Insurance industry customer information analysis method and device and storage medium
Through the combination of logistic regression and deep matching model, and BERT model, customer information analysis in the insurance industry is solved, and the traditional method has been solved in terms of semantic connection and speed, and the customer information analysis method with high accuracy and fast matching is achieved, and the ability to learn is self-learning. .
Patent Information
- Application Number
- CN202411438745.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-15
- Publication Date
- 2025-01-17
AI Technical Summary
In the analysis of customer information in the insurance industry, traditional methods focus on the meaning of the words themselves and ignore semantic connections, resulting in low accuracy and slow language model processing speed, which is difficult to meet the complex needs of insurance analysts.
Using a method combining logistic regression model and deep matching model, we establish the 'scene set' and 'result set' based on historical data analysis logs, and further match the logistic regression model through preliminary matching and deep matching model, and combine the BERT model for text feature extraction. Generate final matching scenarios and analysis results.
It has achieved the improvement of the accuracy and speed of customer information analysis in the insurance industry, solved the problem of fast logistic regression model speed but low accuracy, high accuracy but slow speed of deep matching model, and has the ability to learn and increase the number of analysis times. Continuously improve matching accuracy.
Smart Images
Figure CN119323433A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of text analysis, and in particular to a method, device and storage medium for analyzing customer information in the insurance industry. Background Art
[0002] In order to provide better services to customers, the insurance industry needs to analyze customer data. At present, most of the customer information analysis in the insurance industry is carried out by first manually or machine-based retrieval, and then subsequent analysis to form conclusions. In manual analysis, the scope of customer information retrieval is wide, the number of landings is high, the professional quality and ethics of data operators are high, and information security is difficult to guarantee. Traditional automatic matching methods include matching based on TF-IDF keywords, vectorization methods based on LSA decomposition matrices, and classification and matching of texts based on LDA topic models. These methods only focus on the meaning conveyed by the words themselves, and do not take into account the semantic connection between words and the complex semantic structure of the text itself caused by the use of natural language by insurance analysts in the insurance industry, resulting in low accuracy. The simple use of language models for matching, due to the large amount of data processed, leads to a relatively slow matching speed of language models. Summary of the invention
[0003] In order to solve the above technical problem or at least partially solve the above technical problem, the present invention provides a method, device and storage medium for analyzing customer information in the insurance industry.
[0004] In a first aspect, the present invention provides a method for analyzing customer information in the insurance industry, comprising:
[0005] Establish "scenario sets" and "result sets" based on historical data analysis logs;
[0006] Use the logistic regression model to perform a preliminary match between all scenarios in the "scenario set" and the input analysis requirements, and retain the top n scenarios with the highest matching degree;
[0007] The preliminary matching results are input into the deep matching model for further matching, and k final matching scenes are output from high to low according to the matching probability;
[0008] After obtaining the final matching scenario, based on the association between the results and the scenario, a pre-packaged data query statement is used to generate k analysis results of the "result set" corresponding to the final matching scenario.
[0009] Furthermore, based on the different uses of customer information in historical data analysis logs, different types of data application "scenario sets" and data analysis "result sets" are obtained; each "scenario" in the "scenario set" corresponds to a type of data usage requirement, and each "result" in the "result set" corresponds to a single sensitive data statistical analysis result or a collection of multiple analysis results.
[0010] Furthermore, after the "scenario" is constructed, code is written in the database to extract the corresponding data and perform statistical analysis to obtain the statistical results of the corresponding scenario. Each statistical result corresponds to a result in the "result set", thereby establishing an association between the result and the scenario.
[0011] Furthermore, the logistic regression model is a probability distribution model from scenes to analysis requirements based on a fully connected network. The greater the probability that the input scene is mapped to the analysis requirement through the logistic regression model, the more the scene matches the analysis requirement; the smaller the probability that the input scene is mapped to the analysis requirement through the logistic regression model, the less the scene matches the analysis requirement.
