Risk assessment method and device based on multi-source data analysis, equipment and medium

Abstract

The invention belongs to the field of artificial intelligence, and relates to a risk assessment method and device based on multi-source data analysis, equipment and a medium, and the method comprises the steps: obtaining initial data and a proof file input by a user in a target interface; according to the initial data, processing through a dynamic form engine to generate a personalized interaction form; extracting a key field from the proof file, and performing matching fusion on the interaction form and the key field to generate fusion data; adjusting the reference problem based on the initial data to obtain an adaptive problem of the user; displaying the adaptation problem on the target interface to obtain supplementary data input by the user for the adaptation problem; acquiring multi-source environmental risk factor data, and constructing a multi-dimensional feature vector of the user in combination with the fusion data and the supplementary data; and based on the multi-dimensional feature vector, a deep learning model is adopted to calculate a risk score of the user, and a risk assessment result of the user is obtained. The method can be applied to the business fields of financial insurance and the like, and the accuracy of user risk assessment can be improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology and is applied to online processing business scenarios such as finance and insurance. In particular, it relates to a risk assessment method, device, equipment and medium based on multi-source data analysis. Background Technology

[0002] In the field of front-end development in the insurance industry, business operations currently rely mainly on traditional technical means. These means have many technical bottlenecks in terms of user experience, system adaptability, data processing, and risk assessment.

[0003] Traditional insurance front-end development primarily relies on form filling, human customer service, or simple rule engines (such as automated underwriting based on preset conditions). In terms of user experience, users must manually fill in a large amount of redundant information. For example, during the insurance application process, identity information and health status need to be repeatedly entered, making the process cumbersome, error-prone, and resulting in a poor user experience. Regarding dynamic adaptability, existing systems lack real-time data perception and processing capabilities, failing to dynamically adjust the logic of subsequent business processes based on real-time user input, leading to rigid business processes and difficulty in meeting the personalized needs of different users. For unstructured data processing, existing systems cannot automatically parse user-uploaded medical reports and imaging data, relying solely on manual processing, which is not only inefficient but also prone to missing key information, affecting the accuracy and timeliness of business processing. In the risk assessment stage, traditional underwriting relies on a fixed rule base, unable to dynamically adjust risk scores in real-time with market data such as natural disasters and disease trends, resulting in inaccurate risk assessment decisions and an inability to effectively cope with complex and ever-changing market environments.

[0004] In summary, existing front-end development technologies in the insurance industry suffer from poor user experience, insufficient dynamic adaptability, weak unstructured data processing capabilities, and lagging risk assessment, resulting in low accuracy in user risk assessment. Summary of the Invention

[0005] The purpose of this application is to propose a risk assessment method, apparatus, computer equipment, and storage medium based on multi-source data analysis to solve the problem of low accuracy in existing user risk assessments.

[0006] Firstly, a risk assessment method based on multi-source data analysis is provided, which adopts the following technical solution:

[0007] The system acquires initial data and supporting documents entered by the user on the target interface; processes the initial data using a pre-defined dynamic form engine to generate a personalized interactive form; extracts key fields from the supporting documents and matches and merges the interactive form with the key fields to generate fused data; adjusts pre-defined reference questions based on the initial data to obtain the user's adaptation questions; displays the adaptation questions on the target interface to obtain supplementary data entered by the user regarding the adaptation questions; acquires multi-source environmental risk factor data, and constructs a multi-dimensional feature vector for the user based on the multi-source environmental risk factor data, which reflects the environmental risk status; and calculates the user's risk score using a deep learning model based on the multi-dimensional feature vector to obtain the user's risk assessment result.

[0008] Secondly, a risk assessment device based on multi-source data analysis is provided, which adopts the following technical solution:

[0009] The acquisition module is used to acquire the initial data and supporting documents entered by the user on the target interface;

[0010] The processing module is used to process the initial data using a preset dynamic form engine to generate personalized interactive forms.

[0011] The extraction module is used to extract key fields from the proof documents and match and merge the interactive forms with the key fields to generate merged data;

[0012] The adjustment module is used to adjust the preset reference problem based on the initial data to obtain the user's adaptation problem;

[0013] The display module is used to display adaptation issues on the target interface in order to obtain supplementary data input by the user regarding the adaptation issues;

[0014] The module is used to acquire multi-source environmental risk factor data, and to construct a user's multi-dimensional feature vector based on the multi-source environmental risk factor data, fused data, and supplementary data. The multi-source environmental risk factor data is data that reflects the environmental risk status.

[0015] The calculation module is used to calculate the user's risk score based on multi-dimensional feature vectors and a deep learning model, thereby obtaining the user's risk assessment result.

[0016] Thirdly, a computer device is provided, which adopts the following technical solution:

[0017] The system acquires initial data and supporting documents entered by the user on the target interface; processes the initial data using a pre-defined dynamic form engine to generate a personalized interactive form; extracts key fields from the supporting documents and matches and merges the interactive form with the key fields to generate fused data; adjusts pre-defined reference questions based on the initial data to obtain the user's adaptation questions; displays the adaptation questions on the target interface to obtain supplementary data entered by the user regarding the adaptation questions; acquires multi-source environmental risk factor data, and constructs a multi-dimensional feature vector for the user based on the multi-source environmental risk factor data, which reflects the environmental risk status; and calculates the user's risk score using a deep learning model based on the multi-dimensional feature vector to obtain the user's risk assessment result.

[0018] Fourthly, a computer-readable storage medium is provided, which adopts the following technical solution:

[0019] The system acquires initial data and supporting documents entered by the user on the target interface; processes the initial data using a pre-defined dynamic form engine to generate a personalized interactive form; extracts key fields from the supporting documents and matches and merges the interactive form with the key fields to generate fused data; adjusts pre-defined reference questions based on the initial data to obtain the user's adaptation questions; displays the adaptation questions on the target interface to obtain supplementary data entered by the user regarding the adaptation questions; acquires multi-source environmental risk factor data, and constructs a multi-dimensional feature vector for the user based on the multi-source environmental risk factor data, which reflects the environmental risk status; and calculates the user's risk score using a deep learning model based on the multi-dimensional feature vector to obtain the user's risk assessment result.

