Insurance product full-process management system and method based on big data

Through big data analysis and deep learning technology, combined with multimodal data of social media and credit reporting, intelligent management of the entire process of insurance products is achieved, solving the problems of low efficiency of traditional insurance products and inaccurate risk assessment, and improving the Management efficiency and accuracy.

CN118967330AInactive Publication Date: 2024-11-15YUNSHUHUI (SHAANXI) TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410966703.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-18
Publication Date
2024-11-15
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The traditional insurance product management process relies on manual review and empirical judgment, is inefficient and vulnerable to subjective factors, resulting in inaccurate risk assessment and ineffective response to data growth and competitive pressure.

Method used

Adopt a full-process management system for insurance products based on big data, and through multimodal analysis of social media data and credit report, combined with deep learning algorithms and natural language processing technology, cross-modal mapping fusion feature information is generated to identify possible frauds behavior, realize intelligent management of insurance products.

Benefits of technology

It improves the efficiency and accuracy of insurance product management, can more accurately evaluate applicants' insurance needs and risks, help insurance companies design personalized products, and reduce the risk of fraud.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118967330A_ABST
    Figure CN118967330A_ABST
Patent Text Reader

Abstract

The invention discloses an insurance product full-process management system and method based on big data. The method comprises the following steps: firstly, obtaining social media data and a credit report of an applicant, then, carrying out multi-modal analysis on the social media data to obtain a social media multi-modal fusion representation matrix, and then, carrying out multi-modal fusion representation on the social media; the social media multi-modal fusion representation matrix passes through a content significance device of a foreground-like attention mechanism to obtain a significant social media multi-modal fusion representation matrix, then text semantic analysis is carried out on the credit report to obtain a text semantic coding feature vector of the credit report, and then the text semantic coding feature vector of the credit report is obtained; and determining a discrimination result based on the salient social media multi-modal fusion representation matrix and the cross-modal mapping fusion feature information of the semantic coding feature vector of the credit report text.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure relates to the field of intelligent process management, and specifically to a big data-based insurance product full-process management system and method. Background Art

[0002] Insurance product full-process management refers to a series of management activities carried out by insurance companies during the life cycle of insurance products. These activities are aimed at ensuring that the products can meet market demand while bringing profits to the company.

[0003] In today's digital age, insurance companies face increasing data complexity and immense competitive pressure. Traditional insurance product management processes often rely on manual review and empirical judgment, which is not only inefficient but also susceptible to subjective factors, leading to inaccurate risk assessments.

[0004] With the rapid development of information technology, the application of big data technology has gradually deepened in various industries, and the insurance industry is no exception. Therefore, a full-process management system and method for insurance products based on big data is expected. Summary of the Invention

[0005] The present disclosure is made in consideration of the above problems. One purpose of the present disclosure is to provide an insurance product full-process management system and method based on big data.

[0006] The present disclosure provides an insurance product full-process management system based on big data, which includes:

[0007] Screening data acquisition module, used to obtain applicants' social media data and credit reports;

[0008] A social media data analysis module, configured to perform multimodal analysis on the social media data to obtain a social media multimodal fusion representation matrix;

[0009] A social media data saliency representation module, configured to pass the social media multimodal fusion representation matrix through a content saliency device similar to a foreground attention mechanism to obtain a salient social media multimodal fusion representation matrix;

[0010] A credit report analysis module, configured to perform text semantic analysis on the credit report to obtain a text semantic encoding feature vector of the credit report;

[0011] A screening result generation module is used to determine the screening result based on the cross-modal mapping fusion feature information of the significant social media multimodal fusion representation matrix and the credit report text semantic encoding feature vector.

[0012] For example, in the big data-based insurance product full-process management system according to an embodiment of the present disclosure, the social media data analysis module includes:

[0013] a text and image extraction unit, configured to extract a text portion and an image portion from the social media data to obtain a social media text portion and a social media image portion;

[0014] A multimodal fusion encoding unit is used to input the social media text part and the social media image part into a social media CLIP model including an image encoder and a text encoder to obtain the social media multimodal fusion representation matrix.

[0015] For example, in the big data-based insurance product full-process management system according to an embodiment of the present disclosure, the social media data saliency representation module includes:

[0016] a social media foreground information extraction unit, configured to extract social media foreground information from the social media multimodal fusion representation matrix to obtain a social media foreground focus area representation matrix;

[0017] a foreground information explicitation screening unit, configured to perform foreground information explicitation screening on the social media foreground focus area representation matrix to obtain a dimensionally aligned social media foreground focus area explicitation representation matrix;

[0018] A foreground migration marking unit is used to perform foreground migration marking on the social media multimodal fusion representation matrix based on the dimensionally aligned social media foreground focus area explicit representation matrix to obtain the salient social media multimodal fusion representation matrix.