[0012] Furthermore, when training the logistic regression model, the following loss function is maximized:
[0013]
[0014] d is any scenario. For analysis requirement l, scenario d either matches analysis requirement l or does not match analysis requirement l. Indicates m that matches the analysis requirement l + Any one of the scenes, Indicates that m does not match the analysis requirement l - Any one of the scenes, Scenario-based model for logistic regression Given the probability of analyzing the demand l, Scenario-based model for logistic regression Give the probability of analyzing the demand l.
[0015] Furthermore, the deep matching model includes a text feature extractor built using a pre-trained BERT model, which extracts key features of insurance information in natural language;
[0016] Encode the analysis requirements proposed by insurance analysts and the scenarios in the "scenario set" into text vectors that can be recognized by the BERT model;
[0017] The vectors representing the analysis requirements proposed by the insurance analysts and the scenarios in the "scenario set" are combined with the position vector and input into the BERT model to form the corresponding matching vector; the matching vector is then input into the matching fully connected network, which predicts the matching probability of the matching vector and outputs the final k matching scenarios in descending order of probability.
[0018] Furthermore, when training the matching fully connected network, the following loss function value is maximized:
[0019]
[0020] vec d is the matching vector of any scene, vec l is the matching vector of the analysis requirement and the matching vector of the scenario vec d or a matching vector with the analysis requirements vec l matches, or matches the matching vector required by the analysis vec l does not match, Represents the matching vector with the analysis requirements vec l matches m + Any matching vector in a scene, Represents the matching vector with the analysis requirements vec l does not match m - Any matching vector in a scene, To match the scene-based matching vector of the fully connected network Give the matching vector of the analysis requirements vec The probability of l, To match the scene-based matching vector of the fully connected network Given a matching vector vec The probability of l.
[0021] Furthermore, during the actual operation of the platform, the insurance analyst inputs an analysis requirement, the algorithm system generates recommended analysis results, and the analyst selects the required analysis results; among them, a certain proportion of the analysis results in the "result set" that are not within the display recommendation range are randomly selected from the matching analysis results and displayed to the user; the supervision data is constructed through the insurance analyst's selection results of the analysis results, and the supervision data is used to dynamically update the parameters of the model.
[0022] In a second aspect, the present invention provides an insurance industry customer information analysis device, comprising: at least one processing unit, the processing unit and a storage unit are interconnected via a bus unit, the storage unit stores a computer program, and when the computer program is executed by the processing unit, the insurance industry customer information analysis method is implemented.
[0023] In a third aspect, the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the insurance industry customer information analysis method is implemented.
[0024] The above technical solution provided by the embodiment of the present invention has the following advantages compared with the prior art:
[0025] This application establishes a "scenario set" and a "result set" based on historical data analysis logs; uses a logistic regression model to perform a preliminary match between all scenarios in the "scenario set" and the input analysis requirements, and retains the top n scenarios with the highest matching degree; inputs the preliminary matching results into a deep matching model for further matching, and outputs k final matching scenarios from high to low according to the probability of matching; after obtaining the final matching scenario, based on the association between the results and the scenarios, uses a pre-packaged data query statement to generate k analysis results of the "result set" corresponding to the final matching scenario. The present invention uses a customer information analysis method for the insurance industry that combines logistic regression and deep matching to accurately understand the needs of insurance analysts in the insurance industry, select scenarios from the "scenario set", and output results from the "result set". The algorithm simultaneously solves the problem that the logistic regression matching model is relatively fast but the accuracy is not very high, and the language model has a relatively high accuracy but a slow matching speed, achieving a balance between speed and accuracy. This application also has self-learning capabilities, and can continuously improve matching accuracy as the number of analyses increases. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.
[0027] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.
[0028] Figure 1 A flowchart of a method for analyzing customer information in the insurance industry provided by an embodiment of the present invention;
[0029] Figure 2 A schematic diagram of a depth matching model provided by an embodiment of the present invention;
[0030] Figure 3 A schematic diagram of a customer information analysis device for the insurance industry provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0031] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0032] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device including the element.