[0020] Compared with existing technologies, the embodiments of this application have the following main advantages: Through a dynamic form engine, personalized interactive forms can be generated based on initial user data, avoiding users repeatedly filling in redundant information, greatly simplifying the operation process, and significantly improving the user experience. Key fields are extracted from supporting documents and integrated with the interactive form to obtain fused data, enabling automatic parsing of unstructured data, reducing manual processing, lowering the risk of missing key information, and improving the accuracy and timeliness of business processing. Supplementary data is obtained by adjusting preset reference questions based on initial data, and a multi-dimensional feature vector of the user is constructed by combining the fused data and multi-source environmental risk factor data, enabling real-time perception of multi-source data. Risk scores are calculated based on multi-dimensional feature vectors using a deep learning model, breaking through the limitations of traditional fixed rule bases. It can dynamically adjust the risk assessment logic in real-time by combining external dynamic data, making risk assessment decisions more accurate and improving the accuracy of user risk assessment. Attached Figure Description

[0021] To more clearly illustrate the solutions in this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is an exemplary system architecture diagram to which this application can be applied;

[0023] Figure 2 A flowchart of an embodiment of the risk assessment method based on multi-source data analysis according to this application;

[0024] Figure 3 This is a schematic diagram of a structure of an embodiment of the risk assessment device based on multi-source data analysis according to this application;

[0025] Figure 4 This is a schematic diagram of the structure of one embodiment of the computer device according to this application. Detailed Implementation

[0026] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings of this application, are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings of this application are used to distinguish different objects, not to describe a particular order.

[0027] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0028] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0029] like Figure 1As shown, system architecture 100 may include terminal device 101, network 102, and server 103. Terminal device 101 may be a laptop 1011, tablet 1012, or mobile phone 1013. Network 102 is used as a medium to provide a communication link between terminal device 101 and server 103. Network 102 may include various connection types, such as wired, wireless communication links, or fiber optic cables.

[0030] Users can use terminal device 101 to interact with server 103 via network 102 to receive or send messages, etc. Various communication client applications can be installed on terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social media platform software, etc.

[0031] Terminal device 101 can be various electronic devices with a display screen and support web browsing. In addition to laptop computer 1011, tablet computer 1012 or mobile phone 1013, terminal device 101 can also be e-book reader, MP3 player (Moving Picture Experts Group Audio Layer III), MP4 player (Moving Picture Experts Group Audio Layer IV), laptop computer and desktop computer, etc.

[0032] Server 103 can be a server that provides various services, such as a backend server that provides support for the pages displayed on terminal device 101.

[0033] It should be noted that the risk assessment method based on multi-source data analysis provided in this application embodiment is generally executed by a server / terminal device, and correspondingly, the risk assessment device based on multi-source data analysis is generally set in the server / terminal device.

[0034] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0035] Continue to refer to Figure 2 The diagram illustrates a flowchart of an embodiment of a risk assessment method based on multi-source data analysis according to this application. The risk assessment method based on multi-source data analysis includes the following steps:

[0036] Step S201: Obtain the initial data and supporting documents entered by the user on the target interface.

[0037] In this context, "user" refers to an individual who submits operational requests and provides relevant information to complete the insurance business process within an insurance business scenario. For example, a natural person applying for insurance will input their identity, health, and other information.

[0038] The target interface refers to the interactive platform within the insurance front-end system, providing users with a visual interface for receiving user input and displaying system feedback. It represents the window through which users interact with the system, facilitating information exchange. Examples include insurance business operation pages on web pages or mobile apps.

[0039] Initial data refers to the basic information that users enter for the first time on the target interface to initiate the insurance business process, including but not limited to personal identity information and health status information.

[0040] Supporting documents refer to documents and materials related to insurance business that users upload to prove certain circumstances, such as images of medical documents and insurance documents. For example, an image of a hospital diagnosis report uploaded by a user.

[0041] Step S202: Based on the initial data, process it through a preset dynamic form engine to generate a personalized interactive form.

[0042] Among them, the dynamic form engine refers to the software module in the insurance front-end system that can dynamically generate personalized interactive forms in real time based on the initial data entered by the user.

[0043] In this context, "processing" refers to a series of calculations, analyses, and transformations performed by the dynamic form engine on the user's initial data to generate personalized interactive forms that meet the user's needs.

[0044] Interactive forms refer to forms generated by a dynamic form engine that match the user's initial data, allowing the user to perform further operations and fill in supplementary information, thus collecting more detailed information from the user.

[0045] Step S203: Extract key fields from the proof document and match and merge the interactive form with the key fields to generate merged data.

[0046] Key fields refer to information elements extracted from user-uploaded supporting documents that play a crucial role in insurance risk assessment, such as disease names and medication records. These represent the core information in the supporting documents and are used to accurately construct user characteristics. For example, the disease name field extracted from medical reports.

[0047] Among them, matching and fusion refers to the operation of associating and integrating the extracted key fields with relevant information in the interactive form to generate more comprehensive and accurate fused data, such as matching and fusion of the disease name field with the user's health status section in the form.

[0048] Among them, fused data refers to the dataset generated after the matching and fusion operation, which integrates the user's initial data, key fields of the supporting documents, and interactive form information, and represents more comprehensive and accurate user information.

[0049] Step S204: Based on the initial data, adjust the preset reference problem to obtain the user's adaptation problem.

[0050] The reference questions refer to a set of questions pre-set in the insurance front-end system to guide users to provide more information related to risk assessment.

[0051] Among them, adjustment refers to modifying and improving the preset reference questions based on the user's initial data to make them more in line with the user's actual situation.

[0052] Among these, adaptation questions refer to questions that have been adjusted to match the user's initial data and are more effective in obtaining supplementary information from the user, thus collecting more targeted user data. For example, disease-related questions adjusted for a user's specific age group.

[0053] Step S205: Display the adaptation issues on the target interface to obtain supplementary data input by the user regarding the adaptation issues.

[0054] Supplementary data refers to the data entered by users on the target interface in response to adaptation issues. This data is used to further improve user information, representing more detailed user circumstances and constructing a more comprehensive user feature vector.

[0055] Step S206: Obtain multi-source environmental risk factor data. Based on the multi-source environmental risk factor data, fused data, and supplementary data, construct a multi-dimensional feature vector for the user. The multi-source environmental risk factor data is data that reflects the environmental risk status.

[0056] Multi-source environmental risk factor data refers to data outside of the insurance front-end system that can provide additional information for risk assessment, such as data provided by meteorological APIs and disease databases. This data represents a broader range of market and environmental information, enhancing the accuracy and comprehensiveness of risk assessment. Specifically, the acquisition methods for multi-source environmental risk factor data vary depending on the business scenario. For auto insurance, disaster data can be obtained from meteorological APIs and information on accident-prone road sections from traffic department interfaces; for health insurance, disease databases and hospital electronic medical record systems can be utilized; and for life insurance, demographic databases and macroeconomic data platforms can be referenced to enhance risk assessment.

