation feature records on each dimension are merged to obtain an operation data sequence. The operation data sequence includes standard processed operation records on each dimension. Feature extraction is performed on the standard processed operation records on each dimension to obtain the operation sequence features of the target object on the promotional content. According to the operation sequence features, the probability that the operation of the target object on the promotional content belongs to each preset abnormal operation type is predicted. According to the probability, the target abnormal operation type of the target object on the promotional content is determined from each preset abnormal operation type. This embodiment can identify whether the operation of the target object on the promotional content is abnormal by processing the operation records of the target object on the promotional content in at least one dimension, thereby improving the accuracy of abnormal operation identification.
[0198] To better implement the above methods, embodiments of this application also provide an abnormal operation identification device, such as... Figure 3 As shown, the abnormal operation identification device may include an acquisition unit 301, a standardization unit 302, an extraction unit 303, and an identification unit 304, as follows:
[0199] (1) Obtain unit 301;
[0200] The acquisition unit 301 is used to acquire the operation data of the target object on the promotional content within a preset time period, and the operation data includes operation records of at least one dimension.
[0201] (2) Standardized Unit 302;
[0202] The standardization unit 302 is used to perform data standardization processing on the operation records of each dimension to obtain an operation data sequence, wherein the operation data sequence includes the standardized operation records of each dimension.
[0203] Optionally, in some embodiments of this application, the operation record includes at least one operation feature record, and the operation feature record includes at least one operation feature;
[0204] The standardized unit may include identification sub-units, merging sub-units, and constructing sub-units, as follows:
[0205] The identification subunit is used to identify the feature type of the corresponding operational feature in each dimension for each dimension;
[0206] The merging subunit is used to merge the operation feature records in each dimension according to the feature type of the operation feature, so as to obtain the standard processed operation record in each dimension.
[0207] Construct sub-units to build operation data sequences based on the standard-processed operation records in each dimension.
[0208] Optionally, in some embodiments of this application, the merging subunit may be specifically used to perform statistical analysis on operation feature records with the same operation features in each dimension according to the feature type of the operation feature, and obtain statistical analysis results; and to merge operation feature records with the same operation features in each dimension according to the statistical analysis results, and obtain standard processed operation records in each dimension.
[0209] Optionally, in some embodiments of this application, the step "based on the feature type of the operation feature, perform statistical analysis on the operation feature records with the same operation feature in each dimension to obtain statistical analysis results" may include:
[0210] Based on the feature type of the operation feature, frequency analysis is performed on the operation feature records with the same operation feature in each dimension to obtain frequency information.
[0211] Optionally, in some embodiments of this application, the step "based on the feature type of the operation feature, perform frequency analysis on the operation feature records with the same operation feature in each dimension to obtain frequency information" may include:
[0212] Based on the feature type of the operation feature, the frequency of operation feature records with the same operation feature in each dimension is calculated to obtain the operation frequency;
[0213] The frequency of the operation is compared with a preset frequency threshold to obtain the frequency information after comparison.
[0214] (3) Extraction unit 303;
[0215] Extraction unit 303 is used to extract features from the standard processed operation records in each dimension to obtain the operation sequence features of the target object for the promotion content.
[0216] Optionally, in some embodiments of this application, the extraction unit may include an extraction subunit and a determination subunit, as follows:
[0217] The extraction subunit is used to extract features from the standard processed operation records in each dimension to obtain the target feature information of the standard processed operation records in each dimension.
[0218] A subunit is defined to determine the operation sequence characteristics of the target object for the promotional content based on the target feature information of the operation record after standard processing on each dimension.
[0219] Optionally, in some embodiments of this application, the extraction subunit may be specifically used to predict reference feature information corresponding to the standard processed operation record in each reference dimension based on the standard processed operation record in each dimension, wherein the standard processed operation record in the reference dimension is the standard processed operation record in other dimensions besides the dimension in the operation data sequence; and the reference feature information corresponding to the standard processed operation record in each reference dimension is fused to obtain the target feature information of the standard processed operation record in the dimension.
[0220] Optionally, in some embodiments of this application, the determining subunit may be specifically used to determine the weight corresponding to the standard processed operation record in each dimension; based on the weight, the target feature information of the standard processed operation record in each dimension is fused to obtain the operation sequence feature of the target object for the promotion content.
[0221] (4) Identification unit 304;
[0222] The identification unit 304 is used to identify abnormal operations of the target object based on the operation sequence characteristics.
[0223] Optionally, in some embodiments of this application, the identification unit includes a prediction subunit and an anomaly identification subunit, as follows:
[0224] The prediction subunit is used to predict the probability that the target object's operation on the promotional content belongs to each preset abnormal operation type based on the operation sequence characteristics.
[0225] An anomaly identification subunit is used to determine the target anomaly operation type of the target object for the promotional content from various preset anomaly operation types based on the probability.