[0033] Example 1
[0034] like Figure 1 As shown, the technology of the present invention implements a method for analyzing customer information in the insurance industry, the steps comprising:
[0035] There are three outstanding features of customer information analysis in the insurance industry: First, the process of customer information analysis in the insurance industry is recordable and traceable. According to supervision and company regulations, the operations of insurance analysts involving customer information are recorded in the system with data analysis logs, which makes it possible to record, trace and quantify these operations. Second, the analyzed customer information in the insurance industry is easy to classify. Customer information analysis is mostly data such as preservation type and proportion of new insurance policies. The content of analysis is similar by year, quarter, month and week, and there are many repeated analyses in different periods, which are formatted in a certain way, making it possible for content modeling. Third, the correlation between the analysis results and specific customer information is low. The analysis content is mostly reflected in the form of quantity, percentage, etc., and there are few cases involving specific information of a specific customer or a part of a customer, making it possible for specific customer information not to be output or landed.
[0036] Based on the above three characteristics, the present invention distinguishes different uses of customer information in historical data analysis logs, and obtains different types of data application "scenario sets" and data analysis "result sets". Each "scenario" in the "scenario set" corresponds to a type of data usage demand, and each "result" in the "result set" corresponds to a single sensitive data statistical analysis result or a collection of multiple analysis results.
[0037] In the actual operation process, since each "scenario" in the "scenario set" corresponds to a data usage requirement of the insurance analyst, after the "scenario" is constructed, a corresponding connection between the "scenario" and the "result" is established. Specifically, after the "scenario" is constructed, the person responsible for data extraction writes code in the database to extract the corresponding data and conducts manual statistical analysis, and feeds back the statistical results to the insurance analyst. Each feedback information corresponds to a result in the "result set". Since the codes for extracting data in the database are relatively fixed and can be modeled and encapsulated, when the insurance analyst puts forward an analysis requirement, it becomes a question of how to accurately match the proposed requirements with one or several scenarios in the "scenario set".
[0038] Traditional automatic matching methods include matching based on TF-IDF keywords, vectorization based on LSA decomposition matrix, and classification matching based on LDA topic model. These methods only focus on the meaning conveyed by the words themselves, and do not take into account the semantic connection between words and the complex semantic structure of the text itself caused by the use of natural language by insurance analysts in the insurance industry, resulting in low accuracy. The simple use of language models for matching, due to the large amount of data processed, leads to a relatively slow matching speed of language models.
[0039] Therefore, the present invention uses a method for analyzing customer information in the insurance industry that combines logistic regression and deep matching, which is used to accurately understand the needs of insurance analysts in the insurance industry, select scenarios from the "scenario set", and output results from the "result set". The algorithm simultaneously solves the problem that the logistic regression matching model is relatively fast but the accuracy is not very high, and the deep matching model has a relatively high accuracy but a slow matching speed, achieving a balance between speed and accuracy. The present application also has self-learning capabilities, and can continuously improve matching accuracy as the number of analyses increases.
[0040] In the first step, when the insurance analyst proposes a data analysis requirement, this application first uses the logistic regression model to perform a preliminary match between the input analysis requirement and all the scenarios in the "scenario set", and retains the top n scenarios with the highest degree of match. The logistic regression model is a probability distribution model from scenario to analysis requirement. The greater the probability that the input scenario is mapped to the analysis requirement through the logistic regression model, the more the scenario matches the analysis requirement; the smaller the probability that the input scenario is mapped to the analysis requirement through the logistic regression model, the less the scenario matches the analysis requirement. Specifically, the logistic regression model uses the dot product of the analysis requirement and the scenario representation, and uses the Sigmoid function to probabilize it to represent the degree of match between the analysis requirement and the scenario. When training the logistic regression model, the following loss function value is maximized:
[0041]
[0042] d is any scenario. For analysis requirement l, scenario d either matches analysis requirement l or does not match analysis requirement l. Indicates m that matches the analysis requirement l + Any one of the scenes, Indicates that m does not match the analysis requirement l - Any one of the scenes, Scenario-based model for logistic regression Given the probability of analyzing the demand l, Scenario-based model for logistic regression The probability of the analysis requirement l is given. In this way, each scenario will get a matching probability, which indicates the degree of the scenario and the input analysis requirement, and the n scenarios with the highest probability are taken as the preliminary matching results.
