Cross-platform online application program system and construction method thereof

Abstract

The invention relates to a cross-platform online application program system and a construction method thereof, and relates to the technical field of mobile application development, the system comprises an H5 business layer configured to realize business process control, form interaction and native high-performance View embedding through same-layer rendering and an H5 general development technology; an H5 base layer configured to define an H5 tag; the primary layer is configured to focus on audio and video acquisition and rendering, equipment authority management and Native capability to form a high-security component; and the bridging layer is configured to realize pixel-level same-layer rendering of the H5 and the native component in the native layer based on a dynamic coordinate mapping and event penetration technology, and complete bidirectional instruction interaction through a standardized communication interface. According to the method, high-interactivity, high-security and high-compliance services such as online websites in financial application programs can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of mobile application development, and particularly relates to a cross-platform online application system and a construction method thereof. Background Art

[0002] In current financial applications, for online video services (such as remote account opening and online face signing), due to the need to implement real-person interaction, real-time audio and video transmission, and complex business process control, there are the following pain points in technical implementation: (1) Duplicate development on multiple platforms: The audio and video modules and business logics need to be independently implemented on iOS, Android, and HarmonyOS platforms, resulting in high development costs and difficulty in unifying quality; the cross-platform reuse rate of pure native solutions is low.

[0003] (2) Lag in version iteration: New features rely on client releases, resulting in discontinuous user experiences; (3) Poor performance experience: The pure H5 (HTML5) solution has insufficient interaction fluency: For hybrid development frameworks such as Flutter and RN, the experience has been improved compared to H5, but there are generally problems such as heavy external dependencies and high development technical costs; in Flutter, due to the additional overhead of the Skia engine in audio and video rendering, frame rate fluctuations occur (the measured average value is 25FPS, while the native is 30FPS); the bridging communication delay of React Native is relatively high (average 150ms), and the business coherence is poor.

[0004] (4) Security and compliance: The financial system has high requirements for the security and compliance of data applications. The implementations of technologies such as H5, Flutter, and RN are biased towards the front end and have natural shortcomings in terms of security compared to pure native.

[0005] Therefore, it is necessary to provide a cross-platform online application system and a construction method thereof to implement high-interactivity, high-security, and high-compliance services such as online branches in financial applications. Summary of the Invention

[0006] The purpose of the present invention is to provide a cross-platform online application system and a construction method thereof to implement high-interactivity, high-security, and high-compliance services such as online branches in financial applications.

[0007] To solve the problems existing in the prior art, the present invention provides a cross-platform online application system, including: An H5 service layer configured to implement business process control, form interaction, and native high-performance View embedding through same-layer rendering and H5 general development technology; An H5 basic layer configured to define H5 tags; The native layer is configured to focus on audio - video capture and rendering, device permission management, and Native capabilities, forming highly secure components; The bridging layer is configured to achieve pixel - level same - layer rendering of H5 and native components in the native layer based on dynamic coordinate mapping and event penetration technology, and complete two - way instruction interaction through a standardized communication interface.

[0008] Optionally, in the cross - platform online application system, the pixel - level same - layer rendering of H5 and native components in the native layer is achieved as follows: On the iOS side: Bind the H5 tag to the native layer visual layout through WKChildScrollView; On the Android side: Dynamically synchronize the H5 and Native rendering levels through the H5EmbedView component of Chromium WebView; On the HarmonyOS side: Align the views by using the XComponent of ArkUI and the H5 component.

[0009] Optionally, in the cross - platform online application system, it further includes: The intelligent analysis layer: Configured to build an end - side intelligent analysis engine based on user behavior characteristics, and achieve abnormal transaction detection, risk scoring, dynamic process decision - making, and intention prediction; The dynamic business adaptation module, configured to enable H5 to support cloud - side hot updates.

[0010] The present invention also provides a method for constructing a cross - platform online application system, including the following steps: Establish a hybrid architecture, which are the H5 business layer, the H5 basic layer, the native layer, and the bridging layer respectively; Insert H5 tags in the H5 page layer. On the iOS side, bind the H5 tag to the native layer visual layout through WKChildScrollView. On the Android side, dynamically synchronize the H5 and Native rendering levels through the H5EmbedView component of Chromium WebView. On the HarmonyOS side, align the views by using the XComponent of ArkUI and the H5 component; Perform event penetration and interaction optimization. Events penetrate through the bridging layer to each component in the native layer; the corresponding component in the native layer starts the camera and renders the picture in real - time; the verification result is transmitted back to the H5 page through the jsbridge protocol.