[0019] For example, in the big data-based insurance product full-process management system according to an embodiment of the present disclosure, the social media prospect information extraction unit is used to:

[0020] Performing convolution coding and nonlinear activation processing on the social media multimodal fusion representation matrix to obtain a social media multimodal fusion weight information feature map;

[0021] Along the channel dimension of the social media multimodal fusion weight information feature map, the cross-channel global mean of each pixel position in the social media multimodal fusion weight information feature map is calculated to obtain the social media foreground focus area representation matrix.

[0022] For example, in the big data-based insurance product full-process management system according to an embodiment of the present disclosure, the prospect information visualization and screening unit includes:

[0023] a mask screening subunit, configured to perform masking processing on the social media foreground focus area representation matrix based on a predetermined threshold to obtain a masked social media foreground focus area representation matrix;

[0024] The upsampling subunit is configured to upsample the masked social media foreground focus region representation matrix to obtain the dimensionally aligned social media foreground focus region explicit representation matrix.

[0025] For example, according to the big data-based insurance product full-process management system of an embodiment of the present disclosure, the mask screening unit is used to:

[0026] The eigenvalues ​​in the social media foreground focus area representation matrix that are less than or equal to the predetermined threshold are set to zero.

[0027] For example, in the big data-based insurance product full-process management system according to an embodiment of the present disclosure, the prospect migration marking unit is used to:

[0028] The product of the dimensionally aligned social media foreground focus area explicit representation matrix and the social media multimodal fusion representation matrix is ​​calculated to obtain the salient social media multimodal fusion representation matrix.

[0029] For example, in the big data-based insurance product full-process management system according to an embodiment of the present disclosure, the credit report analysis module is used to:

[0030] extracting a text portion from the credit report to obtain a credit report text portion;

[0031] After word segmentation processing is performed on the text portion of the credit report, a semantic encoder including a word embedding layer is used to obtain a semantic encoding feature vector of the credit report text.

[0032] For example, in the insurance product full-process management system based on big data according to an embodiment of the present disclosure, the screening result generation module includes:

[0033] A cross-modal mapping fusion unit, configured to use a cross-modal mapper to fuse the credit report text semantic encoding feature vector and the significant social media multimodal fusion representation matrix to obtain a multimodal representation vector of the applicant's credit qualifications;

[0034] An information identification unit, configured to use the multimodal representation vector of the applicant's credit qualifications as the cross-modal mapping fusion feature information;

[0035] a risk screening unit, configured to pass the multimodal representation vector of the applicant's credit qualifications through a classifier-based credit qualification screener to obtain a screening result, wherein the screening result is used to indicate whether the risk of fraud exceeds a predetermined threshold;

[0036] Wherein, the cross-modal mapping fusion unit is used to:

[0037] The product of the significant social media multimodal fusion representation matrix and the semantic encoding feature vector of the credit report text is calculated to obtain the multimodal representation vector of the applicant's credit qualification.

[0038] The embodiments of the present disclosure also provide a method for managing the entire insurance product process based on big data, which includes:

[0039] Accessing applicants’ social media data and credit reports;

[0040] Performing multimodal analysis on the social media data to obtain a social media multimodal fusion representation matrix;

[0041] Passing the social media multimodal fusion representation matrix through a content saliency device similar to a foreground attention mechanism to obtain a salient social media multimodal fusion representation matrix;

[0042] Performing text semantic analysis on the credit report to obtain a text semantic encoding feature vector of the credit report;

[0043] The identification result is determined based on the cross-modal mapping fusion feature information of the significant social media multimodal fusion representation matrix and the credit report text semantic encoding feature vector.

[0044] According to the big data-based full-process insurance product management system and method of the embodiments of the present disclosure, it utilizes big data analysis technology, combined with deep learning algorithms and natural language processing technology, to conduct in-depth analysis and comprehensive evaluation of multi-source data of insurance applicants to identify possible fraudulent behaviors, providing the most important foundation for realizing intelligent management of the entire insurance product process. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the following briefly introduces the drawings of the embodiments of the present disclosure. Obviously, the drawings described below only relate to some embodiments of the present disclosure, and are not intended to limit the present disclosure.