[0043] In the second step, the preliminary matching results are input into the deep matching model for further matching, and k final matching scenes are outputted from high to low according to the matching probability. Figure 2 As shown, the deep matching model includes a text feature extractor built using the BERT model. The BERT model extracts key features of insurance information such as policy loans, refund amounts, policy cancellation progress, and business growth in natural language.
[0044] The analysis requirements proposed by the insurance analysts and the scenarios in the "scenario set" are encoded into texts and converted into vectors that can be recognized by the BERT model. Specifically, the analysis requirements proposed by the insurance analysts and the scenarios in the "scenario set" are segmented and converted into vectors through the embedding model.
[0045] The vectors representing the analysis requirements proposed by the insurance analyst and the scenarios in the "scenario set" are combined with the position vector and input into the BERT model to form the corresponding matching vector. The matching vector is then input into the matching fully connected network, which predicts the matching probability of the matching vector and outputs the final k matching scenarios in descending order of probability. When training the matching fully connected network, the following loss function value is maximized:
[0046]
[0047] vec d is the matching vector of any scene, vec l is the matching vector of the analysis requirement and the matching vector of the scenario vec d or a matching vector with the analysis requirements vec l matches, or matches the matching vector required by the analysis vec l does not match, Represents the matching vector with the analysis requirements vec l matches m + Any matching vector in a scene, Represents the matching vector with the analysis requirements vec l does not match m - Any matching vector in a scene, To match the scene-based matching vector of the fully connected network Give the matching vector of the analysis requirements vec The probability of l, To match the scene-based matching vector of the fully connected network Given a matching vector vecThe probability of l.
[0048] The third step is to obtain the final matching scenario and, based on the association between the results and the scenario, use the pre-packaged data query statement to generate k analysis results of the "result set" corresponding to the final matching scenario.
[0049] The initial data for the training of the logistic regression model and the deep matching model in the first and second steps are derived from the historical data analysis log records provided by the insurance company database background for a period of time. However, due to the incompleteness of the training data, the initial parameters of the model will inevitably have errors, and the following two problems may occur: one is that the insurance analyst does not match the corresponding scenario after proposing the data analysis requirements, resulting in no analysis results output; the other is that the results in the front do not meet the analysis requirements. The logistic regression model and deep matching model of this application support continuous self-optimization during use. During the actual operation of the platform, the insurance analyst inputs an analysis requirement, the algorithm system generates recommended analysis results, and the analyst selects the required analysis results. For the first problem, during the operation of the platform, a certain proportion of the analysis results in the "result set" that are not within the display recommendation range will be randomly selected from the matching analysis results and displayed to the user to prevent some results in the "result set" from being excluded by the model forever, and the displayed data can be adjusted according to the user's selection. For the second problem, the supervision data is constructed through the selection results of the analysis results by the insurance analyst, and the supervision data is used to dynamically update the parameters of the model so that the model can match reasonable data more accurately.
[0050] When training a matching fully connected network, the following loss function is maximized:
[0051]
[0052] vec d is the matching vector of any scene, vec l is the matching vector of the analysis requirement and the matching vector of the scenario vec d or a matching vector with the analysis requirements vec l matches, or matches the matching vector required by the analysis vec l does not match, Represents the matching vector with the analysis requirements vec l matches m + Any matching vector in a scene, Represents the matching vector with the analysis requirements vec l does not match m - Any matching vector in a scene, To match the scene-based matching vector of the fully connected network Give the matching vector of the analysis requirements vec The probability of l, To match the scene-based matching vector of the fully connected network Given a matching vector vec The probability of l.