[0011] Optionally, in the method for constructing the cross - platform online application system, The H5 business layer is used to implement the business logic control, business forms, and business pop-ups of the online branch system in the front-end presentation layer through H5 technology; The H5 basic layer is used to define H5 tags; The native layer includes: a real-time audio and video communication unit, a face recognition verification component, a voice dual-recording software development kit, and a real-time transaction status push unit; The bridging layer is used to embed the native video component and the native verification component in the native layer into the H5 page through H5 tags.

[0012] Optionally, in the method for constructing the cross-platform online application system, the H5 tags are custom tags, and the custom tags include video tags and security verification tags.

[0013] Optionally, in the method for constructing the cross-platform online application system, When the custom tag is for face recognition; iOS side: Map the custom tag coordinates to the native face recognition View through WKChildScrollView; Android side: Dynamically bind the rendering level of the face recognition component using H5EmbedView; HarmonyOS side: Through the pixel-level rendering interface of XComponent, implement the embedding of the native face recognition component into the H5 page to ensure the visual consistency of the H5 form and the native verification component.

[0014] Optionally, in the method for constructing the cross-platform online application system, the following steps are further included: If new functions are required, the cloud pushes an H5 update package to overwrite the previous H5 tags.

[0015] Optionally, in the method for constructing the cross-platform online application system, the following steps are further included: Set up an intelligent analysis engine.

[0016] Optionally, in the method for constructing the cross-platform online application system, the intelligent analysis engine includes setting a dynamic process decision tree algorithm, a multi-modal risk scoring model, and an intention prediction long short-term memory network; The dynamic process decision tree algorithm generates a decision tree based on the user behavior sequence, and the node weights are dynamically adjusted through testing; The multi-modal risk scoring model uses a random forest algorithm to fuse transaction data and device environment characteristics, and the risk scoring formula is as follows: ; where is the risk score, is the feature weight, is the normalized feature value, is the bias term, and are the total number of risk characteristics and the th risk characteristic respectively; The intention prediction long short-term memory network is: input the user operation sequence, output the probability distribution of the next operation, and support dynamic process jump.

[0017] Compared with the prior art, the present invention has the following advantages: (1) Efficiency index: After the business function is H5-ified, the code reuse rate of the three terminals is greatly improved, the development cycle is effectively shortened, and the labor cost is significantly reduced. Function iteration is realized through H5 hot update, and the online production cycle is shortened.

[0018] (2) Performance index: Memory occupancy: The H5 business layer runs independently across processes, and the memory consumption is greatly reduced compared with the native solution; Rendering efficiency: The same-layer rendering technology makes the audio and video basically the same as the pure native solution, and the native experience effect is obtained in the H5 page; Cross-terminal consistency: The alignment error rate of UI components is low, and pixel-level consistency of iOS / Android / HarmonyOS three terminals is supported.

[0019] (3) Compliance index: Sensitive information components such as security compliance are developed natively, with the same security as native applications, which is conducive to ensuring compliance. Brief Description of the Drawings

[0020] Figure 1 is the architecture diagram of the cross-platform online application system provided by the embodiment of the present invention; Figure 2 is the cross-platform online business process diagram provided by the embodiment of the present invention; Figure 3 is the consistency algorithm flow diagram provided by the embodiment of the present invention. Detailed Embodiments

[0021] The following will describe the detailed embodiments of the present invention in more detail with reference to the schematic diagrams. According to the following description, the advantages and features of the present invention will be clearer. It should be noted that the drawings are all in a very simplified form and use non-precise scales, only for the purpose of facilitating and clearly assisting in explaining the purpose of the embodiments of the present invention.

[0022] In the following text, if the method described herein includes a series of steps, the order of these steps presented herein is not necessarily the only order in which these steps can be executed, and some of the described steps may be omitted and / or some other steps not described herein may be added to the method.

[0023] In current financial applications, online video services (such as remote account opening and online face-to-face signing) have pain points in technical implementation, such as multi-terminal repeated development, lagging version iteration, poor performance experience, and security compliance, due to the need to implement real-person interaction, real-time audio and video transmission, and complex business process control.

