Page monitoring method, device, computer equipment and storage medium

Abstract

The present application provides a page monitoring method, apparatus, computer equipment and storage medium, wherein the method obtains a page snapshot and a timestamp corresponding to each page snapshot; obtains a feature vector corresponding to each page snapshot; sorts each feature vector according to multiple timestamps; obtains a page monitoring model, wherein the model includes multiple serially connected neuron nodes to form a neuron queue; inputs each feature vector into a neuron node of the neuron queue, each neuron node parses the received feature vector, outputs a page monitoring result corresponding to the feature vector, and inputs the page monitoring result into the next neuron node in the neuron queue for updating, and the last neuron node in the neuron queue outputs the final monitoring result; if it is determined that a page exception exists in the application, monitoring is completed according to the exception type corresponding to the page exception; the exception type includes any one of a continuously blank page and an abnormal block structure on the page.

Description

Technical Field

[0001] The present application relates to the field of software monitoring and is applied in the financial field, and in particular to a page monitoring method, device, computer equipment and storage medium. Background Art

[0002] In the financial sector, the pages of financial applications are generally sensitive and have zero tolerance for error messages. Applications for major financial systems today often include banners, security areas, and various sub-categorized pages, making financial application pages extremely complex. Consequently, developers and operations personnel can make mistakes in information or configuration, resulting in subtle error messages on the page. These inadvertent errors can lead to production failures, causing users of financial system applications to see error messages when opening the page, impacting their user experience.

[0003] Therefore, there is an urgent need for a page monitoring method for financial systems, which focuses on monitoring abnormal information on financial application pages and improving and optimizing the user experience. Summary of the Invention

[0004] This application provides a page monitoring method, apparatus, computer device, and storage medium designed to address the problem of existing applications for major financial systems, which often include banners, sections, and sub-categorized pages, resulting in complex and cumbersome pages. Furthermore, if developers and operators are not careful, they may err on the side of the pen, mistype certain information or misconfiguration, resulting in obscure error messages on the page. Furthermore, due to negligence, these errors can cause the release of production lines, causing users of financial system applications to see error messages when opening the page, impacting their user experience.

[0005] In a first aspect, the present application provides a page monitoring method, comprising:

[0006] In the application to be monitored, obtain page snapshots and the timestamp corresponding to each page snapshot;

[0007] Perform feature extraction on each page snapshot to obtain the feature vector corresponding to each page snapshot;

[0008] Sort the feature vectors corresponding to each page snapshot according to multiple timestamps;

[0009] Obtain a pre-built page monitoring model; the page monitoring model includes multiple neuron nodes, and the multiple neuron nodes are serially connected to form a neuron queue;

[0010] Each feature vector is sequentially input into a neuron node in the neuron queue. Each neuron node parses the received feature vector, outputs the page monitoring result corresponding to the feature vector, and inputs the page monitoring result into the next neuron node in the neuron queue for update. The last neuron node in the neuron queue outputs the final monitoring result.

[0011] If it is determined according to the final monitoring result that the application has a page exception, the monitoring of the application is completed according to the exception type corresponding to the page exception; wherein the exception type includes at least one of a continuous blank page and an abnormal block structure on the page.

[0012] In some embodiments, before obtaining a page snapshot and a timestamp corresponding to each page snapshot in an application to be monitored, it also includes: obtaining the program type and format type of the application; determining multiple sensitive behavior types based on the program type; generating a behavior tracking point corresponding to each sensitive behavior type based on the format type; adding each behavior tracking point in the application to obtain a page snapshot and a timestamp when the behavior tracking point is triggered.

[0013] In some embodiments, before obtaining a pre-built page monitoring model, it also includes: obtaining multiple historical page snapshots and the timestamp corresponding to each historical page snapshot; wherein the historical page snapshots include normal page snapshots and abnormal page snapshots; determining the first sample number and the second sample number according to a preset ratio; constituting a training sample based on the first sample number of normal page snapshots and the second sample number of abnormal page snapshots; completing the training of the page monitoring model to be trained according to the historical anomaly types corresponding to the training samples and the abnormal page snapshots, and obtaining the constructed page monitoring model.

[0014] In some embodiments, based on the exception type corresponding to the page exception, it includes: determining the monitoring time corresponding to the final monitoring result based on multiple timestamps corresponding to the final monitoring result; if it is determined based on the normalized final monitoring result and the monitoring time that the page blank time corresponding to the application is greater than the preset time, the exception type is determined to be that the page is continuously blank; if the value corresponding to the normalized final monitoring result is not within the preset range, the exception type is determined to be that the page has an abnormal block structure.

[0015] In some embodiments, before determining the exception type corresponding to the page exception, the method further includes: obtaining multiple final monitoring results corresponding to the application; obtaining a result vector corresponding to each final monitoring result; obtaining a weight matrix corresponding to each result vector; normalizing each result vector according to a preset normalization formula and a weight matrix, so as to determine the exception type corresponding to each final monitoring result according to the corresponding value after normalization of each final monitoring result; wherein the expression of the normalization formula includes:

[0016]

[0017] Among them, Y is the corresponding value of the normalized final monitoring result, n is the number of final monitoring results, U is the weight matrix, and X is the result vector.

[0018] In some embodiments, feature extraction is performed on each page snapshot to obtain a feature vector corresponding to each page snapshot, including: decomposing the page snapshot according to a preset page element type, obtaining multiple page element information and the position information corresponding to each page element information in the page snapshot; wherein the page element type includes at least images, texts and buttons; if the page element type corresponding to the page element information is an image, obtaining the visual features corresponding to the page element information, and generating a feature vector based on the visual features and the corresponding position information; if the page element type corresponding to the page element information is text, obtaining the text features corresponding to the page element information, and generating a feature vector based on the text features and the corresponding position information; if the page element type corresponding to the page element information is a button, obtaining the interactive features corresponding to the page element information, and generating a feature vector based on the interactive features and the corresponding position information.

