Claims Prediction Method and Apparatus, Electronic Device, and Storage Medium

Abstract

An embodiment of the present application provides a claim prediction method, apparatus, electronic device, and storage medium, belonging to the field of artificial intelligence technology. The method includes: obtaining case information data and case description data of a target case, where the case information data is used to record the basic case information of the target case, and the case description data is used to describe the case situation of the target case; extracting features from the case information data to obtain a case feature vector; performing node prediction on the case feature vector through a tree model to obtain target tree node data and a first prediction value; performing encoding processing on the case description data to obtain a target semantic hidden vector; performing prediction processing on the target tree node data and the target semantic hidden vector through a claim prediction model to obtain a second prediction value; calculating a claim based on the first prediction value and the second prediction value to obtain target claim data. The embodiment of the present application can improve the accuracy of claim situation judgment.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular, to a claim prediction method and apparatus, an electronic device, and a storage medium. Background Art

[0002] Currently, in the claim settlement business, the relevant data of the claim cases to be settled are often verified by manual review, and then the claim settlement situation of the claim cases to be settled is determined according to the verification results. This method often has a large degree of subjectivity and a large error, which affects the accuracy of the judgment of the claim settlement situation. Therefore, how to improve the accuracy of the judgment of the claim settlement situation has become an urgent technical problem to be solved. Summary of the Invention

[0003] The main purpose of the embodiments of this application is to propose a claim prediction method and apparatus, an electronic device, and a storage medium, aiming to improve the accuracy of the judgment of the claim settlement situation.

[0004] To achieve the above object, a first aspect of the embodiments of this application proposes a claim prediction method, and the method includes:

[0005] Obtain the case information data and case description data of the target case, where the case information data is used to record the basic case information of the target case, and the case description data is used to describe the case situation of the target case;

[0006] Extract features from the case information data to obtain a case feature vector;

[0007] Perform node prediction on the case feature vector through a preset tree model to obtain target tree node data and a first prediction value;

[0008] Perform encoding processing on the case description data to obtain a target semantic hidden vector;

[0009] Perform prediction processing on the target tree node data and the target semantic hidden vector through a pre-trained claim prediction model to obtain a second prediction value;

[0010] Perform claim calculation according to the first prediction value and the second prediction value to obtain target claim data.

[0011] In some embodiments, the step of performing node prediction on the case feature vector through a preset tree model to obtain target tree node data and a first prediction value includes:

[0012] Generate nodes for the case feature vector through the tree model to obtain target leaf nodes;

[0013] Perform combination processing on the target leaf nodes to obtain the target tree node data;

[0014] Calculate the probability of the target tree node data through a preset first function to obtain the first prediction value.

[0015] In some embodiments, the claim prediction model includes an encoding layer and a prediction layer. The step of performing prediction processing on the target tree node data and the target semantic hidden vector through a pre-trained claim prediction model to obtain a second prediction value includes:

[0016] Perform encoding processing on the target tree node data through the encoding layer to obtain a tree node encoding vector;

[0017] Perform claim prediction on the tree node encoding vector and the target semantic hidden vector through the prediction layer to obtain the second prediction value.

[0018] In some embodiments, the step of performing claim prediction on the tree node encoding vector and the target semantic hidden vector through the prediction layer to obtain the second prediction value includes:

[0019] Perform vector addition on the tree node encoding vector and the target semantic hidden vector according to preset weight parameters to obtain a target case hidden vector;

[0020] Calculate the claim rate of the target case hidden vector through a second function of the prediction layer to obtain the second prediction value.

[0021] In some embodiments, the case information data includes first discrete data, second discrete data, first numerical data, and second numerical data. The step of performing feature extraction on the case information data to obtain a case feature vector includes:

[0022] Perform encoding processing on the first discrete data according to a preset encoding serial number to obtain a first encoding vector;

[0023] Perform multi-hot encoding processing on the second discrete data to obtain a second encoding vector;

[0024] Perform encoding processing on the first numerical data according to preset padding parameters to obtain a third encoding vector;

[0025] Perform numerical extraction on the second numerical data to obtain a target numerical feature, and perform encoding processing on the target numerical feature to obtain a fourth encoding vector;

[0026] Perform vector splicing on the first encoding vector, the second encoding vector, the third encoding vector, and the fourth encoding vector to obtain the case feature vector.