[0024] To solve the problems existing in the prior art, the present invention provides a cross-platform online application system, as Figure 1 shown, including: First, the H5 business layer is configured to implement reusable modules such as business process control, form interaction, and native high-performance View embedding through same-layer rendering and H5 general development technology; specifically, the H5 business layer is used to implement the business logic control, business forms, and business pop-ups of the online branch system in the front-end presentation layer through H5 technology. Second, the H5 basic layer is configured to define H5 tags (i.e., custom tags) using Shadow DOM, where Shadow DOM is the core component of the Web Components technology stack and is used to implement the encapsulation and isolation of Web components. Third, the native layer is configured to focus on high-performance modules such as audio and video capture and rendering, device permission management, and Native capabilities to form highly secure components to ensure compliance; among them, Native capabilities refer to the ability of an application to directly access and utilize device hardware and operating system functions. In mobile application development, Native capabilities usually refer to the ability of native applications to directly interact with devices, including accessing the device's camera, GPS, notification system, etc. Native is a development mode that uses the development language and framework specified by the operating system.

[0025] Specifically, the native layer includes: a real-time audio and video communication unit, a WebView, an event mechanism, a face recognition verification component, a voice dual-recording software development kit (i.e., voice dual-recording SDK), and a real-time transaction status push unit, etc.; among them, WebView is the core view class in the WebKit framework and is used to manage the interaction between the WebFrame and WebFrameView classes.

[0026] Fourth, the bridging layer is configured to implement pixel-level same-layer rendering of native components in the H5 and native layers based on dynamic coordinate mapping and event penetration technology, and complete two-way instruction interaction through a standardized communication interface (such as jsbridge, which is a bridging protocol for two-way communication between JavaScript and native code), and jsbridge requires permission application, NFC, or face recognition.

[0027] Specifically, pixel-level same-layer rendering between H5 and native components in the native layer is achieved. Here, same-layer rendering means that H5 and native components are superimposed and displayed at the same rendering level, avoiding the hierarchical occlusion problem of traditional WebViews. The same-layer rendering method is as follows: iOS side: Bind the visual layout of H5 tags to the native layer through WKChildScrollView, and optimize the communication efficiency using JSCore. Here, WKChildScrollView is a subclass of WKWebView in iOS (Apple's operating system) and is used to handle scroll views; Specifically, for the iOS side, when creating a Web view using WKWebView, during the parsing and rendering process, WKWebView will render Web components onto WKCompositingView. WKCompositingView is a subclass of the native UIView. Generally, the WKWebView kernel will render multiple components onto the same WKCompositingView. However, when the style of a certain tag sets the overflow: scroll property and the content exceeds the container size, WKWebView will create a separate WKChildScrollView for it. Therefore, the present invention needs to find the View containing the overflow: scroll property and associate the found View with the corresponding Web component one by one, so that the native component can be rendered into the View, achieving the effect of same-layer rendering.

[0028] Android side: Dynamically synchronize the rendering levels of H5 and Native through the H5EmbedView component of Chromium WebView. Here, H5EmbedView is a general term for embedding an H5 page in an Android application; HarmonyOS side: Align views by using the XComponent of ArkUI and H5 components, and support high-precision event penetration. Here, XComponent is a component in HarmonyOS used to meet complex custom drawing requirements; it is mainly used in scenarios such as the display of camera preview streams and the drawing of game screens. The XComponent of ArkUI is a component for complex custom rendering requirements and is mainly used for EGL / OpenGLES rendering and media data writing.

[0029] Specifically, for the Android and HarmonyOS platforms, since the system WebView already supports the mapping from H5 tags to native Views, the application layer only needs to register custom Views with the WebView to complete the rendering of native components into the H5 page with the help of the system (the main difference is that HarmonyOS requires an additional attribute starting with "native / ").

[0030] In summary, there are platform differences in the basic implementation of same-layer rendering for the three platforms (i.e., iOS, Android, and HarmonyOS). To ensure the consistency of the H5 basic layer, it is necessary to encapsulate the same-layer rendering components at the H5 basic layer. We use the Shadow Dom technology to implement custom tags (such as native-video-view), which inherit from HtmlElement. Then, in the constructor initialization method of the custom tag, add a div (iOS) or embed (Android, HarmonyOS) tag according to the platform, and pass in different input parameters according to different platforms. The summary of the overall consistency algorithm is referred to Figure 3 。