[0019] In some embodiments, after completing the monitoring of the application according to the exception type corresponding to the page exception, it also includes: parsing the final monitoring result to obtain the exception timestamp corresponding to the page exception; matching the exception timestamp and the timestamps of multiple page snapshots corresponding to the final monitoring result, and determining the page snapshot corresponding to the successfully matched timestamp as the target page snapshot; inputting the exception type and the target page snapshot into a preset page repair model, the page repair model parses the exception type and the target page snapshot, and outputs a repair plan corresponding to the page exception; completing the page exception repair of the application according to the repair plan.

[0020] In a second aspect, the present application provides a page monitoring device, comprising:

[0021] A snapshot acquisition unit, used to acquire page snapshots and a timestamp corresponding to each page snapshot in the application to be monitored;

[0022] A feature extraction unit is used to extract features from each page snapshot and obtain a feature vector corresponding to each page snapshot;

[0023] A vector sorting unit, used to sort the feature vectors corresponding to each page snapshot according to multiple timestamps;

[0024] A model acquisition unit is used to acquire a pre-built page monitoring model; the page monitoring model includes a plurality of neuron nodes, and the plurality of neuron nodes are serially connected to form a neuron queue;

[0025] A result output unit is used to sequentially input each feature vector into a neuron node in the neuron queue. Each neuron node parses the received feature vector, outputs the page monitoring result corresponding to the feature vector, and inputs the page monitoring result into the next neuron node in the neuron queue for updating. The last neuron node in the neuron queue outputs the final monitoring result.

[0026] The monitoring completion unit is used to complete the monitoring of the application according to the exception type corresponding to the page exception if it is determined that the application has a page exception based on the final monitoring result; wherein the exception type includes at least one of the page being continuously blank and the page having an abnormal block structure.

[0027] In a third aspect, the present application further provides a computer device, comprising:

[0028] memory and processor;

[0029] The memory is used to store computer programs;

[0030] The processor is used to execute the computer program and implement the steps of the page monitoring method described in the first aspect when executing the computer program.

[0031] In a fourth aspect, the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements the steps of the page monitoring method described in the first aspect above.

[0032] Embodiments of the present application provide a page monitoring method, apparatus, computer device, and storage medium. The method first obtains snapshots of pages within a financial application periodically or when necessary, and records a corresponding timestamp for each snapshot. This helps monitor page changes at different time points. Feature extraction is then performed on each page snapshot to generate a feature vector. These vectors capture the page's structural and content characteristics by analyzing the page's visual and textual content. The feature vectors are then sorted by timestamp to ensure that page changes are analyzed in the order in which they occurred. Further analysis is performed using a pre-built page monitoring model. This model consists of multiple serially connected neuron nodes, forming a neuron queue. Each feature vector is sequentially input into the neuron queue. Each neuron node parses and outputs a page monitoring result for that feature vector, which is then input into the next node for update. The last node in the neuron queue outputs the final monitoring result. The final monitoring result is then used to determine whether the page has an anomaly. If an anomaly is present, the application is monitored and processed based on the anomaly type (e.g., a persistently blank page or an unusually blocky structure) to improve page reliability. Through structured monitoring and analysis, the method ensures a high-quality user experience for financial system applications with complex page changes.

[0033] The provided method has the following beneficial effects:

[0034] 1. Improve page stability and reliability: By promptly detecting potential problems in the page, we can avoid poor user experience caused by unintentional errors by developers or operators.

[0035] 2. Automated monitoring and real-time response: The automated monitoring process can greatly reduce the burden of manual inspection, improve detection efficiency and accuracy, and ensure that any abnormal situations can be quickly responded to and repaired.

[0036] 3. Reduce user impact: Detect and fix problems before they affect users' normal use, reducing user churn and complaints caused by incorrect information.

[0037] 4. Accurate Problem Location and Resolution: Detailed classification of exception types allows for more precise problem identification, helping the team quickly locate and resolve issues. This approach, through structured monitoring and analysis, ensures a high-quality user experience for financial system applications across complex and changing interfaces.

[0038] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0040] Figure 1 This is a schematic flow chart of the steps of a page monitoring method provided by an embodiment of the present application;

[0041] Figure 2 This is a schematic flow chart of the steps of a method for obtaining an exception type provided in one embodiment of the present application;

[0042] Figure 3 This is a schematic flow chart of the steps of a method for obtaining a feature vector provided in one embodiment of the present application;

[0043] Figure 4 This is a structural diagram of a page monitoring device provided by an embodiment of the present application;

[0044] Figure 5 This is a schematic block diagram of the structure of a computer device provided in one embodiment of the present application.

[0045] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0047] The flowcharts shown in the accompanying drawings are for illustrative purposes only and do not necessarily include all contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may be decomposed, combined, or partially merged, so the actual execution order may vary depending on the actual situation.

[0048] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present invention, in the embodiments of the present invention, the words "first" and "second" are used to distinguish between identical or similar items with substantially the same functions and effects. Those skilled in the art will understand that the words "first" and "second" do not limit the quantity or execution order, and the words "first" and "second" do not necessarily mean different.

[0049] It should be understood that the terms used in this specification are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in this specification and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0050] It will also be understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0051] The following describes some embodiments of the present application in detail with reference to the accompanying drawings. In the absence of conflict, the following embodiments and features therein may be combined with each other.

[0052] In the financial sector, the pages of financial applications are generally sensitive and have zero tolerance for error messages. Applications for major financial systems today often include banners, security areas, and various sub-categorized pages, making financial application pages extremely complex. Consequently, developers and operations personnel can make mistakes in information or configuration, resulting in subtle error messages on the page. These inadvertent errors can lead to production failures, causing users of financial system applications to see error messages when opening the page, impacting their user experience.