[0046] Figure 1 A schematic diagram of the application architecture of an insurance product full-process management system based on big data in an embodiment of the present disclosure is shown;

[0047] Figure 2 A schematic diagram of the structure of an insurance product full-process management system based on big data in an embodiment of the present disclosure is shown;

[0048] Figure 3A schematic diagram illustrating the structure of a social media data analysis module of a big data-based insurance product full-process management system according to an embodiment of the present disclosure is shown;

[0049] Figure 4 A schematic diagram illustrating the structure of a social media data saliency representation module of a big data-based insurance product full-process management system according to an embodiment of the present disclosure is shown;

[0050] Figure 5 A schematic diagram showing the structure of a screening result generation module of a big data-based insurance product full-process management system according to an embodiment of the present disclosure is shown;

[0051] Figure 6 A flowchart of a method for managing the entire insurance product process based on big data in an embodiment of the present disclosure is shown;

[0052] Figure 7 An application scenario diagram of the insurance product full-process management system based on big data in an embodiment of the present disclosure is shown; and

[0053] Figure 8 A schematic diagram of a storage medium according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0054] The technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present disclosure, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present disclosure without making any creative efforts are also within the scope of protection of the present disclosure.

[0055] The terms used in this specification are those commonly used in the art in view of the functions of the present disclosure, but these terms may vary according to the intentions of those skilled in the art, precedents, or new technologies in the art. In addition, specific terms may be selected by the applicant, and in such cases, their detailed meanings will be described in the detailed description of the present disclosure. Therefore, the terms used in the specification should not be understood as simple names, but rather as a general description based on the meaning of the terms and the present disclosure.

[0056] Although the present disclosure makes various references to certain modules in the system according to embodiments of the present disclosure, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0057] Flowcharts are used throughout this disclosure to illustrate the operations performed by systems according to embodiments of the present disclosure. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, various steps may be processed in reverse order or simultaneously, as needed. Furthermore, other operations may be added to these processes, or one or more operations may be removed from these processes.

[0058] Figure 1 A schematic diagram of the application architecture of an insurance product full-process management system based on big data in an embodiment of the present disclosure is shown, including a server 100 and a terminal device 200.

[0059] The terminal device 200 and the server 100 can be connected via the Internet to achieve mutual communication. Optionally, the above-mentioned Internet uses standard communication technologies and / or protocols. The Internet is typically the Internet, but it can also be any network, including but not limited to a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN), a mobile, wired or wireless network, a private network or any combination of a virtual private network. In some embodiments, technologies and / or formats including Hypertext Markup Language (HTML), Extensible Markup Language (XML), etc. are used to represent data exchanged over the network. In addition, conventional encryption technologies such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Network (VPN), and Internet Protocol Security (IPsec) can also be used to encrypt all or some links. In other embodiments, customized and / or dedicated data communication technologies can also be used to replace or supplement the above-mentioned data communication technologies.

[0060] The server 100 can provide various network services to the terminal device 200. The server 100 can be a single server, a server cluster consisting of multiple servers, or a cloud computing center. Specifically, the server 100 may include a processor 110 (Center Processing Unit, CPU), a memory 120, an input device 130, and an output device 140. The input device 130 may include a keyboard, a mouse, a touch screen, etc. The output device 140 may include a display device such as a liquid crystal display (LCD) or a cathode ray tube (CRT).

[0061] The memory 120 may include a read-only memory (ROM) and a random access memory (RAM), and provides program instructions and data stored in the memory 120 to the processor 110. In the embodiment of the present disclosure, the memory 120 may be used to store the corresponding method of the big data-based insurance product full-process management system in the embodiment of the present disclosure.

[0062] The processor 110 calls the program instructions stored in the memory 120, and the processor 110 is used to execute the steps of any one of the big data-based insurance product full-process management methods in the embodiments of the present disclosure according to the obtained program instructions.

[0063] In addition, the application architecture diagram in the embodiment of the present disclosure is intended to more clearly illustrate the technical solution in the embodiment of the present disclosure, and does not constitute a limitation on the technical solution provided by the embodiment of the present disclosure. Of course, for other application architectures and business applications, the technical solution provided by the embodiment of the present disclosure is also applicable to similar problems.

[0064] The following non-restrictive description of the big data-based insurance product full-process management system provided according to at least one embodiment of the present disclosure is given through several examples or embodiments. As described below, different features in these specific examples or embodiments can be combined with each other without conflicting with each other to obtain new examples or embodiments, and these new examples or embodiments also fall within the scope of protection of the present disclosure.

[0065] In response to the above technical problems, the technical concept of this application is to use big data analysis technology, combined with deep learning algorithms and natural language processing technology, to conduct in-depth analysis and comprehensive evaluation of multi-source data of insurance applicants to identify possible fraud and provide the most important foundation for realizing intelligent management of the entire process of insurance products.