[0031] Optionally, in the cross-platform online application system, it further includes: Intelligent analysis layer: configured to build an end-side intelligent analysis engine based on user behavior characteristics. The intelligent analysis engine conducts intelligent analysis based on the IP address, screen recording status, and jailbreaking status to achieve abnormal transaction detection, risk scoring, dynamic process decision-making, and intent prediction, effectively improving the fluency and security of user function usage. The intelligent analysis layer of the present invention pre-trains the intelligent analysis engine model in advance on the server according to historical data through machine learning algorithms, and pre-installs it on the client. By analyzing user behavior in real time, it optimizes the end-side process and experience, avoiding the risk that the risk control measures on the server are easily bypassed, and further enhancing security. Among them, (1) Behavior data processor: cleans and standardizes user behavior data in real time; (2) Model inference engine: executes algorithms such as risk scoring and intent prediction; (3) Policy executor: dynamically adjusts the business process according to the analysis results (such as triggering enhanced verification or skipping steps, etc.).

[0032] Dynamic business adaptation module: configured to enable H5 to support cloud hot updates, and new business functions (such as optimization of the face-to-face signing process) can take effect without client version release. Among them, hot update realizes the versionless update of client functions by dynamically loading code packages from the cloud.

[0033] The present invention realizes cross-platform reuse of business logic and native optimization of high-performance and high-security modules through the same-layer rendering technology, solves problems such as low multi-end development efficiency, slow function iteration, and insufficient core experience, and at the same time can ensure security and compliance.

[0034] The present invention also provides a method for constructing a cross-platform online application system, comprising the following steps: S1: Establish a hybrid architecture, including an H5 business layer, an H5 basic layer, a native layer, and a bridging layer; The H5 business layer is used to implement the business logic control, business forms, and business pop-ups of the online network system in the front-end presentation layer through H5 technology; The H5 basic layer is used to define H5 tags, where the H5 tags are custom tags, such as tags containing videos ( <native-video-view type="video" id="communication">), security verification label ( <native-security-view type="face-recognition" id="face-auth">< / native-security-view> ) or ( <native-video-view>, <native-security-view>); The native layer includes: a real-time audio and video communication unit, a face recognition and verification component, a voice double-recording software development kit, and a real-time transaction status push unit; The bridging layer is used to embed the native video component and the native verification component in the native layer into the H5 page through H5 tags.

[0035] S2: Insert H5 tags in the H5 page layer. On the iOS side, the WKChildScrollView is used to bind the H5 tags to the visual layout of the native layer. On the Android side, the H5EmbedView component of Chromium WebView is used to dynamically synchronize the rendering levels of H5 and Native. On the HarmonyOS side, the XComponent of ArkUI and the H5 component are used to align the views; Specifically, for the iOS side, when creating a Web view using WKWebView, during the parsing and rendering process, WKWebView will render Web components onto the WKCompositingView. WKCompositingView is a subclass of the native UIView. Generally, the WKWebView kernel will render multiple components onto the same WKCompositingView. However, when the style of a certain tag sets the overflow: scroll property and the content exceeds the container size, WKWebView will create a separate WKChildScrollView for it. Therefore, the present invention needs to find the View containing the overflow: scroll property and associate the found View with the corresponding Web component one by one, so that the native component can be rendered into the View to achieve the effect of same-layer rendering. For the Android side and the HarmonyOS side, since the system WebView layer already supports the mapping from H5 tags to native Views, the application layer only needs to register custom Views with the WebView to complete the rendering of native components into the H5 page with the help of the system (the main difference is that HarmonyOS requires an additional property starting with "native / ").

[0036] In summary, there are platform differences in the basic implementation of same-layer rendering for the three platforms (i.e., iOS, Android, and HarmonyOS). To ensure the consistency of the H5 basic layer, it is necessary to encapsulate the same-layer rendering components at the H5 basic layer. We use the Shadow Dom technology to implement custom tags (such as native-video-view), which inherit from HtmlElement. Then, in the constructor initialization method of the custom tag, add a div (iOS) or embed (Android, HarmonyOS) tag according to the platform, and pass in different parameters according to different platforms. For the summary of the overall consistency algorithm, please refer to Figure 3 。

[0037] S3: Perform event penetration and interaction optimization. The event penetrates through the bridge layer to each component of the native layer; the corresponding component in the native layer starts the camera and renders the picture in real time; the verification result is passed back to the H5 page through the jsbridge protocol; S4: If new functions need to be added, the cloud pushes an H5 update package to overwrite the previous H5 tag; the user can complete the function upgrade without awareness and does not need to redownload the application.