[0057] Step S207: Based on the multidimensional feature vector, a deep learning model is used to calculate the user's risk score and obtain the user's risk assessment result.

[0058] Among them, the multidimensional feature vector refers to a vector that contains multiple dimensions of user feature information, constructed based on multi-source environmental risk factor data sources, fused data and supplementary data, and is used by deep learning models for risk assessment.

[0059] Deep learning models refer to algorithmic models used in insurance front-end systems to analyze and calculate multi-dimensional feature vectors to derive user risk scores, thereby achieving accurate risk assessment. For example, risk assessment models built on neural networks.

[0060] Among them, the risk score refers to the numerical value calculated by the deep learning model based on multi-dimensional feature vectors, which is used to quantify the degree of insurance risk for users.

[0061] The risk assessment result refers to the final assessment conclusion of the user's insurance risk based on the risk score. It represents a comprehensive judgment of the user's risk and is used to provide insurance advice and decision-making basis for the user. For example, it may determine whether the user's risk level is high, medium, or low.

[0062] This application embodiment utilizes a dynamic form engine to generate personalized interactive forms based on initial user data, avoiding repeated filling of redundant information, greatly simplifying the operation process, and significantly improving the user experience. Key fields are extracted from supporting documents and integrated with the interactive form to obtain fused data, enabling automatic parsing of unstructured data, reducing manual processing, mitigating the risk of missing key information, and improving the accuracy and timeliness of business processing. Supplementary data is obtained by adjusting preset reference questions based on initial data, and a multi-dimensional feature vector of the user is constructed by combining the fused data and multi-source environmental risk factor data, enabling real-time perception of multi-source data. A risk score is calculated based on the multi-dimensional feature vector using a deep learning model, breaking through the limitations of traditional fixed rule bases. It can dynamically adjust the risk assessment logic in real-time by combining external dynamic data, making risk assessment decisions more accurate and improving the accuracy of user risk assessment.

[0063] In some optional implementations of this embodiment, step 202, based on the initial data, is processed by a preset dynamic form engine to generate a personalized interactive form, specifically including the following steps:

[0064] The system uses a pre-defined dynamic form engine to extract features from the initial data to obtain user feature data. Based on the user feature data, it determines the field information related to risk assessment to obtain a field set. The system then uses the dynamic form engine to map the fields in the field set to structured data to generate a personalized interactive form.

[0065] Feature extraction refers to the process of using specific algorithms and rules pre-set in the dynamic form engine to deeply analyze and process the initial data entered by the user on the target interface, and to extract representative and key information elements. For example, when processing health information entered by the user, feature extraction algorithms can identify key features such as whether the user has a specific disease and the severity of the disease.

[0066] Among them, user feature data is a dataset obtained from the initial user data after feature extraction, which can comprehensively and accurately reflect the user's personal characteristics and insurance-related attributes.

[0067] The field information is further determined based on user characteristic data and consists of specific data items directly related to insurance risk assessment. For example, when determining field information related to health risk assessment, it may include disease type, onset time, treatment status, etc.

[0068] The field set is an ordered dataset formed by integrating and classifying the various fields related to risk assessment.

[0069] Structured data, in particular, is data with a clear structure and standard format formed by mapping and transforming each field in a field set according to predetermined rules and formats through a dynamic form engine.

[0070] In one example, in a health insurance application scenario, the technical solution of this embodiment can be used to improve user experience and the accuracy of risk assessment. First, the initial data entered by the user on the application target interface is obtained, such as age 45, occupation as an office worker, and a recent medical examination report. Using a preset dynamic form engine, features are extracted from the initial data. Natural language processing and data analysis algorithms are used to extract user characteristic data from information such as age and occupation, identifying that the user is in a high-risk age group for diseases and may have health risks due to prolonged sitting at work. Next, based on the user characteristic data, fields related to health risk assessment are determined, such as past medical history, family medical history, and exercise frequency, resulting in a field set. Then, using the dynamic form engine, the fields in the field set are mapped to structured data to generate a personalized interactive form. The form specifically asks the user about their past medical history, family history of hereditary diseases, and weekly exercise frequency.

[0071] This application embodiment utilizes a dynamic form engine to extract features from initial data, accurately mining key user characteristic data from massive amounts of initial information. Based on this user characteristic data, risk assessment-related fields are determined and a field set is formed, effectively focusing on core data and avoiding interference from irrelevant information. The dynamic form engine then maps these fields to structured data to generate personalized interactive forms, providing customized forms based on different user characteristics. This reduces redundant user input, improves operational convenience, and facilitates efficient subsequent system processing, enhancing the system's dynamic adaptability to different users.

[0072] In some optional implementations, the supporting documents include images of medical documents and insurance documents. Step 203 involves extracting key fields from the supporting documents, specifically including the following steps:

[0073] Using pre-defined optical character recognition technology, text information is extracted from medical document images and insurance document images respectively to obtain target text sets for medical document images and insurance document images; using pre-defined natural language processing technology, semantic parsing is performed on the target text sets to identify key information and obtain key fields.

[0074] Among them, medical document images represent a visual carrier of users' medical information, used to carry medical document content such as medical records, diagnostic reports, and physical examination reports. For example, in the context of health insurance, users upload medical record images containing information such as past medical history and recent physical examination indicators.

[0075] Among them, insurance document images are generated by image conversion of user-provided documents directly related to insurance business, representing a visual representation of specific information related to insurance business. For example, an image of an insurance claim application form uploaded by a user.

[0076] Among these technologies, optical character recognition (OCR) is used to extract text information from image data such as medical documents and insurance documents. For example, OCR can extract text information such as symptom descriptions and examination values ​​from user-uploaded medical report images.

[0077] The target text set is a collection obtained by extracting text information from medical document images and insurance document images using optical character recognition technology.

[0078] Natural Language Processing (NLP) technology is used to achieve deep understanding and mining of textual information. NLP technology can perform semantic parsing and information extraction on target text sets extracted from medical document images and insurance document images.

[0079] Semantic parsing is an important part of natural language processing technology. It represents the ability to analyze and interpret the semantics of words, sentences, and paragraphs in a target text set, and is used to reveal the true meaning and internal logical relationships expressed in the text.

[0080] Among them, the key information is information that has a significant impact on insurance business, extracted from the target text sets of medical document images and insurance document images through semantic parsing using natural language processing technology.