[0066] Based on this, Figure 2A schematic diagram of the structure of a big data-based insurance product full-process management system 800 according to an embodiment of the present disclosure is shown. The big data-based insurance product full-process management system 800 includes: a screening data acquisition module 810 for acquiring an applicant's social media data and credit report; a social media data analysis module 820 for performing multimodal analysis on the social media data to obtain a social media multimodal fusion representation matrix; a social media data saliency representation module 830 for applying a foreground attention-like content saliency filter to the social media multimodal fusion representation matrix to obtain a salient social media multimodal fusion representation matrix; a credit report analysis module 840 for performing textual semantic analysis on the credit report to obtain a credit report textual semantic encoding feature vector; and a screening result generation module 850 for determining a screening result based on cross-modal mapping fusion feature information between the salient social media multimodal fusion representation matrix and the credit report textual semantic encoding feature vector.

[0067] It should be understood that the big data-based insurance product full-process management system 800 uses a series of modules to comprehensively evaluate insurance applicants to determine their insurance product suitability and pricing. The screening data acquisition module 810 is responsible for collecting applicants' social media data and credit reports. This is the starting point of the entire system, ensuring sufficient data for subsequent analysis and evaluation. The social media data analysis module 820 conducts in-depth analysis of the collected social media data. Through multimodal analysis, it transforms the social media data into a representation matrix that integrates multiple types of information, which helps to more comprehensively understand the applicant's social behavior and preferences. The social media data saliency representation module 830 further processes the social media data to highlight important information. It uses a content saliency function similar to the foreground attention mechanism to extract important feature information from the social media multimodal fusion representation matrix to form a salient social media multimodal fusion representation matrix. The credit report analysis module 840 is used to analyze the applicant's credit report. It extracts key information from the credit report through text semantic analysis and converts it into a text semantic encoding feature vector, which helps to assess the applicant's credit status. Result Generation Module 850 comprehensively analyzes and generates the final screening results. It cross-modally maps and fuses the significant social media multimodal fusion representation matrix and the textual semantic encoding feature vectors of the credit report to generate comprehensive feature information, ultimately determining the applicant's insurance product suitability and pricing. Through the collaborative operation of these modules, the entire system can more accurately assess the applicant's insurance needs and risks, providing decision support to insurance companies and helping them design more personalized and reasonable insurance products.

[0068] Specifically, the technical solution of this application first obtains the applicant's social media data and credit report. These data and credit reports provide multi-dimensional information about the applicant, helping insurance companies assess the applicant's risk profile from different perspectives. Specifically, social media data can reveal the applicant's lifestyle, spending habits, and social relationships, while the credit report records the applicant's credit history and financial status. This information helps identify potential fraud.

[0069] Considering that social media data is multimodal, it contains different types of information, such as text and images. Each modality in social media data may contain unique information to provide comprehensive behavioral characteristics about the user. Therefore, in the technical solution of the present application, it is expected to extract the text part and the image part from the social media data to obtain a social media text part and a social media image part; and input the social media text part and the social media image part into a social media CLIP model including an image encoder and a text encoder to obtain a social media multimodal fusion representation matrix. In this way, different encoders in the social media CLIP model are used to extract features from the social media text part and the social media image part respectively, thereby obtaining a more comprehensive character portrait of the applicant.

[0070] Accordingly, if Figure 3 As shown, the social media data analysis module 820 includes: a text and image extraction unit 821, used to extract the text part and the image part from the social media data to obtain a social media text part and a social media image part; a multimodal fusion encoding unit 822, used to input the social media text part and the social media image part into a social media CLIP model including an image encoder and a text encoder to obtain the social media multimodal fusion representation matrix.

[0071] It's worth noting that the social media CLIP model is a multimodal learning model that combines computer vision (image processing) and natural language processing (text processing) techniques to understand and generate text descriptions associated with images. CLIP stands for "Contrastive Language-Image Pre-training," meaning that this model is pre-trained through contrastive learning to achieve associations between images and text. The image encoder converts images into numerical vector representations, known as image features. Deep learning techniques, such as convolutional neural networks (CNNs), are used to extract visual features from images, such as color, texture, and shape. The text encoder converts text into numerical vector representations, known as text features. Natural language processing techniques, such as Transformers or recurrent neural networks (RNNs), are used to extract semantic information from text, such as keywords, phrases, and sentiment. In the social media data analysis module 820, the social media CLIP model uses the image encoder and text encoder to extract image and text features from social media data. These features are then fused into a multimodal fusion representation matrix. This matrix more comprehensively captures the information in social media data, supporting subsequent data analysis and decision-making.