[0053] Therefore, there is an urgent need for a page monitoring method for financial systems, which focuses on monitoring abnormal information on financial application pages and improving and optimizing the user experience.

[0054] To resolve the above issues, please refer to Figure 1 , Figure 1 This is a schematic flow chart of a page monitoring method provided in one embodiment of the present application. The page monitoring method can be implemented by a computer device, which can be deployed on a single server or a server cluster. It can also be deployed on a handheld terminal, a laptop computer, a wearable device, or a robot.

[0055] It should be noted that the acquisition of relevant information involved in the acquisition process of any information such as page snapshots by the provided method complies with relevant regulations and is carried out with the user's consent. It will not infringe on the user's privacy and will not violate relevant laws and regulations.

[0056] At the same time, the provided method can also monitor pages of applications in other fields, such as medical and educational fields where the pages of applications are very heavy and complex, and the embodiments of this application do not limit this.

[0057] Specifically, if Figure 1As shown, the provided page monitoring method includes steps S101 to S106, which are detailed as follows:

[0058] Step S101: In the application to be monitored, obtain page snapshots and a timestamp corresponding to each page snapshot.

[0059] Specifically, use tools or custom scripts to capture snapshots of application pages at different time points. You may use a headless browser (such as Puppeteer or Selenium) to automatically obtain rendered screenshots or DOM structures of the page. Attach a timestamp accurate to milliseconds to each captured page snapshot to determine the time it was obtained. This can help track the moment when changes occurred in subsequent analysis. For example, obtain the complete DOM structure and visual presentation of the login page once an hour, and automatically collect these page snapshots through scripts. This provides a detailed time record and retains the real-time page status at each system update or configuration change, allowing developers and operations personnel to trace back and compare historical status.

[0060] In some embodiments, before obtaining a page snapshot and a timestamp corresponding to each page snapshot in an application to be monitored, it also includes: obtaining the program type and format type of the application; determining multiple sensitive behavior types based on the program type; generating a behavior tracking point corresponding to each sensitive behavior type based on the format type; adding each behavior tracking point in the application to obtain a page snapshot and a timestamp when the behavior tracking point is triggered.

[0061] Before implementing monitoring, identify the specific type of application (such as web application, mobile application, etc.) and content format (such as HTML5, Naive App, etc.). This identification helps to set monitoring strategies more accurately in subsequent steps. According to the type of application, identify the "sensitive behaviors" related to the program, that is, those important interactions or operations that may cause page abnormalities. These behaviors may include login, payment, transfer, data input, etc. Based on the identified sensitive behavior types, define and configure corresponding tracking points to track the specific information when users perform these behaviors in the application. These tracking points need to be compatible with the application format to accurately capture the corresponding data. Integrate the above-generated tracking points into the application. The tracking points are designed to be automatically triggered when sensitive behaviors occur, thereby generating a page snapshot and the corresponding timestamp. This design ensures that the monitoring system can capture the real-time status of the page at key interaction points.

[0062] For example, in a web application scenario, the application type is "Financial Trading System" and the format type is "HTML5." Sensitive behaviors include user login, fund transfer, and investment operations. JavaScript tracking is set for each sensitive behavior. When a user clicks the "Login" button or performs a "Transfer," the tracking is triggered, capturing a real-time page snapshot and timestamp for subsequent analysis.

[0063] By conducting real-time monitoring of key user interactions, we ensure that possible abnormal states are captured immediately, thereby improving the speed and quality of monitoring response. By tracking sensitive behavior, we avoid extensive monitoring of all user operations, thereby reducing system load and focusing on monitoring key areas and critical operations. At the same time, by collecting accurate page status when anomalies and abnormal inducements occur, detailed background information is provided for fault diagnosis and repair. When the anomaly monitoring mechanism is more accurate and effective, it ultimately reduces problems in user experience and enhances user trust and satisfaction. The above embodiment successfully achieves efficient monitoring of sensitive operations by strategically embedding behavior tracking in the application, avoiding the previous cumbersome comprehensive monitoring strategy and improving the value and effectiveness of monitoring.

[0064] Step S102: Perform feature extraction on each page snapshot to obtain a feature vector corresponding to each page snapshot.

[0065] Specifically, important features are extracted from page snapshots using OCR, image processing libraries, or DOM parsing tools. These features may include: text content (such as titles and button labels), the position and size (coordinates and pixel size) of UI elements, color and style (CSS rules applied), and unstructured data (images and icons). These features must be converted into numerical feature vectors for input into machine learning models. For example, by parsing the DOM, the position coordinates and text content of all buttons on the page can be extracted, forming vectors such as [Btn1_Text, Btn1_CoordX, Btn1_CoordY, ...]. By decomposing visual information and page structure into a manageable digital form, the accuracy and applicability of subsequent detection algorithms are enhanced.

[0066] Step S103: Sort the feature vectors corresponding to each page snapshot according to multiple timestamps.

[0067] Specifically, by sorting captured feature vectors by their timestamps, we can construct a time-series-based dataset that facilitates the detection of trends and sudden changes. For example, we can compare feature vectors for the same page at different times of the day to examine whether there are unusual patterns in their changes, such as missing content or significant layout changes during a certain period. This makes it easier to identify the specific time points and triggering conditions for anomalies, allowing for rapid correction.

[0068] Step S104: Obtain a pre-built page monitoring model; the page monitoring model includes a plurality of neuron nodes, and the plurality of neuron nodes are serially connected to form a neuron queue.

[0069] Specifically, a neural network model is built using a deep learning framework (such as TensorFlow or PyTorch). This can be a model focused on pattern recognition and classification tasks, such as a CNN, or a many-to-one model within an RNN. For example, weighted connections between neuron nodes allow for complex feature mapping and learning of abnormal patterns, providing adaptive detection capabilities for different pages. For example, a pre-trained neural network model is built to focus on learning page layout and visual features to identify normal and abnormal samples. This improves the automation level of the detection system, reduces manual intervention, and improves the efficiency and accuracy of anomaly detection in large-scale systems.