[0081] In one example, in a health insurance application scenario, a user uploads images of medical documents (such as a recent physical examination report) and insurance documents (such as a previously purchased insurance contract) through a target interface. First, optical character recognition (OCR) technology is used to scan and recognize the medical document images. This technology, through image preprocessing, character segmentation, feature extraction, and character recognition, converts the text information in the physical examination report image into an editable text format. The same processing is applied to the insurance document image, ultimately resulting in a target text set of medical and insurance document images. Next, natural language processing (NLP) technology is used to semantically analyze the target text set. Through lexical analysis, syntactic analysis, and semantic role labeling, key information is identified, yielding key fields. For example, "hypertension" and "hyperglycemia" are identified as key health fields from the physical examination information, and "exclusions for specific diseases" and other key insurance fields are identified from the insurance contract.

[0082] This application's embodiments utilize optical character recognition (OCR) technology to extract text information from medical and insurance document images, forming a target text set. This overcomes the limitations of traditional systems that cannot automatically parse unstructured data, enabling automated text extraction from medical and insurance-related documents. This significantly improves data processing efficiency and reduces omissions and errors prone to occur during manual processing. Furthermore, natural language processing (NLP) technology is used to semantically analyze the target text set to identify key information and obtain critical fields. This allows for precise extraction of the core document content, providing accurate and crucial data support for subsequent business processes.

[0083] In some optional implementations, the step "extracting text information from medical document images and insurance document images respectively using preset optical character recognition technology to obtain target text sets for medical document images and insurance document images" specifically includes the following steps:

[0084] The process involves: acquiring the pixel density of medical and insurance document images; determining the sharpness of the images using a preset resolution detection mechanism and pixel density to obtain a quality score; if the quality score is greater than or equal to a preset threshold, extracting the text information from the images using a preset optical character recognition (OCR) technique to obtain the target text set; if the quality score is lower than the preset threshold, enhancing the images using a preset image enhancement algorithm to obtain an enhanced image; and finally, extracting the text content from the enhanced image using text extraction and edge detection techniques to obtain the target text set.

[0085] Pixel density refers to the number of pixels contained in a unit area of ​​medical document images and insurance document images. The higher the value, the more densely the pixels are arranged in the image, and the richer the details are displayed.

[0086] The resolution detection mechanism is a system based on specific algorithms and rules used to detect the resolution of medical document images and insurance document images.

[0087] The quality score is obtained by using a resolution detection mechanism and pixel density to determine the clarity of medical document images and insurance document images.

[0088] The preset threshold is a pre-set standard value that can be derived from the analysis of a large amount of image processing experience and actual business needs. It represents the boundary for judging whether the image quality is acceptable.

[0089] Image enhancement algorithms are a collection of mathematical operations and processing methods used to improve image quality. For example, histogram equalization algorithms.

[0090] Among them, enhancement processing refers to the operation process of improving and optimizing poor-quality images by using image enhancement algorithms.

[0091] Image enhancement refers to the image obtained after processing with image enhancement algorithms. For example, after enhancement processing, the text edges of an originally blurry medical report image become clearer, and the overall image contrast is improved.

[0092] Among them, text extraction methods are a set of technical means and processes for extracting text information from images. They represent the specific ways to convert text in an image into editable text, such as text extraction methods based on connected component analysis.

[0093] Edge detection technology is an image processing technique used to identify the edges of objects in an image. It is based on the variation characteristics of the gray values ​​of image pixels and uses a specific algorithm to detect the boundaries between different objects in the image, thus representing a method for determining the contours and boundaries of objects in an image.

[0094] In one example, in a health insurance underwriting scenario, a user uploads initial data, such as basic personal information and insurance intentions, through the target interface of the insurance business front end. Simultaneously, they upload images of medical documents (such as physical examination reports) and insurance documents (such as previous policies). The system first obtains the pixel density of the medical and insurance document images; for example, a physical examination report image contains 5000 pixels per square centimeter. Next, a resolution detection mechanism combined with pixel density is used to determine image clarity, resulting in a quality score of 75. The preset threshold is 80. Since 75 is below the preset threshold, the system uses a preset image enhancement algorithm, such as histogram equalization, to enhance the image, making the originally dark and low-contrast image clearer, resulting in an enhanced image. Then, a text extraction method based on connected component analysis and Canny edge detection technology are used to accurately extract text content from the enhanced image, obtaining the target text set. If the quality score is greater than or equal to 80, the text information is directly extracted using preset optical character recognition technology, obtaining the target text set.

[0095] This application embodiment obtains the pixel density of medical document images and insurance document images, and uses a resolution detection mechanism to derive a quality score, which can accurately determine the image clarity. If the quality score meets the standard, optical character recognition technology is directly used to extract text information, which is efficient and fast. If the quality score does not meet the standard, an image enhancement algorithm is used to obtain an enhanced image, and then text extraction methods and edge detection technology are used to extract text, which can effectively solve the problem of the traditional method's weak ability to process unstructured data. This solution avoids the tediousness and error-proneness of manual image processing.

[0096] In some optional implementations, step 203 involves matching and merging the interactive form with key fields to generate merged data, specifically including the following steps:

[0097] A preset text similarity algorithm is used to match key fields with form fields of interactive forms to obtain a matching result set. Based on the matching result set, a preset weighted fusion method is used to integrate the form fields and key fields to generate fused data.

[0098] Text similarity algorithms are methods used to measure the semantic similarity between different texts. They are used to calculate the similarity between key fields extracted from supporting documents and form fields in interactive forms, determining whether they match semantically. For example, a cosine similarity algorithm can be used to convert key fields and form fields into vector forms, and the similarity is measured by calculating the cosine of the angle between the vectors, thus providing a basis for subsequent field matching.

[0099] The matching result set is a collection of matching results between key fields and form fields obtained after text similarity algorithm calculation.

[0100] Weighted fusion is a technique that comprehensively considers the importance of different data sources or data items and merges them. It is used to integrate form fields and key fields according to pre-defined weighting rules based on the matching result set, thereby fully leveraging the information value of different fields and improving data accuracy and completeness.

[0101] In one example, during a health insurance application process, the user's uploaded medical documentation includes the key field "suffering from mild hypertension," and the interactive form has a "health status" field. Using a pre-defined cosine similarity text similarity algorithm, the key field and form field are converted into vectors to calculate their similarity, resulting in a matching set (e.g., a similarity of 0.8). Based on this set, a pre-defined weighted fusion method is used, assigning a higher weight to "hypertension" due to its significant impact on health insurance risk, thus integrating the form field and key field. This accurately integrates user health information, avoiding errors and omissions from manual entry, improving data processing accuracy and efficiency, and providing reliable data for subsequent risk assessment.