[0072] When users post social media updates, they may use both text and images to express the same content or emotion, for example, a photo with descriptive text. In addition, some social media platforms provide functions, such as automatically generating descriptive text for pictures, which may also cause overlap of text and image content. In this way, redundant information may be generated in the social media text part and the social media image part. A large amount of redundant information may cause deviations in the description and characterization of user behavior patterns. Therefore, in the technical solution of the present application, the social media multimodal fusion representation matrix is ​​further passed through a content saliency device similar to a foreground attention mechanism to obtain a salient social media multimodal fusion representation matrix.

[0073] In an embodiment of the present application, the specific encoding process of passing the social media multimodal fusion representation matrix through a content saliency device similar to a foreground attention mechanism to obtain a significant social media multimodal fusion representation matrix includes: firstly extracting social media foreground information from the social media multimodal fusion representation matrix to obtain a social media foreground focus area representation matrix; then, performing foreground information explicitation screening on the social media foreground focus area representation matrix to obtain a dimensionally aligned social media foreground focus area explicitation representation matrix; and then, based on the dimensionally aligned social media foreground focus area explicitation representation matrix, performing foreground migration marking on the social media multimodal fusion representation matrix to obtain the significant social media multimodal fusion representation matrix.

[0074] Here, the content salience detector of the foreground attention mechanism introduces an attention mechanism and an information explicitation and screening mechanism to highlight the importance of foreground information in the social media multimodal fusion representation matrix, such as the user's interactive behavior on social media and the multimedia content that can directly reflect the user's thoughts and emotions. Specifically, in the content salience detector of the foreground attention mechanism, a global comprehensive foreground information distribution feature pattern expression in the social media multimodal fusion representation matrix is ​​extracted through convolutional coding, nonlinear activation processing, and cross-channel global averaging processing. In this way, this global comprehensive foreground information distribution feature pattern expression is used to extract the corresponding foreground target feature information of interest, and the foreground target feature information of interest is marked and purified through mask processing, so that the network module can focus on the foreground content in the social media multimodal fusion representation matrix.

[0075] Accordingly, if Figure 4 As shown, the social media data saliency representation module 830 includes: a social media foreground information extraction unit 831, used to extract social media foreground information from the social media multimodal fusion representation matrix to obtain a social media foreground focus area representation matrix; a foreground information explicitation screening unit 832, used to perform foreground information explicitation screening on the social media foreground focus area representation matrix to obtain a dimensionally aligned social media foreground focus area explicitation representation matrix; a foreground migration marking unit 833, used to perform foreground migration marking on the social media multimodal fusion representation matrix based on the dimensionally aligned social media foreground focus area explicitation representation matrix to obtain the salient social media multimodal fusion representation matrix.

[0076] Among them, the social media foreground information extraction unit 831 is used to: perform convolution coding and nonlinear activation processing on the social media multimodal fusion representation matrix to obtain a social media multimodal fusion weight information feature map; along the channel dimension of the social media multimodal fusion weight information feature map, calculate the cross-channel global mean of each pixel position in the social media multimodal fusion weight information feature map to obtain the social media foreground focus area representation matrix.

[0077] The foreground information explicitation and screening unit 832 includes: a mask screening subunit for performing masking processing on the social media foreground focus region representation matrix based on a predetermined threshold to obtain a masked social media foreground focus region representation matrix; and an upsampling subunit for upsampling the masked social media foreground focus region representation matrix to obtain the dimensionally aligned social media foreground focus region explicitation representation matrix. Specifically, the mask screening subunit is configured to set eigenvalues ​​in the social media foreground focus region representation matrix that are less than or equal to the predetermined threshold to zero.

[0078] Among them, the foreground migration marking unit 833 is used to: calculate the product between the dimension-aligned social media foreground focus area explicit representation matrix and the social media multimodal fusion representation matrix to obtain the significant social media multimodal fusion representation matrix.

[0079] At the same time, a text portion is extracted from the credit report to obtain a credit report text portion; the credit report text portion is then segmented and passed through a semantic encoder including a word embedding layer to obtain a semantically encoded feature vector of the credit report text. Here, word segmentation is used to divide the continuous text into meaningful units (words or phrases), and the word embedding layer and semantic encoder are used to extract key features and semantic meaning from the text to capture the semantic characteristics of the credit report text portion.

[0080] Correspondingly, the credit report analysis module 840 is used to: extract the text part from the credit report to obtain the credit report text part; perform word segmentation on the credit report text part and then pass it through a semantic encoder including a word embedding layer to obtain the semantic encoding feature vector of the credit report text.