[0070] In some embodiments, before obtaining a pre-built page monitoring model, it also includes: obtaining multiple historical page snapshots and the timestamp corresponding to each historical page snapshot; wherein the historical page snapshots include normal page snapshots and abnormal page snapshots; determining the first sample number and the second sample number according to a preset ratio; constituting a training sample based on the first sample number of normal page snapshots and the second sample number of abnormal page snapshots; completing the training of the page monitoring model to be trained according to the historical anomaly types corresponding to the training samples and the abnormal page snapshots, and obtaining the constructed page monitoring model.

[0071] Collect multiple historical page snapshots and attach a timestamp to each snapshot. These snapshots include pages captured during normal operation and pages recorded when anomalies occurred in the past.

[0072] The collected historical page snapshots are classified into two categories, "normal" and "abnormal", for the subsequent training process. According to the pre-set ratio (for example, 80% normal samples, 20% abnormal samples), the first number of samples (normal page snapshots) and the second number of samples (abnormal page snapshots) for training are calculated. According to the determined ratio, normal and abnormal pages are selected from the collected historical snapshots to form a model training sample set. This step ensures the representativeness and balance of the training data, especially when there is less abnormal data. Through appropriate proportion design, the model is prevented from being too biased towards the normal state. The model is trained using the selected training samples and the corresponding historical anomaly types. The model learns the feature differences between normal and abnormal pages. The anomaly type information is used to help the model refine its recognition capabilities for different abnormal conditions, such as distinguishing different abnormal block structures or page blank types.

[0073] For example, if there are 1000 normal pages and 100 abnormal pages in a historical dataset, and the training set is set to 80% normal and 20% abnormal, then 800 normal pages and 100 abnormal pages are selected for model training. Using deep learning techniques such as convolutional neural networks (CNNs), these training samples are fed into the model and trained multiple times. Parameters are adjusted based on training metrics (such as accuracy and recall) to optimize model performance.

[0074] By using rich historical data for training, the model can more accurately identify and distinguish between normal and abnormal page states. By using the specific types of past anomalies for training, the model can not only detect whether there is an anomaly, but also identify the specific type of anomaly, thereby improving fault diagnosis capabilities. By setting a reasonable sample ratio and a variety of abnormal samples, the model reduces the misjudgment of normal samples, reduces the false alarm rate, and improves the reliability of the monitoring system. New abnormal events can be added to the historical data after they occur, and the model can be continuously updated with the latest data to maintain the cutting-edge and dynamic adaptability of the model's detection capabilities. Through the above embodiment, the page monitoring model is fully trained with historical data before construction, which improves the adaptability and detection effect to different application scenarios.

[0075] Step S105. Input each feature vector into a neuron node in the neuron queue in sequence. Each neuron node parses the received feature vector, outputs the page monitoring result corresponding to the feature vector, and inputs the page monitoring result into the next neuron node in the neuron queue for updating. The last neuron node in the neuron queue outputs the final monitoring result.

[0076] Specifically, each feature vector is sequentially input into a neuron queue, and the intermediate results are calculated layer by layer, with the final monitoring result being provided by the output layer. Local feature analysis is performed at each intermediate node, highlighting changes in deeper features and refining the final judgment. For example, for each input feature vector, the model outputs "normal" or "abnormal." It can also output more fine-grained anomaly types, such as "blank module" or "misplaced element." This layer-by-layer analysis enhances the ability to learn the normal behavior and abnormal characteristics of complex pages, improving the accuracy of the system's anomaly identification.

[0077] Step S106. If it is determined according to the final monitoring result that the application has a page anomaly, the application is monitored according to the anomaly type corresponding to the page anomaly; wherein the anomaly type includes at least one of a continuously blank page and an abnormal block structure on the page.

[0078] Specifically, based on the output of abnormal results, automated scripts are executed or operations personnel are prompted for further manual analysis. Appropriate remediation and optimization measures are implemented for the identified abnormality types. Abnormal events are logged for future model training and adjustment. For example, if a persistently blank page is detected, an alarm is automatically triggered and backend logs or network request links are checked. By reducing manual processing time, the system ensures timely response and reduces user churn and trust issues caused by page errors.

[0079] Steps S101 to S106 work together to identify and handle exceptions in a timely manner based on the continuous acquisition and analysis of page data, thereby improving the stability of financial applications and user experience.

[0080] In some embodiments, as Figure 2 As shown, according to the exception type corresponding to the page exception, steps S106a to S106b are included.

[0081] Step S106a. Determine the monitoring duration corresponding to the final monitoring result based on multiple timestamps corresponding to the final monitoring result;

[0082] Step S106b. If the page blank duration corresponding to the application is determined to be greater than the preset duration based on the normalized final monitoring result and the monitoring duration, the abnormality type is determined to be a continuous blank page; if the value corresponding to the normalized final monitoring result is not within the preset range, the abnormality type is determined to be an abnormal block structure on the page.

[0083] By collecting multiple timestamps corresponding to the final monitoring results, the length of time the monitoring activity lasts, that is, the monitoring duration, is calculated. The final monitoring results are normalized. These normalized results can standardize the monitoring data of different indicators and time periods, making it easier to compare and analyze. By comparing the normalized monitoring results and monitoring duration, if the cumulative duration of the page being blank (the time it takes to load or render a page without content) exceeds the preset threshold during the monitoring duration, it is judged as a "continuously blank page" anomaly. This detection focuses on identifying the problem of a page displaying no content for a long time, which is a serious problem for user experience.

[0084] Observe the normalized monitoring results. If the result value deviates from the pre-set normal range, it indicates that the page has an abnormal block structure (such as unexpected blanks, misaligned or repeated modules).