[0102] This application embodiment uses a text similarity algorithm to match key fields with form fields in an interactive form, accurately locating the relationship between the two and obtaining an accurate matching result set. Then, based on a preset weighted fusion method, the form fields and key fields are scientifically integrated to generate fused data. This process achieves effective fusion of unstructured and structured data, improves data processing efficiency, and avoids the problem of information omissions during manual processing.

[0103] In some optional implementations, step 204, based on the initial data, adjusts the preset reference problem to obtain the user's adaptation problem, specifically including the following steps:

[0104] Based on the field set, a pre-defined knowledge graph is used to query the user's corresponding health risk indicators and safety assessment related factor indicators to identify risk points. Based on the field set and risk points, a question sequence is generated through a pre-defined rule engine to obtain a preliminary question list. Based on the preliminary question list, the pre-defined reference questions are adjusted to obtain the user's appropriate questions.

[0105] Knowledge graphs are pre-built structured semantic knowledge bases derived from in-depth mining and integrated analysis of massive amounts of insurance business data, medical expertise, industry research reports, and other multi-source data. They represent various concepts in the insurance field and their interrelationships through elements such as entities, relationships, and attributes. For example, in the health insurance scenario, a knowledge graph contains entities such as basic user information (e.g., age, gender, occupation), health status (e.g., medical history, physical examination indicators), and insurance product information (e.g., coverage, claims conditions), as well as relationships between them such as "relationship between age and disease incidence" and "association between occupation and occupational disease."

[0106] Among them, health risk indicators are a series of quantitative or qualitative parameters used in insurance front-end development to measure the degree to which a user's health status affects insurance risk.

[0107] Among them, the security assessment related factor indicators are a series of factors closely related to user security status and insurance business security assessment during the insurance front-end development process.

[0108] In this context, risk points are specific aspects or links that may increase insurance risk, identified in the insurance front-end development technical solution based on initial user input data and health risk indicators and safety assessment correlation factors obtained through knowledge graph queries. For example, in a health insurance scenario, if a knowledge graph query reveals that a user's age falls within a high-risk age group for a certain disease, their blood pressure is high, and their occupation involves prolonged sitting in an office, then "cardiovascular disease risk" might be identified as a risk point.

[0109] Among them, the rule engine is the core component in the insurance front-end development technology solution used to automate the processing of business rules.

[0110] The question sequence refers to a series of related questions generated by a pre-defined rule engine in a certain logical order based on the user's initial data field set and identified risk points. For example, in a health insurance scenario, if the system determines that a user has a risk of cardiovascular disease, the rule engine can generate the following question sequence: First, ask "Do you have a family history of cardiovascular disease?" If the user answers yes, then ask "What specific type of cardiovascular disease is in your family?", and then ask "Is your daily diet high in salt and fat?", etc.

[0111] The preliminary question list, generated by the rule engine in the insurance front-end development technical solution, is a collection of multiple initially proposed questions after the rule engine produces a sequence of questions based on the user's initial data field set and identified risk points. For example, in the health insurance example of identifying cardiovascular disease risk points, the preliminary question list formed after organizing the question sequence generated by the rule engine might include questions such as "Is there a family history of cardiovascular disease?", "What types of cardiovascular diseases are there in the family?", "Is the daily diet high in salt and fat?", and "How often do you exercise?".

[0112] In one example, during a health insurance underwriting scenario, user Xiao Li inputs initial data on the target interface, including his age (35 years old), occupation (programmer), and uploads a recent medical examination report as proof. After obtaining the initial data's field set, the system uses a pre-defined knowledge graph for querying. The knowledge graph stores a large amount of data related to medical and insurance businesses. Based on Xiao Li's age and occupation, it retrieves corresponding health risk indicators such as potential risk of cervical spondylosis (prone to prolonged sitting in a programmer's profession) and cardiovascular disease risk (the incidence of this disease is at a certain level in the 35-year-old population), as well as safety assessment related factors such as work stress (the high workload of programmers may affect their health). The system comprehensively identifies "potential risk of cervical spondylosis" and "cardiovascular disease risk" as risk points. Based on the field set and risk points, a pre-defined rule engine generates a sequence of questions, obtaining a preliminary question list, such as "Do you have a family history of cardiovascular disease?" and "How long do you spend working at a desk?". Next, based on the preliminary question list, the system adjusts the pre-defined reference questions, considering Xiao Li's programmer profession, and refines the question to more closely match his work scenario: "Do you work continuously in front of a computer for more than 4 hours a day?". After displaying compatibility issues on the target interface, supplementary data is retrieved. Through this process, the system can accurately pinpoint risks, obtain targeted information, improve the accuracy of risk assessment, optimize user experience, and meet personalized needs.

[0113] This application embodiment identifies risk points by using a knowledge graph to query user health risk indicators and safety assessment correlation factors based on the initial data's field set. This accurately pinpoints potential user risks and provides crucial evidence for subsequent assessments. Based on the field set and risk points, a preliminary question list is generated by a rule engine and further adjusted to obtain tailored questions, making the questions more targeted and personalized. This process not only reduces the need for users to repeatedly input redundant information, improving user experience, but also dynamically generates tailored questions based on the user's actual situation, enhancing the system's dynamic adaptability, effectively meeting the personalized needs of different users, and improving the accuracy and comprehensiveness of risk assessment.

[0114] In some optional implementations, step S206 involves constructing a multidimensional feature vector for the user based on multi-source environmental risk factor data, fused data, and supplementary data. The multi-source environmental risk factor data reflects the environmental risk status and specifically includes the following steps:

[0115] The system removes duplicates and missing values ​​from multi-source environmental risk factor data, fused data, and supplementary data using preset data cleaning rules, resulting in a cleaned dataset. A preset field mapping mechanism is then used to extract key fields from the cleaned dataset, resulting in a structured field set. A preset principal component analysis algorithm is then employed to extract features from the structured field set, resulting in a dimensionality-reduced feature set. Finally, a multidimensional feature vector for the user is constructed based on the dimensionality-reduced feature set.