[0081] Next, a cross-modal mapper is used to fuse the semantically encoded feature vector of the credit report text and the multimodal fusion representation matrix of the significant social media to obtain a multimodal representation vector of the applicant's credit qualifications. In other words, the cross-modal mapper is used to integrate two data sources, the semantically encoded feature vector of the credit report text and the multimodal fusion representation matrix of the significant social media, to form a comprehensive and integrated representation of the user, providing a richer feature expression. Subsequently, the multimodal representation vector of the applicant's credit qualifications is passed through a classifier-based credit qualification discriminator to obtain a screening result, which is used to indicate whether the risk of fraud exceeds a predetermined threshold.

[0082] Accordingly, if Figure 5As shown, the identification result generation module 850 includes: a cross-modal mapping fusion unit 851, which is used to use a cross-modal mapper to fuse the semantic encoding feature vector of the credit report text and the significant social media multimodal fusion representation matrix to obtain the applicant's credit qualification multimodal representation vector; an information identification unit 852, which is used to use the applicant's credit qualification multimodal representation vector as the cross-modal mapping fusion feature information; a risk identification unit 853, which is used to pass the applicant's credit qualification multimodal representation vector through a classifier-based credit qualification discriminator to obtain a screening result, and the screening result is used to indicate whether the risk of fraud exceeds a predetermined threshold; wherein, the cross-modal mapping fusion unit 851 is used to: calculate the product between the significant social media multimodal fusion representation matrix and the credit report text semantic encoding feature vector to obtain the applicant's credit qualification multimodal representation vector.

[0083] Preferably, the risk identification unit 853 is used to: calculate the vector one norm and the vector two norm of the multimodal representation vector of the credit qualification of the applicant; add each eigenvalue of the multimodal representation vector of the credit qualification of the applicant to the square root of the length of the multimodal representation vector of the credit qualification of the applicant and the inverse of the square root of the vector two norm to obtain the characteristic sum value corresponding to each eigenvalue; use the characteristic sum value corresponding to each eigenvalue as the exponent of an exponential function with a natural constant as the base to obtain the characteristic index value corresponding to each eigenvalue; multiply each eigenvalue of the multimodal representation vector of the credit qualification of the applicant with the vector one norm and a weight hyperparameter of the multimodal representation vector of the credit qualification of the applicant to obtain the characteristic product value corresponding to each eigenvalue; add and arrange the characteristic index value corresponding to each eigenvalue and the characteristic product value corresponding to each eigenvalue to obtain an optimized multimodal representation vector of the credit qualification of the applicant; and pass the optimized multimodal representation vector of the credit qualification of the applicant through a classifier-based credit qualification discriminator to obtain a screening result.

[0084] Specifically, considering the difference in feature expression dimensions between the semantic encoding feature vector of the credit report text and the multimodal fusion representation matrix of the significant social media, after fusion using the cross-modal mapper, the multimodal representation vector of the credit qualification of the applicant will be affected by the class regression constraint due to the multi-dimensional feature fusion distribution. Therefore, in the above preferred example, the structured norm representation of the multimodal representation vector of the credit qualification of the applicant is used as the local canonical coordinate for each eigenvalue of the multimodal representation vector of the credit qualification of the applicant to determine the overall distribution of the features of the multimodal representation vector of the credit qualification of the applicant relative to the rotation offset of the eigenvalue for each eigenvalue of the multimodal representation vector of the credit qualification of the applicant as the center, and the bounding box of the eigenvalue distribution of the multimodal representation vector of the credit qualification of the applicant is used to perform eigenvalue constraints to improve the overall class distribution regression constraint of the multimodal representation vector of the credit qualification of the applicant, thereby improving the training speed of the model and the accuracy of the discrimination result obtained by the credit qualification discriminator based on the classifier of the multimodal representation vector of the credit qualification of the applicant.

[0085] It should be understood that the role of a classifier is to use given categories and known training data to learn classification rules and classifiers, and then classify (or predict) unknown data. Logistic regression and SVM are commonly used to solve binary classification problems. For multi-class classification problems, logistic regression or SVM can also be used, but multiple binary classifications are required to form a multi-class classification. However, this is prone to errors and is inefficient. A commonly used multi-classification method is the Softmax classification function.