[0085] For example, if monitoring results show that a page has been blank for 6 minutes out of the past 10 minutes, exceeding the preset 5-minute threshold, the anomaly type is determined to be "continuously blank page." Furthermore, in the normalized results, certain structural metrics, such as layout integrity and module display ratio, are outside the preset range (e.g., [0.8, 1.2]), confirming the presence of an "abnormal block structure" on the page, possibly due to ad loading failures or misplaced modules.

[0086] The above-mentioned embodiment improves the ability to analyze abnormal states of complex pages by subdividing specific abnormal types, making troubleshooting and resolution more targeted and efficient. Specifically, the detection of persistent blank pages ensures that users will not wait for a long time or leave due to the lack of content on the page, thereby enhancing the consistency of user experience. Through an intelligent mechanism based on monitoring time and normalized data, false positives for normal pages are reduced, and the reliability and accuracy of the overall monitoring system are improved. At the same time, detailed analysis of abnormal types provides effective data support for subsequent automated debugging, helping the system to respond and adjust more quickly. Furthermore, through multi-layer data analysis, the detailed identification of different types of page abnormalities is enhanced, thereby improving the practicality and diagnostic effect of page monitoring.

[0087] In some embodiments, before determining the exception type corresponding to the page exception, the method further includes: obtaining multiple final monitoring results corresponding to the application; obtaining a result vector corresponding to each final monitoring result; obtaining a weight matrix corresponding to each result vector; normalizing each result vector according to a preset normalization formula and a weight matrix, so as to determine the exception type corresponding to each final monitoring result according to the corresponding value after normalization of each final monitoring result; wherein the expression of the normalization formula includes:

[0088]

[0089] Among them, Y is the corresponding value of the normalized final monitoring result, n is the number of final monitoring results, U is the weight matrix, and X is the result vector.

[0090] First, multiple final monitoring results corresponding to the application are obtained and a result vector is generated for each result. Simultaneously, a corresponding weight matrix is ​​obtained for each result vector. The weight matrix is ​​typically designed based on the features contained in the result vector, so that different features have different influences during the normalization process. Each result vector is then processed using a preset normalization formula. Based on the normalized values ​​of each final monitoring result, the corresponding anomaly type is determined. Normalized results help more clearly identify values ​​that deviate from the normal range, enabling accurate anomaly classification. By considering the complex relationships and influences between multiple features, this normalization method enables more accurate and granular assessment of the normality of monitoring results. Furthermore, by using an adjustable weight matrix and formula, this method is adaptable to different application scenarios and monitoring metrics, maintaining monitoring accuracy in a changing environment. Through sophisticated normalization, results can not only be identified as anomalies but also clearly classified as specific anomaly types, enabling developers to conduct precise debugging and rapid remediation. This method has learning capabilities, and through repeated training and adjustment of the weight matrix, its recognition capabilities can be gradually improved, demonstrating greater adaptability to evolving applications. The above examples greatly improve the analysis depth and accuracy of page monitoring results through complex mathematical processing and machine learning algorithms, providing support for more efficient exception management.

[0091] In some embodiments, as Figure 3 The process of performing feature extraction on each page snapshot to obtain a feature vector corresponding to each page snapshot includes steps S102a to S102b.

[0092] Step S102a decomposes the page snapshot according to the preset page element type, obtains multiple page element information and the corresponding position information of each page element information in the page snapshot; wherein the page element type includes at least an image, text, and a button;

[0093] Step S102b. If the page element type corresponding to the page element information is an image, obtain the visual features corresponding to the page element information, and generate a feature vector based on the visual features and the corresponding position information; if the page element type corresponding to the page element information is text, obtain the text features corresponding to the page element information, and generate a feature vector based on the text features and the corresponding position information; if the page element type corresponding to the page element information is a button, obtain the interactive features corresponding to the page element information, and generate a feature vector based on the interactive features and the corresponding position information.

[0094] Each page snapshot is decomposed to identify and extract multiple elements in the page and their location information. These page elements include images, text, and buttons. Visual features of the image are extracted, such as color histogram, shape, texture, etc. Combined with the location information of the image in the page, a comprehensive feature vector is generated. Text features are extracted, such as word frequency, keywords, semantic information, etc. Combined with the location information of the text, a feature vector is generated. Interactive features of the button are extracted, including clickability, function description, user behavior data, etc. Feature vectors are also generated in combination with location information. The feature vector generated for each type of page element contains a combination of the specific features of the element (visual, textual or interactive features) and location information, ensuring that the characteristics of specific elements and their impact can be accurately located and identified in the analysis.

[0095] For example, for an image on a news website, we extract its primary color, edge detection outline information, and record its position on the page (such as at the top or bottom). We also extract the most frequent words and sentiment analysis results (such as positive or negative) from the article section of the page, combining them with the page area where the text appears. For buttons on the page, such as "Submit," we record their location, default state (clickable or disabled), and historical click event data.

[0096] Through detailed feature extraction, this method can more accurately distinguish between normal and abnormal page states, especially detecting subtle changes in complex pages. This extensive analysis, including visual, textual, and interactive features, can cover the characteristics of most page elements and provide a more comprehensive portrait of page states. It also supports deeper user behavior analysis. For example, based on the interactive characteristics of buttons, user usage patterns and preferences can be inferred, helping to improve user interface design. This fine-grained feature vector generation method is applicable to a variety of business scenarios, from advertising monitoring to content management, and can provide strong support and optimization suggestions.

[0097] Through the above embodiments, the method can extract multi-level information from the page snapshot, providing a solid foundation for subsequent page anomaly identification and processing.