[0027] In some embodiments, the step of encoding the case description data to obtain the target semantic hidden vector includes:

[0028] Convert the format of the case description data to obtain an initial description text;

[0029] Perform segmentation processing on the initial description text according to a preset text length to obtain a target description text;

[0030] Encode the target description text through a preset encoding model to obtain the target semantic hidden vector.

[0031] In some embodiments, the case description data includes first text data, second text data, and third text data. The step of converting the format of the case description data to obtain an initial description text includes:

[0032] Convert the format of the first text data according to a preset first format template to obtain a first description text;

[0033] Convert the format of the second text data according to a preset second format template to obtain a second description text;

[0034] Convert the format of the third text data according to a preset third format template to obtain a third description text;

[0035] Perform splicing processing on the first description text, the second description text, and the third description text according to a preset splicing order to obtain the initial description text.

[0036] To achieve the above object, a second aspect of the embodiments of the present application proposes a claim settlement prediction device, which includes:

[0037] A data acquisition module, configured to acquire case information data and case description data of a target case, where the case information data is used to record the basic case information of the target case, and the case description data is used to describe the case situation of the target case;

[0038] A feature extraction module, configured to extract features from the case information data to obtain a case feature vector;

[0039] A node prediction module, configured to perform node prediction on the case feature vector through a preset tree model to obtain target tree node data and a first prediction value;

[0040] An encoding module, configured to encode the case description data to obtain a target semantic hidden vector;

[0041] A claim prediction module, configured to perform prediction processing on the target tree node data and the target semantic hidden vector through a pre-trained claim prediction model to obtain a second prediction value;

[0042] A calculation module, configured to perform claim calculation based on the first prediction value and the second prediction value to obtain target claim data.

[0043] To achieve the above object, a third aspect of the embodiments of the present application provides an electronic device, which includes a memory, a processor, a program stored on the memory and executable on the processor, and a data bus for realizing connection communication between the processor and the memory. When the program is executed by the processor, the method described in the first aspect above is realized.

[0044] To achieve the above object, a fourth aspect of the embodiments of the present application provides a storage medium, which is a computer-readable storage medium for computer-readable storage. The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to realize the method described in the first aspect above.

[0045] The claim prediction method, claim prediction device, electronic device and storage medium proposed by the present application obtain the case information data and case description data of the target case. Among them, the case information data is used to record the basic case information of the target case, and the case description data is used to describe the case situation of the target case, and can perform claim prediction based on the basic case information and the case situation. Further, feature extraction is performed on the case information data to obtain a case feature vector, and node prediction is performed on the case feature vector through a preset tree model to obtain target tree node data and a first prediction value. The tree model can meet the requirements of processing numerical features and discrete features in the case information data, and can identify relatively complex semantic information in the case information data, improving the comprehensiveness of feature processing and the accuracy of semantic recognition. Further, encoding processing is performed on the case description data to obtain a target semantic hidden vector, and prediction processing is performed on the target tree node data and the target semantic hidden vector through a pre-trained claim prediction model to obtain a second prediction value, which can perform prediction by combining the basic case information and the case situation of the target case, improving the diversity and comprehensiveness of the feature information used for claim prediction. Finally, claim calculation is performed based on the first prediction value and the second prediction value to obtain target claim data, which can perform claim analysis by comprehensively considering the prediction situations of the tree model and the claim prediction model, making the target claim data more reasonable and objective, thereby improving the accuracy of the judgment of the claim situation. Description of the Drawings

[0046] Figure 1 is a flowchart of the claim prediction method provided by the embodiments of the present application;

[0047] Figure 2 is Figure 1 the flowchart of step S102 in

[0048] Figure 3 is Figure 1 the flowchart of step S103 in

[0049] Figure 4 is Figure 1 the flowchart of step S104 in

[0050] Figure 5 is Figure 4 the flowchart of step S401 in

[0051] Figure 6 is Figure 1 the flowchart of step S105 in

[0052] Figure 7 is Figure 6 the flowchart of step S602 in

[0053] Figure 8 is the structural schematic diagram of the claim prediction device provided by the embodiments of the present application;

[0054] Figure 9 is the hardware structural schematic diagram of the electronic device provided by the embodiments of the present application. Detailed implementation manners

[0055] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0056] It should be noted that although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order from the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the description and claims and the above accompanying drawings are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence.