[0086] Based on the above embodiments, see Figure 6 As shown in FIG, a flowchart of a method for managing the entire process of insurance products based on big data in an embodiment of the present disclosure is provided. For example, the method for managing the entire process of insurance products based on big data can be executed by a server, which can be Figure 1 The server 100 shown in FIG. Figure 6As shown, according to the embodiment of the present disclosure, the full-process management method of insurance products based on big data includes the following steps: S510, obtaining the social media data and credit report of the applicant; S520, performing multimodal analysis on the social media data to obtain a social media multimodal fusion representation matrix; S530, passing the social media multimodal fusion representation matrix through a content saliency device similar to a foreground attention mechanism to obtain a salient social media multimodal fusion representation matrix; S540, performing text semantic analysis on the credit report to obtain a credit report text semantic encoding feature vector; S550, determining the identification result based on the cross-modal mapping fusion feature information of the salient social media multimodal fusion representation matrix and the credit report text semantic encoding feature vector.

[0087] Here, those skilled in the art will understand that the specific operations of each step in the above-mentioned insurance product full-process management method based on big data have been referred to above. Figures 2 to 5 The description of the insurance product full-process management system 800 based on big data has been introduced in detail, and therefore, its repeated description will be omitted.

[0088] Figure 7 FIG is an application scenario diagram of the insurance product full process management system based on big data according to an embodiment of the present disclosure. Figure 7 As shown, in this application scenario, first, the applicant's social media data (e.g., Figure 7 D1) and credit reports (e.g. Figure 7 Then, the social media data and the credit report are input into a server that is deployed with an insurance product full-process management algorithm based on big data (for example, Figure 7 In S) shown in , the server is capable of using the big data-based insurance product full-process management algorithm to process the social media data and the credit report to obtain a screening result indicating whether the risk of fraud exceeds a predetermined threshold.

[0089] Based on the above embodiments, the present disclosure also provides another exemplary embodiment of an electronic device. In some possible implementations, the electronic device in the present disclosure may include a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it can implement the steps of the big data-based insurance product full-process management method in the above embodiments.

[0090] For example, the electronic device is disclosed Figure 1 Taking the server 100 in the example for explanation, the processor in the electronic device is the processor 110 in the server 100, and the memory in the electronic device is the memory 120 in the server 100.

[0091] An embodiment of the present disclosure also provides a computer-readable storage medium. Figure 8 1 shows a schematic diagram of a computer-readable storage medium 1000 according to an embodiment of the present disclosure. Figure 8 As shown, the computer-readable storage medium 1000 stores computer-executable instructions 1001. When the computer-executable instructions 1001 are executed by a processor, the big data-based insurance product full-process management method according to the embodiments of the present disclosure described with reference to the above figures can be executed. The computer-readable storage medium includes, but is not limited to, volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), a hard disk, flash memory, etc.

[0092] Embodiments of the present disclosure also provide a computer program product or computer program, which includes computer-executable instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer-executable instructions from the computer-readable storage medium and executes the computer-executable instructions, causing the computer device to implement the big data-based full-process insurance product management method according to an embodiment of the present disclosure.

[0093] Those skilled in the art will appreciate that the contents disclosed in this disclosure may be subject to various modifications and improvements. For example, the various devices or components described above may be implemented through hardware, software, firmware, or a combination of some or all of the three.

[0094] In addition, although the present disclosure makes various references to certain units in the system according to embodiments of the present disclosure, any number of different units can be used and run on the client and / or server. The units are only illustrative, and different aspects of the systems and methods can use different units.

[0095] Those skilled in the art will appreciate that all or part of the steps in the above method can be performed by instructing the relevant hardware through a program, and the program can be stored in a computer-readable storage medium, such as a read-only memory, a magnetic disk, or an optical disk. Alternatively, all or part of the steps in the above embodiment can also be implemented using one or more integrated circuits. Accordingly, each module / unit in the above embodiment can be implemented in the form of hardware or in the form of software functional modules. The present disclosure is not limited to any particular combination of hardware and software.

[0096] Unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. It should also be understood that terms such as those defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and should not be interpreted in an idealized or highly formal sense, unless explicitly defined as such herein.

[0097] The above is an illustration of the present disclosure and should not be considered as a limitation thereof. Although several exemplary embodiments of the present disclosure have been described, it will be readily understood by those skilled in the art that many modifications may be made to the exemplary embodiments without departing from the novel teachings and advantages of the present disclosure. Therefore, all such modifications are intended to be included within the scope of the present disclosure as defined by the claims. It should be understood that the above is an illustration of the present disclosure and should not be considered as limited to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present disclosure is defined by the claims and their equivalents.

Claims

1. A full-process insurance product management system based on big data, characterized in that: include: Screening data acquisition module, used to obtain applicants’ social media data and credit reports; A social media data analysis module, used to perform multimodal analysis on the social media data to obtain a social media multimodal fusion representation matrix; A social media data saliency representation module, used for passing the social media multimodal fusion representation matrix through a content saliency device similar to a foreground attention mechanism to obtain a salient social media multimodal fusion representation matrix; A credit report analysis module, used for performing text semantic analysis on the credit report to obtain a text semantic coding feature vector of the credit report; A screening result generation module is used to determine the screening result based on the cross-modal mapping fusion feature information of the significant social media multimodal fusion representation matrix and the credit report text semantic encoding feature vector.