[0098] In some embodiments, after completing the monitoring of the application according to the exception type corresponding to the page exception, it also includes: parsing the final monitoring result to obtain the exception timestamp corresponding to the page exception; matching the exception timestamp and the timestamps of multiple page snapshots corresponding to the final monitoring result, and determining the page snapshot corresponding to the successfully matched timestamp as the target page snapshot; inputting the exception type and the target page snapshot into a preset page repair model, the page repair model parses the exception type and the target page snapshot, and outputs a repair plan corresponding to the page exception; completing the page exception repair of the application according to the repair plan.

[0099] After completing the monitoring of page anomalies, the final monitoring results are parsed to obtain the anomaly timestamp corresponding to the page anomaly. The anomaly timestamp is matched with the timestamps of multiple page snapshots recorded in the final monitoring results. The page snapshots corresponding to the successfully matched timestamps are identified and determined as the target page snapshots. The identified anomaly type and target page snapshot are input into the pre-set page repair model. This model is trained to parse the input anomaly type and page snapshot.

[0100] The page repair model outputs a repair plan for page anomalies based on the input content. Based on the repair plan, the method automatically or assistedly completes the repair operation of the application page anomaly to ensure that the page returns to normal function and display.

[0101] For example, if an application detects a page loading anomaly at a specific time (e.g., 10:30:45), it can analyze the timestamp to identify the specific page snapshot where the anomaly occurred. This snapshot and the anomaly type (e.g., "page remains blank") are input into the page repair model. Based on the trained knowledge, the model outputs recommendations, such as adjusting the page loading logic or updating specific resource files. The model then retrieves repair measures, such as updating resource files and redeploying, to restore the page state.

[0102] Through timestamp matching, it is possible to accurately locate which page snapshot caused the exception, supporting the rapid location and tracing of the problem. The use of repair models to automatically generate and implement repair plans greatly improves the efficiency of exception handling and reduces the need for manual intervention. Due to the ability to quickly match and repair exceptions, user experience is guaranteed, and the availability and reliability of applications are improved. The page repair model can accumulate experience from past exceptions and repair plans through continuous "learning", gradually optimizing repair capabilities and the accuracy of suggestions. It reduces the workload of operation and maintenance personnel, allowing them to focus on more complex development issues rather than daily exception repairs. The embodiment demonstrates an intelligent page exception repair mechanism, which greatly improves the timeliness and effectiveness of exception handling through automation and model-driven methods.

[0103] The proposed method first takes snapshots of pages in financial applications, either periodically or as needed, and records a timestamp for each snapshot. This helps monitor page changes at different points in time. Feature extraction is then performed on each page snapshot to generate feature vectors. These vectors capture the page's structural and content characteristics by analyzing the visual and textual content. The feature vectors are then sorted by timestamp to ensure that page changes are analyzed in the order in which they occurred. Further analysis is performed using a pre-built page monitoring model. This model consists of multiple serially connected neuron nodes, forming a neuron queue. Each feature vector is sequentially input into the neuron queue. Each neuron node parses and outputs a page monitoring result for that feature vector, which is then fed into the next node for update. The last node in the neuron queue outputs the final monitoring result. The final monitoring result is used to determine whether the page is abnormal. If an anomaly is present, the application is monitored and addressed based on the anomaly type (such as persistently blank pages or unusually blocky structures) to improve page reliability. Through structured monitoring and analysis, this method ensures a high-quality user experience in financial system applications with complex page changes.

[0104] The provided method has the following beneficial effects:

[0105] 1. Improve page stability and reliability: By promptly detecting potential problems in the page, we can avoid poor user experience caused by unintentional errors by developers or operators.

[0106] 2. Automated monitoring and real-time response: The automated monitoring process can greatly reduce the burden of manual inspection, improve detection efficiency and accuracy, and ensure that any abnormal situations can be quickly responded to and repaired.

[0107] 3. Reduce user impact: Detect and fix problems before they affect users' normal use, reducing user churn and complaints caused by incorrect information.

[0108] 4. Accurate Problem Location and Resolution: Detailed classification of exception types allows for more precise problem identification, helping the team quickly locate and resolve issues. This approach, through structured monitoring and analysis, ensures a high-quality user experience for financial system applications across complex and changing interfaces.

[0109] See also Figure 4 As shown, Figure 42 is a schematic diagram of the structure of a page monitoring device 200 provided in an embodiment of the present application. The page monitoring device 200 is used to perform the steps of the page monitoring method shown in the above embodiments. The page monitoring device 200 can be a single server or a server cluster, or the page monitoring device 200 can be a terminal, such as a handheld terminal, a laptop computer, a wearable device, or a robot.

[0110] like Figure 4 As shown, the page monitoring device 200 includes:

[0111] The snapshot obtaining unit 201 is used to obtain page snapshots and a timestamp corresponding to each page snapshot in the application to be monitored;

[0112] A feature extraction unit 202 is used to extract features from each page snapshot and obtain a feature vector corresponding to each page snapshot;

[0113] A vector sorting unit 203 is configured to sort the feature vectors corresponding to each page snapshot according to multiple timestamps;

[0114] The model acquisition unit 204 is used to acquire a pre-built page monitoring model; the page monitoring model includes a plurality of neuron nodes, and the plurality of neuron nodes are serially connected to form a neuron queue;

[0115] The result output unit 205 is used to sequentially input each feature vector into a neuron node in the neuron queue. Each neuron node parses the received feature vector, outputs the page monitoring result corresponding to the feature vector, and inputs the page monitoring result into the next neuron node in the neuron queue for updating. The last neuron node in the neuron queue outputs the final monitoring result.

[0116] The monitoring completion unit 206 is used to complete the monitoring of the application according to the exception type corresponding to the page exception if it is determined according to the final monitoring result that the application has a page exception; wherein the exception type includes at least one of a continuous blank page and an abnormal block structure on the page.

[0117] It should be noted that technical personnel in the relevant field can clearly understand that for the convenience and conciseness of description, the specific working processes of the page monitoring device and each module described above can refer to the corresponding processes in the page monitoring method embodiments described in the above embodiments, and will not be repeated here.