[0057] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0058] First, several nouns involved in the present application are analyzed:

[0059] Artificial Intelligence (AI): It is a new technical science that studies and develops theories, methods, technologies, and application systems for simulating, extending, and expanding human intelligence. Artificial intelligence is a branch of computer science. It attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a way similar to human intelligence. The research in this field includes robots, speech recognition, image recognition, natural language processing, and expert systems, etc. Artificial intelligence can simulate the information process of human consciousness and thinking. It also refers to the theory, method, technology, and application system that uses a digital computer or a machine controlled by a digital computer to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to obtain the best results.

[0060] Natural Language Processing (NLP): NLP uses computers to process, understand, and apply human languages (such as Chinese, English, etc.). NLP is a branch of artificial intelligence and an interdisciplinary field of computer science and linguistics, and is often referred to as computational linguistics. Natural language processing includes syntactic analysis, semantic analysis, discourse understanding, etc. Natural language processing is commonly used in technical fields such as machine translation, handwritten and printed character recognition, speech recognition and text-to-speech conversion, information intention recognition, information extraction and filtering, text classification and clustering, public opinion analysis, and opinion mining. It involves data mining, machine learning, knowledge acquisition, knowledge engineering, artificial intelligence research related to language processing, and linguistic research related to language computing.

[0061] Information Extraction: A text processing technology that extracts factual information such as specified types of entities, relationships, events, etc. from natural language texts and forms structured data for output. Information extraction is a technology for extracting specific information from text data. Text data is composed of some specific units, such as sentences, paragraphs, and texts. Text information is exactly composed of some small specific units, such as words, phrases, sentences, paragraphs, or combinations of these specific units. Extracting noun phrases, personal names, place names, etc. from text data are all text information extraction. Of course, the information extracted by text information extraction technology can be various types of information.

[0062] A decision tree is a decision analysis method that, based on the known probabilities of various situations, calculates the probability that the expected net present value is greater than or equal to zero by constructing a decision tree, evaluates project risks, and determines its feasibility. It is a graphical method that intuitively applies probability analysis. Since the decision branches are drawn in a graph that resembles the branches of a tree, it is called a decision tree. In machine learning, a decision tree is a prediction model that represents a mapping relationship between object attributes and object values. A decision tree is a tree structure where each internal node represents a test on an attribute, each branch represents a test output, and each leaf node represents a category.

[0063] Encoding: Encoding is the process of converting information from one form or format to another. Use a pre-specified method to encode text, numbers, or other objects into digital codes, or convert information and data into specified electrical pulse signals. Encoding is widely used in electronic computers, televisions, remote controls, and communications. Encoding is the process of converting information from one form or format to another.

[0064] Currently, in the claims settlement business, the relevant data of claims settlement cases are often verified by manual review, and then the claims settlement situation of the claims settlement cases to be processed is determined according to the verification results. This method often has a large degree of subjectivity and a large error, which affects the accuracy of the judgment of the claims settlement situation. Therefore, how to improve the accuracy of the judgment of the claims settlement situation has become a technical problem to be solved urgently.

[0065] Based on this, the embodiments of the present application provide a claims settlement prediction method, a claims settlement prediction device, an electronic device, and a storage medium, aiming to improve the accuracy of the judgment of the claims settlement situation.

[0066] The claims settlement prediction method, the claims settlement prediction device, the electronic device, and the storage medium provided by the embodiments of the present application are specifically described in the following embodiments. First, the claims settlement prediction method in the embodiments of the present application is described.

[0067] The embodiments of the present application can obtain and process relevant data based on artificial intelligence technology. Among them, 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 knowledge to obtain the best results.

[0068] The basic technologies of artificial intelligence generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation / interaction systems, and mechatronics. The software technologies of artificial intelligence mainly include several major directions such as computer vision technology, robotics, biometric technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0069] The claim prediction method provided by the embodiments of this application relates to the field of artificial intelligence technology. The claim prediction method provided by the embodiments of this application can be applied to a terminal, or to a server side, or can also be software running on a terminal or a server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or can be configured as a server cluster or a distributed system composed of multiple physical servers, or can also be configured as a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the claim prediction method, etc., but is not limited to the above forms.