2. The insurance product full-process management system based on big data according to claim 1 is characterized in that: The social media data analysis module includes: A text and image extraction unit, configured to extract a text portion and an image portion from the social media data to obtain a social media text portion and a social media image portion; The multimodal fusion encoding unit is used to input the social media text part and the social media image part into a social media CLIP model including an image encoder and a text encoder to obtain the social media multimodal fusion representation matrix.

3. The insurance product full-process management system based on big data according to claim 2 is characterized in that: The social media data saliency representation module includes: A social media foreground information extraction unit, used to extract social media foreground information from the social media multimodal fusion representation matrix to obtain a social media foreground focus area representation matrix; A foreground information explicitation screening unit, configured to perform foreground information explicitation screening on the social media foreground focus area representation matrix to obtain a dimensionally aligned social media foreground focus area explicitation representation matrix; A foreground migration marking unit is used to perform foreground migration marking on the social media multimodal fusion representation matrix based on the dimensionally aligned social media foreground focus area explicit representation matrix to obtain the significant social media multimodal fusion representation matrix.

4. The insurance product full-process management system based on big data according to claim 3 is characterized in that: The social media prospect information extraction unit is used to: Performing convolutional coding and nonlinear activation processing on the social media multimodal fusion representation matrix to obtain a social media multimodal fusion weight information feature map; Along the channel dimension of the social media multimodal fusion weight information feature map, the cross-channel global mean of each pixel position in the social media multimodal fusion weight information feature map is calculated to obtain the social media foreground focus area representation matrix.

5. The insurance product full-process management system based on big data according to claim 4 is characterized in that: The foreground information explicit screening unit includes: A mask screening subunit, configured to perform masking processing on the social media foreground focus area representation matrix based on a predetermined threshold to obtain a masked social media foreground focus area representation matrix; The upsampling subunit is used to upsample the masked social media foreground focus area representation matrix to obtain the dimension-aligned social media foreground focus area explicit representation matrix.

6. The insurance product full-process management system based on big data according to claim 5 is characterized in that: The mask screening unit is used for: The eigenvalues ​​in the social media foreground focus area representation matrix that are less than or equal to the predetermined threshold are set to zero.

7. The insurance product full-process management system based on big data according to claim 6 is characterized in that: The foreground migration marking unit is used to: The product of the dimensionally aligned social media foreground focus area explicit representation matrix and the social media multimodal fusion representation matrix is ​​calculated to obtain the significant social media multimodal fusion representation matrix.

8. The insurance product full-process management system based on big data according to claim 7 is characterized in that: The credit report analysis module is used to: Extracting a text portion from the credit report to obtain a credit report text portion; After word segmentation processing is performed on the credit report text part, a semantic encoder including a word embedding layer is passed through to obtain a semantic encoding feature vector of the credit report text.

9. The insurance product full-process management system based on big data according to claim 8 is characterized in that: The screening result generating module comprises: A cross-modal mapping fusion unit, configured to use a cross-modal mapper to fuse the credit report text semantic encoding feature vector and the significant social media multimodal fusion representation matrix to obtain a multimodal representation vector of the credit qualification of the applicant; An information identification unit, used to use the multimodal representation vector of the credit qualification of the applicant as the cross-modal mapping fusion feature information; A risk identification unit, used for passing the multimodal representation vector of the credit qualification of the applicant through a credit qualification identifier based on a classifier to obtain an identification result, wherein the identification result is used to indicate whether the risk of fraudulent behavior exceeds a predetermined threshold; Wherein, the cross-modal mapping fusion unit is used to: The product of the significant social media multimodal fusion representation matrix and the credit report text semantic encoding feature vector is calculated to obtain the applicant's credit qualification multimodal representation vector.

10. A full-process management method for insurance products based on big data, characterized in that: include: Access to applicants’ social media data and credit reports; Performing multimodal analysis on the social media data to obtain a social media multimodal fusion representation matrix; The social media multimodal fusion representation matrix is ​​passed through a content saliency device similar to a foreground attention mechanism to obtain a salient social media multimodal fusion representation matrix; Performing text semantic analysis on the credit report to obtain a text semantic encoding feature vector of the credit report; The identification result is determined based on the cross-modal mapping fusion feature information of the significant social media multimodal fusion representation matrix and the credit report text semantic encoding feature vector.