[0226] As can be seen from the above, this embodiment can acquire the operation data of the target object on the promotional content within a preset time period through the acquisition unit 301. The operation data includes operation records in at least one dimension. The standardization unit 302 performs data standardization processing on the operation records in each dimension to obtain an operation data sequence. The operation data sequence includes the standardized operation records in each dimension. The extraction unit 303 extracts features from the standardized operation records in each dimension to obtain the operation sequence features of the target object on the promotional content. The identification unit 304 identifies abnormal operations of the target object based on the operation sequence features. This embodiment can identify whether the target object's operations on the promotional content are abnormal by processing the operation records of the target object on at least one dimension, thereby improving the accuracy of abnormal operation identification.
[0227] This application also provides an electronic device, such as... Figure 4 The diagram shows a schematic representation of the structure of an electronic device according to an embodiment of this application. This electronic device can be a terminal or a server, specifically:
[0228] The electronic device may include components such as a processor 401 with one or more processing cores, a memory 402 with one or more computer-readable storage media, a power supply 403, and an input unit 404. Those skilled in the art will understand that... Figure 4 The electronic device structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein:
[0229] The processor 401 is the control center of the electronic device. It connects various parts of the electronic device via various interfaces and lines. By running or executing software programs and / or modules stored in the memory 402, and by calling data stored in the memory 402, it performs various functions and processes data, thereby providing overall monitoring of the electronic device. Optionally, the processor 401 may include one or more processing cores; preferably, the processor 401 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 401.
[0230] The memory 402 can be used to store software programs and modules. The processor 401 executes various functional applications and data processing by running the software programs and modules stored in the memory 402. The memory 402 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the electronic device, etc. In addition, the memory 402 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 402 may also include a memory controller to provide the processor 401 with access to the memory 402.
[0231] The electronic device also includes a power supply 403 that supplies power to the various components. Preferably, the power supply 403 can be logically connected to the processor 401 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply 403 may also include one or more DC or AC power supplies, recharging systems, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components.
[0232] The electronic device may also include an input unit 404, which can be used to receive input digital or character information, and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.
[0233] Although not shown, the electronic device may also include a display unit, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 401 in the electronic device loads the executable files corresponding to the processes of one or more applications into the memory 402 according to the following instructions, and the processor 401 runs the applications stored in the memory 402 to realize various functions, as follows:
[0234] The system acquires operation data of a target object on promotional content within a preset time period. The operation data includes operation records in at least one dimension. The system performs data standardization processing on the operation records in each dimension to obtain an operation data sequence, which includes the standardized operation records in each dimension. The system extracts features from the standardized operation records in each dimension to obtain the operation sequence features of the target object on the promotional content. Based on the operation sequence features, the system identifies abnormal operations of the target object.
[0235] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.
[0236] As can be seen from the above, this embodiment can acquire the operation data of the target object regarding the promotional content within a preset time period. The operation data includes operation records in at least one dimension. Data standardization processing is performed on the operation records in each dimension to obtain an operation data sequence, which includes standardized operation records in each dimension. Feature extraction is performed on the standardized operation records in each dimension to obtain the operation sequence features of the target object regarding the promotional content. Based on the operation sequence features, abnormal operation identification is performed on the target object. This embodiment can identify whether there are abnormalities in the target object's operations regarding the promotional content by processing the operation records of the target object in at least one dimension, thus improving the accuracy of abnormal operation identification.
[0237] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.
[0238] Therefore, embodiments of this application provide a computer-readable storage medium storing a plurality of instructions that can be loaded by a processor to execute steps in any of the abnormal operation identification methods provided in embodiments of this application. For example, the instructions can execute the following steps:
[0239] The system acquires operation data of a target object on promotional content within a preset time period. The operation data includes operation records in at least one dimension. The system performs data standardization processing on the operation records in each dimension to obtain an operation data sequence, which includes the standardized operation records in each dimension. The system extracts features from the standardized operation records in each dimension to obtain the operation sequence features of the target object on the promotional content. Based on the operation sequence features, the system identifies abnormal operations of the target object.
[0240] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.
[0241] The computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.
[0242] Since the instructions stored in the computer-readable storage medium can execute the steps in any of the abnormal operation identification methods provided in the embodiments of this application, the beneficial effects that any of the abnormal operation identification methods provided in the embodiments of this application can achieve can be realized, as detailed in the preceding embodiments, and will not be repeated here.
[0243] According to one aspect of this application, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the various alternative implementations of the above-described abnormal operation identification aspect.
[0244] The above provides a detailed description of an abnormal operation identification method and related equipment provided by the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.