[0038] It also includes the following steps: S5: Set up an intelligent analysis engine. The intelligent analysis engine includes setting a dynamic process decision tree algorithm, a multi-modal risk scoring model, and an intention prediction long short-term memory network. The intelligent analysis engine performs intelligent analysis based on the IP address, screen recording status, and jailbreak status to achieve abnormal transaction detection, risk scoring, dynamic process decision-making, and intention prediction; (1) The dynamic process decision tree algorithm generates a decision tree based on the user behavior sequence, and the node weights are dynamically adjusted through testing; for example, when it is detected that the user frequently modifies the ID card information, the AI-assisted recognition function is automatically enabled.

[0039] (2) The multi-modal risk scoring model uses the random forest algorithm to fuse transaction data and device environment characteristics. The risk scoring formula is as follows: ; where is the risk score, is the feature weight, is the normalized feature value, is the bias term, and are the total number of risk features and the th risk feature respectively;

[0040] (3) The intention prediction long short-term memory network (LSTM) is as follows: Input the user operation sequence and output the probability distribution of the next operation, supporting dynamic process jumps.

[0041] The intelligent analysis engine has the following advantages: (1) High-risk user identification: Transmit features such as the span of the user's IP address, whether the mobile phone is recording the screen, whether the mobile phone is a new device, and whether the user's face remains stable within the screen to the local intelligent analysis engine to make a real-time decision on whether to allow the user's transfer operation or give a prompt for the user's operation (such as prompting the user to keep the face stable within the screen during video communication to prevent the user from being manipulated by a third party). (2) Transfer limit restriction: If the intelligent analysis engine returns a high-risk flag, the transfer limit is adjusted in real time to prevent the user from being deceived. (3) The interception accuracy rate of high-risk transactions on the device reaches 60%, and the false alarm rate is <2%.

[0042] In one embodiment, when the custom tag is face recognition; Insert a custom tag in the H5 page: The real-time communication is: <native-video-view type="video" id="communication">< / native-video-view> ; The face verification is as follows: <native-security-view type="face-recognition" id="face-auth">< / native-security-view> ; iOS side: Map the coordinates of custom tags to the native face recognition View through WKChildScrollView; Android side: Dynamically bind the rendering level of the face recognition component using H5EmbedView, with an error rate < 0.3px; HarmonyOS side: Through the pixel-level rendering interface of XComponent, embed the native face recognition component into the H5 page to ensure the visual consistency between the H5 form and the native verification component.

[0043] As Figure 2 shown, in another embodiment, in the transfer transaction process of a banking application, the user needs to complete the following core operations: Real-time video: The user enters the online branch page and communicates with the business handling personnel in real time through video (i.e., clicking the call button); Form filling: Enter basic data such as payee information and amount; Security verification: Complete identity verification through real-time interactions such as face recognition and voice double recording; Transaction confirmation: Real-time display of transaction status and feedback results. This embodiment realizes the seamless integration of the H5 form with the native video module and the native security verification module through the same-layer rendering technology, solving problems such as multi-terminal parallel development, cross-terminal UI misalignment, unsmooth audio and video, stuck verification process, and low security level in traditional solutions.

[0044] The performance comparison and effects are as follows: (1) Rendering efficiency: The rendering frame rate of the native face recognition component in the H5 page is stable at 30+FPS, with a difference of < 3% from the pure native solution; The cross-terminal loading time of the form page is uniformly 400ms (±20ms).

[0045] (2) Memory occupancy: The memory occupancy of the H5 form module can be reduced by 40%+ compared to the native solution; The native verification component and the H5 component run in separate processes respectively, avoiding the risk of WebView memory overflow.

[0046] (3) User experience: The UI alignment error rate < 0.5px, and the operation experience on the three terminals is consistent; The success rate of the verification process is the same as that of the native, with a significant improvement compared to technology stacks such as H5, and the anomalies caused by rendering level conflicts are reduced by 90%.

[0047] The technical advantages are as follows: (1)Development efficiency: The code reuse rate of the H5 form is 95%, and the development cycle of the three-terminal verification module is shortened by 60%; (2)Dynamic expansion: By means of hot update, the "fingerprint verification" process is added, and the function online cycle is compressed from 3 weeks to 1 day; (3)Compliance: The native layer independently processes sensitive biological data, meeting the requirements of financial security supervision; (4)Security: The intelligent analysis engine detects abnormal transactions in real time to ensure transaction security.