[0053] Suppose that an insurance analyst inputs an analysis requirement: "How much has the premium increased year-on-year, and how is the growth rate compared with previous years?" The platform finally outputs k analysis results, namely: c1, c2, c3, c4, c5, ..., c k , among which, the analyst selected k1 results such as "premium increased by XX%", "ranked third in the past 10 years", and "annual growth rate ranked last in the past 5 years". Then these k1 results are the successfully matched analysis results, which meet the analyst's needs and can be used as positive examples, recorded as The remaining k2 items have no clicks, so these k2 items are the analysis results of unsuccessful matching, which do not meet the needs of analysts and can be used as counterexamples, recorded as, Based on the association between results and scenarios, and k new supervisory data that can generate the analysis requirement of "how much the premium has increased year-on-year, and how the growth rate is compared with previous years" and As the platform is used more frequently, more supervised data samples that meet the needs of insurance analysts can be generated. When a certain amount of supervised data samples are accumulated, the logistic regression and deep matching models are updated and trained, and the relevant parameters are corrected. In this way, as insurance analysts continue to use the platform, the recommended results will become more intelligent, reasonable, and in line with the actual needs of the insurance industry.
[0054] Example 2
[0055] See also Figure 3 As shown, an embodiment of the present invention provides an insurance industry customer information analysis device, including: the processing unit and the storage unit are interconnected through a bus unit, and the storage unit can be used to store software programs, computer executable programs and modules, such as the software programs, computer executable programs and modules corresponding to an insurance industry customer information analysis method in an embodiment of the present invention. The processing unit implements the above-mentioned insurance industry customer information analysis method by running the software programs, computer executable programs and modules stored in the storage unit, including:
[0056] Establish "scenario sets" and "result sets" based on historical data analysis logs;
[0057] Use the logistic regression model to perform a preliminary match between all scenarios in the "scenario set" and the input analysis requirements, and retain the top n scenarios with the highest matching degree;
[0058] The preliminary matching results are input into the deep matching model for further matching, and k final matching scenes are output from high to low according to the matching probability;
[0059] After obtaining the final matching scenario, based on the association between the results and the scenario, a pre-packaged data query statement is used to generate k analysis results of the "result set" corresponding to the final matching scenario.
[0060] Of course, the computer program stored in the storage unit of the insurance industry customer information analysis device provided by the embodiment of the present invention is not limited to the method operations described above, and can also execute related operations in the insurance industry customer information analysis method provided by any embodiment of the present invention.
[0061] Example 3
[0062] An embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program. When the computer program is executed, the insurance industry customer information analysis method is implemented, including:
[0063] Establish "scenario sets" and "result sets" based on historical data analysis logs;
[0064] Use the logistic regression model to perform a preliminary match between all scenarios in the "scenario set" and the input analysis requirements, and retain the top n scenarios with the highest matching degree;
[0065] The preliminary matching results are input into the deep matching model for further matching, and k final matching scenes are output from high to low according to the matching probability;
[0066] After obtaining the final matching scenario, based on the association between the results and the scenario, a pre-packaged data query statement is used to generate k analysis results of the "result set" corresponding to the final matching scenario.
[0067] A computer-readable storage medium provided in an embodiment of the present invention stores a computer program which is not limited to the method operations described above, but can also execute related operations in an insurance industry customer information analysis method provided in any embodiment of the present invention.
[0068] In the embodiments provided by the present invention, it should be understood that the disclosed structures and methods can be implemented in other ways. For example, the structural embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, structures or units, which can be electrical, mechanical or other forms.
[0069] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0070] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0071] The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features claimed herein.
Claims
1. A method for analyzing customer information in the insurance industry, characterized in that: include: Establish "scenario set" and "result set" based on historical data analysis logs; Use the logistic regression model to perform a preliminary match between all scenarios in the "scenario set" and the input analysis requirements, and retain the top n scenarios with the highest matching degree; The preliminary matching results are input into the deep matching model for further matching, and k final matching scenes are output from high to low according to the matching probability; After obtaining the final matching scenario, based on the association between the results and the scenario, a pre-packaged data query statement is used to generate k analysis results of the "result set" corresponding to the final matching scenario.