[0116] In one example, in a health insurance application scenario, the user enters initial data on the target interface, such as age 35 and occupation as an office worker, and uploads a medical report as supporting documentation. The system acquires multi-source environmental risk factor data, such as recent disease prevalence data. First, according to preset data cleaning rules, it removes multi-source environmental risk factor data, fused data after integrating the initial data and key fields from the medical report, and duplicates from subsequently acquired supplementary data, such as duplicate disease diagnosis records. Missing values ​​are filled, such as missing examination dates in the medical report, resulting in a cleaned dataset. Next, through a preset field mapping mechanism, key fields, such as age, occupation, disease type, and incidence frequency, are extracted from the cleaned dataset to form a structured field set. Then, a preset principal component analysis algorithm is used to extract features from the structured field set, removing redundant information to obtain a dimensionality-reduced feature set. Finally, a multi-dimensional feature vector of the user is constructed based on the dimensionality-reduced feature set. This solution can accurately integrate multi-source data, remove invalid information, construct a comprehensive and accurate user feature model, and improve the accuracy of health insurance risk assessment.

[0117] In some optional implementations, after step 207, which calculates the user's risk score based on a deep learning model using a multi-dimensional feature vector to obtain the user's risk assessment result, the following steps are also included:

[0118] Based on the risk assessment results, insurance recommendations are generated for users, including premium estimates.

[0119] Among them, the insurance recommendation is a guide to insurance purchases provided to users based on the results of risk assessment.

[0120] Among them, the premium estimate is used to provide users with a general range of insurance costs, helping them understand the possible financial expenses before purchasing insurance.

[0121] In one example, in a car insurance scenario, a deep learning model determines a user's risk assessment result as high-risk. Based on this result, an insurance recommendation is generated. Regarding premium estimation, due to the high risk, historical data and market conditions are referenced, and a specific premium calculation model is used to estimate the user's annual premium at 8,000 yuan, 30% higher than that of low-risk users. This premium estimate combines the user's risk profile with reasonable pricing. Simultaneously, a risk warning is provided, indicating that high-risk individuals may face higher claims thresholds and exclusion of certain high-risk liabilities.

[0122] This application's embodiments can reasonably estimate insurance premiums based on a user's accurate risk profile using scientific algorithms, providing users with a clear and intuitive reference for insurance costs. This not only allows users to have a clear understanding of insurance costs but also, combined with risk warnings, helps users fully understand their own risks, make insurance decisions that better suit their needs, and improve the user experience.

[0123] It should be emphasized that, in order to further ensure the privacy and security of the aforementioned initial data, supporting documents, interactive forms, integrated data, adaptation issues, supplementary data, multi-source environmental risk factor data, multi-dimensional feature vectors, and risk assessment results, these data can also be stored in a node of a blockchain.

[0124] The blockchain referred to in this application is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Essentially, a blockchain is a decentralized database, a chain of data blocks linked together using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain can include an underlying blockchain platform, a platform product service layer, and an application service layer.

[0125] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0126] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0127] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing related hardware with computer-readable instructions. These computer-readable instructions can be stored in a computer-readable storage medium. When executed, the program can include the processes of the embodiments of the above methods. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, optical disk, or read-only memory (ROM), or random access memory (RAM).

[0128] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0129] Further reference Figure 3 As a response to the above Figure 2 To implement the method shown, this application provides an embodiment of a risk assessment device based on multi-source data analysis. This device embodiment is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0130] like Figure 3 As shown, the risk assessment device 400 based on multi-source data analysis in this embodiment includes: an acquisition module 401, a processing module 402, an extraction module 403, an adjustment module 404, a display module 405, a construction module 406, and a calculation module 407. Wherein:

[0131] Module 401 is used to acquire the initial data and supporting documents entered by the user on the target interface;

[0132] The processing module 402 is used to process the initial data using a preset dynamic form engine to generate a personalized interactive form.

[0133] Extraction module 403 is used to extract key fields from the proof document and match and merge the interactive form with the key fields to generate merged data;

[0134] The adjustment module 404 is used to adjust the preset reference problem based on the initial data to obtain the user's adaptation problem;

[0135] Display module 405 is used to display adaptation issues on the target interface in order to obtain supplementary data input by the user regarding the adaptation issues;

[0136] Module 406 is used to acquire multi-source environmental risk factor data. Based on the multi-source environmental risk factor data, fused data and supplementary data, a multi-dimensional feature vector of the user is constructed. The multi-source environmental risk factor data is data that reflects the environmental risk status.

[0137] The calculation module 407 is used to calculate the user's risk score based on a multi-dimensional feature vector and a deep learning model, thereby obtaining the user's risk assessment result.

[0138] This embodiment utilizes a dynamic form engine to generate personalized interactive forms based on initial user data, avoiding repeated filling of redundant information, greatly simplifying the operation process, and significantly improving the user experience. Key fields are extracted from supporting documents and integrated with the interactive form to obtain fused data, enabling automatic parsing of unstructured data, reducing manual processing, mitigating the risk of missing key information, and improving the accuracy and timeliness of business processing. Supplementary data is obtained by adjusting preset reference questions based on initial data, and a multi-dimensional feature vector of the user is constructed by combining the fused data and multi-source environmental risk factor data, enabling real-time perception of multi-source data. A risk score is calculated based on the multi-dimensional feature vector using a deep learning model, breaking through the limitations of traditional fixed rule bases. It can dynamically adjust the risk assessment logic in real-time by combining external dynamic data, making risk assessment decisions more accurate and improving the accuracy of user risk assessment.

[0139] In one embodiment, the processing module 402 includes:

[0140] The extraction submodule is used to extract features from the initial data using a preset dynamic form engine to obtain user feature data.

[0141] The determination submodule is used to determine the field information related to risk assessment based on user characteristic data, and obtain the field set;

[0142] The mapping submodule is used to map fields in a field set to structured data through a dynamic form engine, generating personalized interactive forms.

[0143] This application embodiment utilizes a dynamic form engine to extract features from initial data, accurately mining key user characteristic data from massive amounts of initial information. Based on this user characteristic data, risk assessment-related fields are determined and a field set is formed, effectively focusing on core data and avoiding interference from irrelevant information. The dynamic form engine then maps these fields to structured data to generate personalized interactive forms, providing customized forms based on different user characteristics. This reduces redundant user input, improves operational convenience, and facilitates efficient subsequent system processing, enhancing the system's dynamic adaptability to different users.