[0048] The above are only the preferred embodiments of the present invention and do not impose any limitation on the present invention. Any person skilled in the art, within the scope of the technical solution of the present invention, makes any form of equivalent replacement or modification and other changes to the technical solution and technical content disclosed by the present invention, which are all within the content of the technical solution of the present invention and still fall within the protection scope of the present invention.

Claims

1. A cross-platform online application system, characterized in that: include: The H5 business layer is configured to achieve business process control, form interaction, and native high-performance View embedding through same-layer rendering and H5 general development technology; H5 base layer, configured to define H5 tags; The native layer is configured to focus on audio and video acquisition and rendering, device permission management, and Native capabilities, forming a highly secure component; The bridge layer is configured to achieve pixel-level same-layer rendering of H5 and native components in the native layer based on dynamic coordinate mapping and event penetration technology, and complete two-way command interaction through a standardized communication interface; achieve pixel-level same-layer rendering of H5 and native components in the native layer in the following ways: iOS: bind the H5 tag to the native layer visual layout through WKChildScrollView; Android: dynamically synchronize H5 and Native rendering levels through the H5EmbedView component of Chromium WebView; HarmonyOS: use ArkUI's XComponent and H5 components to achieve view alignment.

2. The cross-platform online application system according to claim 1, characterized in that: Also includes: Intelligent analysis layer: It is configured to build an end-side intelligent analysis engine based on user behavior characteristics to achieve abnormal transaction detection, risk scoring, dynamic process decision-making and intention prediction; The dynamic business adaptation module is configured to enable H5 to support cloud hot updates.

3. A method for constructing a cross-platform online application system, characterized in that: The following steps are involved: Establish a hybrid architecture, including H5 business layer, H5 basic layer, native layer and bridge layer; Insert H5 tags into the H5 page layer. On iOS, WKChildScrollView is used to bind the H5 tags to the native layer visual layout. On Android, the H5EmbedView component of Chromium WebView is used to dynamically synchronize the H5 and Native rendering layers. On HarmonyOS, ArkUI's XComponent and H5 components are used to achieve view alignment. Perform event penetration and interaction optimization. Events penetrate into the components of the native layer through the bridge layer. The corresponding components of the native layer start the camera and render the image in real time. The verification result is sent back to the H5 page via the jsbridge protocol.

4. The method for constructing a cross-platform online application system according to claim 3, characterized in that: The H5 business layer is used to implement the business logic control, business forms and business pop-up windows of the online outlet system at the front-end presentation layer through H5 technology; The H5 base layer is used to define H5 tags; The native layer includes: audio and video real-time communication unit, face recognition verification component, voice dual recording software development kit and real-time transaction status push unit; The bridge layer is used to embed the native video component and native verification component in the native layer into the H5 page through the H5 tag.

5. The method for constructing a cross-platform online application system according to claim 4, characterized in that: The H5 tag is a custom tag, which includes a video tag and a security verification tag.

6. The method for constructing a cross-platform online application system according to claim 5, characterized in that: When the custom label is face recognition; iOS: Use WKChildScrollView to map the custom label coordinates to the native face recognition View; Android: Use H5EmbedView to dynamically bind the rendering level of the face recognition component; HarmonyOS side: Through the pixel-level rendering interface of XComponent, the native face recognition component is embedded in the H5 page to ensure the visual consistency between the H5 form and the native verification component.

7. The method for constructing a cross-platform online application system according to claim 3, characterized in that: The following steps are also included: If new functions are needed, the H5 update package will be pushed to the cloud to overwrite the previous H5 tag.

8. The method for constructing a cross-platform online application system according to claim 3, characterized in that: The following steps are also included: Set up the intelligent analysis engine.

9. The method for constructing a cross-platform online application system according to claim 8, characterized in that: The intelligent analysis engine includes setting a dynamic process decision tree algorithm, a multimodal risk scoring model, and an intent prediction ‌long short-term memory network; The dynamic process decision tree algorithm generates a decision tree based on the user behavior sequence, and the node weights are dynamically adjusted through testing; The multimodal risk scoring model uses the random forest algorithm to integrate transaction data and device environment characteristics. The risk scoring formula is as follows: ;in, Score the risk, is the feature weight, is the normalized eigenvalue, is the deviation term, and are the total number of risk characteristics and the risk characteristics; The intent prediction ‌long short-term memory network is: input the user operation sequence, output the probability distribution of the next operation, and support dynamic process jumps.

Images