2. The insurance industry customer information analysis method according to claim 1, characterized in that: Based on the different uses of customer information in historical data analysis logs, different types of data application "scenario sets" and data analysis "result sets" are obtained; each "scenario" in the "scenario set" corresponds to a type of data usage demand, and each "result" in the "result set" corresponds to a single sensitive data statistical analysis result or a collection of multiple analysis results.
3. The insurance industry customer information analysis method according to claim 1, characterized in that: After the "scenario" is constructed, code is written in the database to extract the corresponding data and perform statistical analysis to obtain the statistical results of the corresponding scenario. Each statistical result corresponds to a result in the "result set", thereby establishing a relationship between the result and the scenario.
4. The insurance industry customer information analysis method according to claim 1, characterized in that: The logistic regression model is a probability distribution model from scenes to analysis requirements based on a fully connected network. The greater the probability that the input scene is mapped to the analysis requirement through the logistic regression model, the more the scene matches the analysis requirement; the smaller the probability that the input scene is mapped to the analysis requirement through the logistic regression model, the less the scene matches the analysis requirement.
5. The insurance industry customer information analysis method according to claim 4, characterized in that: When training a logistic regression model, the following loss function is maximized: d is any scenario. For analysis requirement l, scenario d either matches analysis requirement l or does not match analysis requirement l. Indicates m that matches the analysis requirement l + Any one of the scenes, Indicates that m does not match the analysis requirement l - Any one of the scenes, Scenario-based model for logistic regression Given the probability of analyzing the demand l, Scenario-based model for logistic regression Give the probability of analyzing the demand l.
6. The insurance industry customer information analysis method according to claim 1, characterized in that: The deep matching model includes a text feature extractor built using a pre-trained BERT model, which extracts key features of insurance information in natural language; Encode the analysis requirements proposed by insurance analysts and the scenarios in the "scenario set" into vectors that can be recognized by the BERT model; The vectors representing the analysis requirements proposed by the insurance analyst and the scenarios in the "scenario set" are combined with the position vector and input into the BERT model to form the corresponding matching vector; the matching vector is then input into the matching fully connected network, which predicts the matching probability of the matching vector and outputs the final k matching scenarios in descending order of probability.
7. The insurance industry customer information analysis method according to claim 6, characterized in that: When training a matching fully connected network, the following loss function is maximized: vec d is the matching vector of any scene, vec l is the matching vector of the analysis requirement and the matching vector of the scenario vec d or a matching vector with the analysis requirements vec l matches, or matches the matching vector required by the analysis vec l does not match, Represents the matching vector with the analysis requirements vec l matches m + Any matching vector in a scene, Represents the matching vector with the analysis requirements vec l does not match m - Any matching vector in a scene, To match the scene-based matching vector of the fully connected network Give the matching vector of the analysis requirements vec The probability of l, To match the scene-based matching vector of the fully connected network Given a matching vector vec The probability of l.
8. The method for analyzing insurance industry customer information according to claim 1, characterized in that: During the actual operation of the platform, the insurance analyst inputs an analysis requirement, the algorithm system generates recommended analysis results, and the analyst selects the required analysis results; among them, a certain proportion of the analysis results in the "result set" that are not within the display recommendation range are randomly extracted from the matching analysis results and displayed to the user; the supervision data is constructed through the selection results of the insurance analyst on the analysis results, and the supervision data is used to dynamically update the parameters of the model.
9. A customer information analysis device for the insurance industry, characterized in that: include: At least one processing unit, the processing unit and the storage unit are interconnected via a bus unit, the storage unit stores a computer program, and when the computer program is executed by the processing unit, the insurance industry customer information analysis method as described in any one of claims 1-8 is implemented.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the insurance industry customer information analysis method as described in any one of claims 1-8 is implemented.