[0144] In one embodiment, the adjustment module 404 includes:

[0145] The query submodule is used to query the user's corresponding health risk indicators and safety assessment related factor indicators based on the field set and a preset knowledge graph to determine risk points;

[0146] The issue generation submodule is used to generate an issue sequence based on the field set and risk points, and obtain a preliminary issue list through a preset rule engine;

[0147] The adjustment submodule is used to adjust the preset reference questions based on the initial question list to obtain the user's suitable questions.

[0148] This application embodiment identifies risk points by using a knowledge graph to query user health risk indicators and safety assessment correlation factors based on the initial data's field set. This accurately pinpoints potential user risks and provides crucial evidence for subsequent assessments. Based on the field set and risk points, a preliminary question list is generated by a rule engine and further adjusted to obtain tailored questions, making the questions more targeted and personalized. This process not only reduces the need for users to repeatedly input redundant information, improving user experience, but also dynamically generates tailored questions based on the user's actual situation, enhancing the system's dynamic adaptability, effectively meeting the personalized needs of different users, and improving the accuracy and comprehensiveness of risk assessment.

[0149] In one embodiment, the extraction module 403 includes:

[0150] The extraction submodule is used to extract text information from medical document images and insurance document images respectively using preset optical character recognition technology, so as to obtain the target text sets of medical document images and insurance document images;

[0151] The parsing submodule is used to perform semantic parsing on the target text set using preset natural language processing techniques, identify key information, and obtain key fields.

[0152] This application's embodiments utilize optical character recognition (OCR) technology to extract text information from medical and insurance document images, forming a target text set. This overcomes the limitations of traditional systems that cannot automatically parse unstructured data, enabling automated text extraction from medical and insurance-related documents. This significantly improves data processing efficiency and reduces omissions and errors prone to occur during manual processing. Furthermore, natural language processing (NLP) technology is used to semantically analyze the target text set to identify key information and obtain critical fields. This allows for precise extraction of the core document content, providing accurate and crucial data support for subsequent business processes.

[0153] In one embodiment, the extraction submodule is further configured to obtain the pixel density of the medical document image and the insurance document image; determine the clarity of the medical document image and the insurance document image by using a preset resolution detection mechanism and pixel density, and obtain a quality score; if the quality score is greater than or equal to a preset threshold, extract the text information of the medical document image and the insurance document image by using a preset optical character recognition technology, and obtain the target text set of the medical document image and the insurance document image; if the quality score is lower than the preset threshold, use a preset image enhancement algorithm to enhance the medical document image and the insurance document image, and obtain an enhanced image; use a text extraction method and edge detection technology to extract the text content of the enhanced image, and obtain the target text set of the medical document image and the insurance document image.

[0154] This application embodiment obtains the pixel density of medical document images and insurance document images, and uses a resolution detection mechanism to derive a quality score, which can accurately determine the image clarity. If the quality score meets the standard, optical character recognition technology is directly used to extract text information, which is efficient and fast. If the quality score does not meet the standard, an image enhancement algorithm is used to obtain an enhanced image, and then text extraction methods and edge detection technology are used to extract text, which can effectively solve the problem of the traditional method's weak ability to process unstructured data. This solution avoids the tediousness and error-proneness of manual image processing.

[0155] In one embodiment, the extraction module 403 includes:

[0156] The matching submodule is used to match key fields with form fields of interactive forms using a preset text similarity algorithm to obtain a matching result set;

[0157] The integration submodule is used to integrate form fields and key fields based on the matching result set using a preset weighted fusion method to generate integrated data.

[0158] This application embodiment uses a text similarity algorithm to match key fields with form fields in an interactive form, accurately locating the relationship between the two and obtaining an accurate matching result set. Then, based on a preset weighted fusion method, the form fields and key fields are scientifically integrated to generate fused data. This process achieves effective fusion of unstructured and structured data, improves data processing efficiency, and avoids the problem of information omissions during manual processing.

[0159] In one embodiment, the risk assessment device 400 based on multi-source data analysis further includes:

[0160] It is recommended to generate a submodule to provide insurance recommendations for users based on the risk assessment results. The insurance recommendations include a premium estimate.

[0161] This application's embodiments can reasonably estimate insurance premiums based on a user's accurate risk profile using scientific algorithms, providing users with a clear and intuitive reference for insurance costs. This not only allows users to have a clear understanding of insurance costs but also, combined with risk warnings, helps users fully understand their own risks, make insurance decisions that better suit their needs, and improve the user experience.

[0162] To address the aforementioned technical problems, embodiments of this application also provide a computer device. Please refer to [link / reference needed]. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0163] Computer device 6 includes a memory 61, a processor 62, and a network interface 63 that are interconnected via a system bus. It should be noted that only computer device 6 with memory 61, processor 62, and network interface 63 is shown in the figure; however, it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described herein is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0164] Computer devices can include desktop computers, laptops, handheld computers, and cloud servers. These devices allow for human-computer interaction with users through keyboards, mice, remote controls, touchpads, or voice-activated devices.

[0165] The memory 61 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 61 may be an internal storage unit of the computer device 6, such as the hard disk or memory of the computer device 6. In other embodiments, the memory 61 may also be an external storage device of the computer device 6, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 6. Of course, the memory 61 may also include both the internal storage unit and the external storage device of the computer device 6. In this embodiment, the memory 61 is typically used to store the operating system and various application software installed on the computer device 6, such as computer-readable instructions based on risk assessment methods of multi-source data analysis. In addition, memory 61 can also be used to temporarily store various types of data that have been output or will be output.

[0166] In some embodiments, processor 62 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip. This processor 62 is typically used to control the overall operation of computer device 6. In this embodiment, processor 62 is used to execute computer-readable instructions stored in memory 61 or to process data, such as executing computer-readable instructions for a risk assessment method based on multi-source data analysis.

[0167] The network interface 63 may include a wireless network interface or a wired network interface, which is typically used to establish a communication connection between the computer device 6 and other electronic devices.

[0168] This application embodiment utilizes a dynamic form engine to generate personalized interactive forms based on initial user data, avoiding repeated filling of redundant information, greatly simplifying the operation process, and significantly improving the user experience. Key fields are extracted from supporting documents and integrated with the interactive form to obtain fused data, enabling automatic parsing of unstructured data, reducing manual processing, mitigating the risk of missing key information, and improving the accuracy and timeliness of business processing. Supplementary data is obtained by adjusting preset reference questions based on initial data, and a multi-dimensional feature vector of the user is constructed by combining the fused data and multi-source environmental risk factor data, enabling real-time perception of multi-source data. A risk score is calculated based on the multi-dimensional feature vector using a deep learning model, breaking through the limitations of traditional fixed rule bases. It can dynamically adjust the risk assessment logic in real-time by combining external dynamic data, making risk assessment decisions more accurate and improving the accuracy of user risk assessment.