[0118] The above page monitoring method can be implemented in the form of a computer program. The computer program can be used in Figure 4 Run on the device shown.

[0119] See also Figure 5 , Figure 5 1 is a schematic block diagram of the structure of a computer device provided in an embodiment of the present application. The computer device includes a processor, a memory, and a network interface connected via a device bus, wherein the memory may include a storage medium and an internal memory.

[0120] The storage medium can store an operating device and a computer program. The computer program includes program instructions, and when the program instructions are executed, the processor can execute any one of the page monitoring methods.

[0121] The processor is used to provide computing and control capabilities and support the operation of the entire computer equipment.

[0122] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor, the processor can execute any page monitoring method.

[0123] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the terminal to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0124] It should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0125] In one embodiment, the processor is configured to execute a computer program stored in the memory to implement the following steps:

[0126] In the application to be monitored, obtain page snapshots and the timestamp corresponding to each page snapshot;

[0127] Perform feature extraction on each page snapshot to obtain the feature vector corresponding to each page snapshot;

[0128] Sort the feature vectors corresponding to each page snapshot according to multiple timestamps;

[0129] Obtain a pre-built page monitoring model; the page monitoring model includes multiple neuron nodes, and the multiple neuron nodes are serially connected to form a neuron queue;

[0130] Each feature vector is sequentially input into a neuron node in the neuron queue. Each neuron node parses the received feature vector, outputs the page monitoring result corresponding to the feature vector, and inputs the page monitoring result into the next neuron node in the neuron queue for update. The last neuron node in the neuron queue outputs the final monitoring result.

[0131] If it is determined according to the final monitoring result that the application has a page exception, the monitoring of the application is completed according to the exception type corresponding to the page exception; wherein the exception type includes at least one of a continuous blank page and an abnormal block structure on the page.

[0132] In some embodiments, before obtaining a page snapshot and a timestamp corresponding to each page snapshot in an application to be monitored, it also includes: obtaining the program type and format type of the application; determining multiple sensitive behavior types based on the program type; generating a behavior tracking point corresponding to each sensitive behavior type based on the format type; adding each behavior tracking point in the application to obtain a page snapshot and a timestamp when the behavior tracking point is triggered.

[0133] In some embodiments, before obtaining a pre-built page monitoring model, it also includes: obtaining multiple historical page snapshots and the timestamp corresponding to each historical page snapshot; wherein the historical page snapshots include normal page snapshots and abnormal page snapshots; determining the first sample number and the second sample number according to a preset ratio; constituting a training sample based on the first sample number of normal page snapshots and the second sample number of abnormal page snapshots; completing the training of the page monitoring model to be trained according to the historical anomaly types corresponding to the training samples and the abnormal page snapshots, and obtaining the constructed page monitoring model.

[0134] In some embodiments, based on the exception type corresponding to the page exception, it includes: determining the monitoring time corresponding to the final monitoring result based on multiple timestamps corresponding to the final monitoring result; if it is determined based on the normalized final monitoring result and the monitoring time that the page blank time corresponding to the application is greater than the preset time, the exception type is determined to be that the page is continuously blank; if the value corresponding to the normalized final monitoring result is not within the preset range, the exception type is determined to be that the page has an abnormal block structure.

[0135] In some embodiments, before determining the exception type corresponding to the page exception, the method further includes: obtaining multiple final monitoring results corresponding to the application; obtaining a result vector corresponding to each final monitoring result; obtaining a weight matrix corresponding to each result vector; normalizing each result vector according to a preset normalization formula and a weight matrix, so as to determine the exception type corresponding to each final monitoring result according to the corresponding value after normalization of each final monitoring result; wherein the expression of the normalization formula includes:

[0136]

[0137] Among them, Y is the corresponding value of the normalized final monitoring result, n is the number of final monitoring results, U is the weight matrix, and X is the result vector.

[0138] In some embodiments, feature extraction is performed on each page snapshot to obtain a feature vector corresponding to each page snapshot, including: decomposing the page snapshot according to a preset page element type, obtaining multiple page element information and the position information corresponding to each page element information in the page snapshot; wherein the page element type includes at least images, texts and buttons; if the page element type corresponding to the page element information is an image, obtaining the visual features corresponding to the page element information, and generating a feature vector based on the visual features and the corresponding position information; if the page element type corresponding to the page element information is text, obtaining the text features corresponding to the page element information, and generating a feature vector based on the text features and the corresponding position information; if the page element type corresponding to the page element information is a button, obtaining the interactive features corresponding to the page element information, and generating a feature vector based on the interactive features and the corresponding position information.

[0139] In some embodiments, after completing the monitoring of the application according to the exception type corresponding to the page exception, it also includes: parsing the final monitoring result to obtain the exception timestamp corresponding to the page exception; matching the exception timestamp and the timestamps of multiple page snapshots corresponding to the final monitoring result, and determining the page snapshot corresponding to the successfully matched timestamp as the target page snapshot; inputting the exception type and the target page snapshot into a preset page repair model, the page repair model parses the exception type and the target page snapshot, and outputs a repair plan corresponding to the page exception; completing the page exception repair of the application according to the repair plan.

[0140] A computer-readable storage medium is also provided in an embodiment of the present application. The computer-readable storage medium stores a computer program. The computer program includes program instructions. The processor executes the program instructions to implement the steps of the page monitoring method provided in the above embodiments of the present application.

[0141] The computer-readable storage medium may be an internal storage unit of the computer device described in the aforementioned embodiment, such as a hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a smart memory card (SMC), a secure digital (SD) card, a flash memory card, etc., equipped on the computer device.