Claims
1. An abnormal operation recognition method characterized by comprising: include: Obtain the operation data of the target object on the promotion content within a preset time period. The operation data includes operation records of at least one dimension. The operation records include at least one operation feature record. The operation feature record includes at least one operation feature. The operation feature includes operation scenario, operation type, and operation time. The operation records for each dimension are standardized to obtain an operation data sequence. This sequence includes standardized operation records for each dimension, including: considering operation feature records with the same operation scenario, operation type, and operation time within the same time period as operation records of the same dimension; counting the frequency of occurrence of operation feature records in the operation records of that dimension, recording this as the operation frequency; and merging the operation feature records in the operation records of that dimension to obtain standardized operation records for that dimension. These standardized operation records include operation type, operation scenario, operation frequency, and operation time. Based on the standardized operation records for each dimension, an operation data sequence is constructed. Feature extraction is performed on the standard processed operation records in each dimension to obtain the operation sequence features of the target object for the promotional content. This includes: for the standard processed operation records in each dimension, predicting reference feature information corresponding to the standard processed operation records in each reference dimension based on the standard processed operation records in the dimension, wherein the standard processed operation records in the reference dimensions are the standard processed operation records in other dimensions besides the dimension in the operation data sequence; fusing the reference feature information corresponding to the standard processed operation records in each reference dimension to obtain the target feature information of the standard processed operation records in the dimension; and determining the operation sequence features of the target object for the promotional content based on the target feature information of the standard processed operation records in each dimension. Based on the operation sequence characteristics, abnormal operations are identified in the target object.
2. The method of claim 1, wherein, The data standardization process for the operation records of each dimension also includes: For each dimension, identify the feature type of the corresponding operational feature in each dimension.
3. The method of claim 2, wherein, The frequency of occurrence of operation feature records in the operation records along the specified dimension is denoted as operation frequency, including: Based on the feature type of the operation feature, the frequency of operation feature records with the same operation feature in each dimension is calculated to obtain the operation frequency; After calculating the frequency of occurrence of operation feature records in the operation records along the specified dimension, and recording it as the operation frequency, the method further includes: The frequency of the operation is compared with a preset frequency threshold to obtain the frequency information after comparison.
4. The method of claim 1, wherein, The step of determining the operation sequence characteristics of the target object for the promotional content based on the target feature information of the operation record after standard processing on each dimension includes: Determine the weights corresponding to the standard post-processing operation records for each dimension; Based on the weights, the target feature information of the standard processed operation records in each dimension is fused to obtain the operation sequence features of the target object for the promotional content.
5. The method of claim 1, wherein, The step of identifying abnormal operations on the target object based on the operation sequence characteristics includes: Based on the operation sequence characteristics, predict the probability that the target object's operation on the promotional content belongs to each preset abnormal operation type; Based on the probability, the target abnormal operation type of the target object for the promotional content is determined from each preset abnormal operation type.
6. An abnormal operation recognizing apparatus characterized by comprising: include: The acquisition unit is used to acquire the operation data of the target object on the promotional content within a preset time period. The operation data includes operation records of at least one dimension. The operation records include at least one operation feature record. The operation feature record includes at least one operation feature. The operation feature includes operation scenario, operation type, and operation time. A standardization unit is used to standardize the operation records for each dimension to obtain an operation data sequence. The operation data sequence includes standardized operation records for each dimension, including: considering operation feature records with the same operation scenario, operation type, and operation time within the same time period as operation records of the same dimension; counting the frequency of occurrence of operation feature records in the operation records of the dimension, recording it as the operation frequency; and merging the operation feature records in the operation records of the dimension to obtain standardized operation records for that dimension. The standardized operation records include operation type, operation scenario, operation frequency, and operation time. An operation data sequence is constructed based on the standardized operation records for each dimension. An extraction unit is used to extract features from the standard processed operation records in each dimension to obtain the operation sequence features of the target object for the promotional content. This includes: an extraction subunit, used to predict reference feature information corresponding to the standard processed operation records in each reference dimension based on the standard processed operation records in that dimension, wherein the standard processed operation records in the reference dimensions are the standard processed operation records in other dimensions besides the specified dimension in the operation data sequence; fusing the reference feature information corresponding to the standard processed operation records in each reference dimension to obtain the target feature information of the standard processed operation records in that dimension; and a determination subunit, used to determine the operation sequence features of the target object for the promotional content based on the target feature information of the standard processed operation records in each dimension. The identification unit is used to identify abnormal operations of the target object based on the operation sequence characteristics.
7. An electronic device, comprising: It includes a memory and a processor; the memory stores an application program, and the processor runs the application program within the memory to perform the operations in the abnormal operation identification method according to any one of claims 1 to 5.
8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a plurality of instructions adapted for loading by a processor to perform the steps of the abnormal operation identification method according to any one of claims 1 to 5.
9. A computer program product comprising computer programs or instructions, characterized in that, When the computer program or instructions are executed by the processor, they implement the steps of the abnormal operation identification method according to any one of claims 1 to 5.
Citation Information
Patent Citations
Internet anti-cheating behavior method, device and equipment and readable storage medium
CN111262854A
Template-oriented Word2vec-based log exception detection method and device
CN111459964A