[0169] This application also provides another embodiment, namely, providing a computer-readable storage medium storing computer-readable instructions that can be executed by at least one processor to cause the at least one processor to perform the steps of the risk assessment method based on multi-source data analysis as described above.

[0170] This application embodiment utilizes a dynamic form engine to generate personalized interactive forms based on initial user data, avoiding repeated filling of redundant information, greatly simplifying the operation process, and significantly improving the user experience. Key fields are extracted from supporting documents and integrated with the interactive form to obtain fused data, enabling automatic parsing of unstructured data, reducing manual processing, mitigating the risk of missing key information, and improving the accuracy and timeliness of business processing. Supplementary data is obtained by adjusting preset reference questions based on initial data, and a multi-dimensional feature vector of the user is constructed by combining the fused data and multi-source environmental risk factor data, enabling real-time perception of multi-source data. A risk score is calculated based on the multi-dimensional feature vector using a deep learning model, breaking through the limitations of traditional fixed rule bases. It can dynamically adjust the risk assessment logic in real-time by combining external dynamic data, making risk assessment decisions more accurate and improving the accuracy of user risk assessment.

[0171] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods of the various embodiments of this application.

[0172] Obviously, the embodiments described above are only some embodiments of this application, not all embodiments. The accompanying drawings show preferred embodiments of this application, but do not limit the patent scope of this application. This application can be implemented in many different forms; rather, the purpose of providing these embodiments is to provide a more thorough and comprehensive understanding of the disclosure of this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this application's specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the scope of patent protection of this application.

[0173] The software tools or components not belonging to our company that appear in the embodiments of this application are merely examples and do not represent actual use.

Claims

1. A risk assessment method based on multi-source data analysis, characterized in that, The method comprises the following steps: obtaining initial data and proof files input by a user on a target interface; processing the initial data through a preset dynamic form engine to generate an individualized interactive form; extracting key fields from the proof files and matching and fusing the interactive form with the key fields to generate fused data; adjusting a preset reference question based on the initial data to obtain an adapted question for the user; displaying the adapted question on the target interface to obtain supplementary data input by the user in response to the adapted question; obtaining multi-source environmental risk factor data, and constructing a multi-dimensional feature vector of the user based on the multi-source environmental risk factor data, the fused data and the supplementary data, wherein the multi-source environmental risk factor data is data reflecting environmental risk conditions; based on the multi-dimensional feature vector, using a deep learning model to calculate a risk score of the user to obtain a risk assessment result of the user.

2. The method of claim 1, wherein, The step of processing the initial data through a preset dynamic form engine to generate an individualized interactive form comprises: extracting features from the initial data through the preset dynamic form engine to obtain user feature data; determining field information related to risk assessment based on the user feature data to obtain a field set; mapping fields in the field set to structured data through the dynamic form engine to generate an individualized interactive form.

3. The method of claim 2, wherein, The step of adjusting a preset reference question based on the initial data to obtain an adapted question for the user comprises: querying health risk indicators and safety assessment associated factor indicators corresponding to the user based on the field set using a preset knowledge graph to determine a risk point; generating a question sequence based on the field set and the risk point through a preset rule engine to obtain a preliminary question list; adjusting a preset reference question based on the preliminary question list to obtain an adapted question for the user.

4. The method of claim 1, wherein, The proof files comprise medical document images and insurance document images, and the step of extracting key fields from the proof files comprises: extracting text information from the medical document images and the insurance document images respectively using a preset optical character recognition technology to obtain target text sets of the medical document images and the insurance document images; performing semantic analysis on the target text sets using a preset natural language processing technology to identify key information to obtain key fields.

5. The method of claim 4, wherein, The step of extracting text information from the medical document images and the insurance document images respectively using a preset optical character recognition technology to obtain target text sets of the medical document images and the insurance document images comprises: obtaining pixel densities of the medical document images and the insurance document images; judging the clarity of the medical document images and the insurance document images through a preset resolution detection mechanism and the pixel densities to obtain quality scores; If the mass fraction is greater than or equal to a preset threshold, text information of the medical document image and the insurance document image is extracted through a preset optical character recognition technology, and a target text set of the medical document image and the insurance document image is obtained; If the mass fraction is less than the preset threshold, the medical document image and the insurance document image are enhanced through a preset image enhancement algorithm, and an enhanced image is obtained; Text content of the enhanced image is extracted through a text extraction method and an edge detection technology, and a target text set of the medical document image and the insurance document image is obtained.

6. The method of claim 1, wherein, The step of matching and fusing the interactive form with the key field to generate fused data specifically includes: The key field and the form field of the interactive form are matched through a preset text similarity algorithm, and a matching result set is obtained; According to the matching result set, the form field and the key field are integrated through a preset weighted fusion method to generate fused data.

7. The method of claim 1, wherein, After the step of calculating the risk score of the user based on the multi-dimensional feature vector through a deep learning model to obtain the risk assessment result of the user, the method further includes: According to the risk assessment result, an insurance application suggestion for the user is generated, and the insurance application suggestion includes premium estimation.

8. A risk assessment device based on multi-source data analysis, characterized in that, It includes: An acquisition module is configured to acquire initial data and proof files input by a user on a target interface; A processing module is configured to generate a personalized interactive form by processing the initial data through a preset dynamic form engine; An extraction module is configured to extract key fields from the proof files and match and fuse the interactive form with the key fields to generate fused data; An adjustment module is configured to adjust a preset reference question based on the initial data to obtain an adapted question for the user; A display module is configured to display the adapted question on the target interface to acquire supplementary data input by the user in response to the adapted question; A construction module is configured to acquire multi-source environmental risk factor data, and construct a multi-dimensional feature vector of the user based on the multi-source environmental risk factor data, the fused data, and the supplementary data, wherein the multi-source environmental risk factor data is data reflecting environmental risk conditions; A calculation module is configured to calculate a risk score of the user based on the multi-dimensional feature vector through a deep learning model to obtain a risk assessment result of the user.

9. A computer device, comprising: The computer readable storage medium stores computer readable instructions, and the computer readable instructions are executed by the processor to realize the steps of the risk assessment method based on multi-source data analysis according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer readable instructions, and the computer readable instructions are executed by the processor to realize the steps of the risk assessment method based on multi-source data analysis according to any one of claims 1 to 7.

Images