[0142] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present application, and such modifications or substitutions should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A page monitoring method, characterized in that: include: In the application to be monitored, obtaining page snapshots and a timestamp corresponding to each of the page snapshots; Performing feature extraction on each of the page snapshots to obtain a feature vector corresponding to each of the page snapshots; sorting the feature vectors corresponding to each of the page snapshots according to the multiple timestamps; Obtaining a pre-built page monitoring model; the page monitoring model includes a plurality of neuron nodes, wherein the plurality of neuron nodes are serially connected to form a neuron queue; Sequentially inputting each of the feature vectors into a neuron node in the neuron queue, each neuron node parses the received feature vector, outputs a page monitoring result corresponding to the feature vector, and inputs the page monitoring result into the next neuron node in the neuron queue for updating, and the last neuron node in the neuron queue outputs a final monitoring result; If it is determined according to the final monitoring result that the application has a page anomaly, the monitoring of the application is completed according to the anomaly type corresponding to the page anomaly; wherein the anomaly type includes at least one of a continuous blank page and an abnormal block structure on the page.

2. The method according to claim 1, characterized in that Before obtaining the page snapshots and the timestamp corresponding to each of the page snapshots in the application to be monitored, the method further includes: Obtaining a program type and a format type of the application; determining a plurality of sensitive behavior types according to the program type; Generate a behavior tracking point corresponding to each sensitive behavior type according to the format type; Each of the behavior tracking points is added to the application to obtain the page snapshot and the timestamp when the behavior tracking point is triggered.

3. The method according to claim 1, characterized in that Before obtaining the pre-built page monitoring model, the following steps are also included: Acquire multiple historical page snapshots and a timestamp corresponding to each of the historical page snapshots; wherein the historical page snapshots include normal page snapshots and abnormal page snapshots; determining a first sample size and a second sample size according to a preset ratio; Forming a training sample based on the first sample number of normal page snapshots and the second sample number of abnormal page snapshots; The training of the page monitoring model to be trained is completed according to the historical anomaly types corresponding to the training samples and the abnormal page snapshots, and the constructed page monitoring model is obtained.

4. The method according to claim 1, wherein The exception type corresponding to the page exception includes: Determining a monitoring duration corresponding to the final monitoring result according to the multiple timestamps corresponding to the final monitoring result; If it is determined based on the normalized final monitoring result and the monitoring duration that the page corresponding to the application is blank for longer than a preset duration, determining that the abnormality type is that the page is continuously blank; If the value corresponding to the normalized final monitoring result is not within a preset range, it is determined that the abnormality type is that an abnormal block structure exists on the page.

5. The method according to claim 1, wherein Before the abnormality type corresponding to the page abnormality, the method further includes: Obtaining a plurality of final monitoring results corresponding to the application; Obtaining a result vector corresponding to each of the final monitoring results; Obtaining a weight matrix corresponding to each of the result vectors; Each of the result vectors is normalized according to a preset normalization formula and the weight matrix, so as to determine the abnormality type corresponding to each of the final monitoring results according to the corresponding value after normalization of each of the final monitoring results; wherein the expression of the normalization formula includes: ； in, is the value corresponding to the normalized final monitoring result, n is the number of the final monitoring results, U is the weight matrix, and X is the result vector.

6. The method according to claim 1, characterized in that The performing feature extraction on each of the page snapshots to obtain a feature vector corresponding to each of the page snapshots includes: Decomposing the page snapshot according to preset page element types to obtain a plurality of page element information and corresponding position information of each page element information in the page snapshot; wherein the page element types include at least images, texts, and buttons; If the page element type corresponding to the page element information is an image, obtaining a visual feature corresponding to the page element information, and generating the feature vector according to the visual feature and the corresponding position information; If the page element type corresponding to the page element information is the text, obtaining text features corresponding to the page element information, and generating the feature vector according to the text features and the corresponding position information; If the page element type corresponding to the page element information is the button, an interaction feature corresponding to the page element information is acquired, and the feature vector is generated according to the interaction feature and the corresponding position information.

7. The method according to claim 1, characterized in that After completing monitoring of the application program according to the exception type corresponding to the page exception, the method further includes: Analyze the final monitoring result and obtain the abnormal timestamp corresponding to the page abnormality; Matching the abnormal timestamp with the timestamps of the plurality of page snapshots corresponding to the final monitoring result, and determining the page snapshot corresponding to the timestamp that successfully matches as the target page snapshot; Inputting the exception type and the target page snapshot into a preset page repair model, the page repair model parses the exception type and the target page snapshot, and outputs a repair solution corresponding to the page exception; Complete the page abnormality repair of the application program according to the repair solution.

8. A page monitoring device, characterized in that: include: A snapshot acquisition unit, configured to acquire page snapshots and a timestamp corresponding to each page snapshot in the application to be monitored; A feature extraction unit, configured to extract features from each of the page snapshots and obtain a feature vector corresponding to each of the page snapshots; a vector sorting unit, configured to sort the feature vectors corresponding to each of the page snapshots according to the multiple timestamps; A model acquisition unit is used to acquire a pre-built page monitoring model; the page monitoring model includes a plurality of neuron nodes, and the plurality of neuron nodes are serially connected to form a neuron queue; A result output unit is used to sequentially input each of the feature vectors into a neuron node in the neuron queue, each neuron node parses the received feature vector, outputs a page monitoring result corresponding to the feature vector, and inputs the page monitoring result into the next neuron node in the neuron queue for updating, and the last neuron node in the neuron queue outputs a final monitoring result; A monitoring completion unit is used to complete the monitoring of the application program according to the exception type corresponding to the page exception if it is determined that the application program has a page exception according to the final monitoring result; wherein the exception type includes at least one of a page being continuously blank and a page having an abnormal block structure.

9. A computer device, characterized in that: The computer device includes a memory and a processor; The memory is used to store computer programs; The processor is configured to execute the computer program and implement the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, causes the processor to implement the method according to any one of claims 1 to 7.

Images