[0070] This application can be used in many general or specific computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and so on. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. This application can also be practiced in a distributed computing environment, where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0071] Figure 1 is an optional flowchart of the claim prediction method provided by the embodiments of this application, Figure 1 The method in may include but is not limited to steps S101 to S106.

[0072] Step S101, obtain the case information data and case description data of the target case, where the case information data is used to record the basic case information of the target case, and the case description data is used to describe the case situation of the target case;

[0073] Step S102: Extract features from the case information data to obtain a case feature vector;

[0074] Step S103: Perform node prediction on the case feature vector through a preset tree model to obtain target tree node data and a first prediction value;

[0075] Step S104: Perform encoding processing on the case description data to obtain a target semantic hidden vector;

[0076] Step S105: Perform prediction processing on the target tree node data and the target semantic hidden vector through a pre-trained claim prediction model to obtain a second prediction value;

[0077] Step S106: Perform claim calculation based on the first prediction value and the second prediction value to obtain target claim data.

[0078] Steps S101 to S106 illustrated in the embodiments of the present application extract features from the case information data to obtain a case feature vector, and perform node prediction on the case feature vector through a preset tree model to obtain target tree node data and a first prediction value. The tree model can meet the requirements of processing numerical features and discrete features in the case information data, and can identify relatively complex semantic information in the case information data, improving the comprehensiveness of feature processing and the accuracy of semantic recognition. Further, perform encoding processing on the case description data to obtain a target semantic hidden vector, and perform prediction processing on the target tree node data and the target semantic hidden vector through a pre-trained claim prediction model to obtain a second prediction value, which can perform prediction by combining the basic case information and the case details of the target case, improving the diversity and comprehensiveness of the feature information for claim prediction. Finally, perform claim calculation based on the first prediction value and the second prediction value to obtain target claim data, which can perform claim analysis by comprehensively considering the prediction situations of the tree model and the claim prediction model, making the target claim data more reasonable and objective, thereby improving the accuracy of judging the claim situation.

[0079] It should be noted that in each specific implementation manner of the present application, when it comes to performing relevant processing based on data related to the user's identity or characteristics, such as user information, user behavior data, the user's case information data and case description data, user historical data, and user location information, etc., the user's permission or consent will be obtained first. Moreover, the collection, use, and processing of these data will comply with the relevant laws, regulations, and standards of the relevant countries and regions. In addition, when the embodiments of the present application need to obtain the user's sensitive personal information, the user's separate permission or separate consent will be obtained through methods such as pop-up windows or jumping to a confirmation page. After clearly obtaining the user's separate permission or separate consent, the necessary user-related data for the normal operation of the embodiments of the present application will be obtained.

[0080] In step S101 of some embodiments, data can be crawled targetedly after setting up the data source by writing a web crawler to obtain the case information data and case description data of the target case. The case information data and case description data can also be obtained by other means, not limited to this. Among them, the target case is generally a case to be settled. The case information data is used to record the basic information of the target case. The case information data can include the diagnosis disease code, insurance type code, date difference between the accident time and the insurance effective period, policy amount, etc. The case description data is text data, which is used to describe the case situation of the target case. The case description data can include the basic information of the disease that the user sees a doctor for (such as disease name, disease symptoms, etc.), the agreed rules on the policy, the communication records when the policy is entered, etc.

[0081] Please refer to Figure 2 , in some embodiments, the case information data includes first discrete data, second discrete data, first numerical data, and second numerical data. Step S102 can include but is not limited to steps S201 to S205:

[0082] Step S201, perform encoding processing on the first discrete data according to a preset encoding serial number to obtain a first encoding vector;

[0083] Step S202, perform multi-hot encoding processing on the second discrete data to obtain a second encoding vector;

[0084] Step S203, perform encoding processing on the first numerical data according to a preset padding parameter to obtain a third encoding vector;

[0085] Step S204, perform numerical extraction on the second numerical data to obtain a target numerical feature, and perform encoding processing on the target numerical feature to obtain a fourth encoding vector;

[0086] Step S205, perform vector concatenation on the first encoding vector, the second encoding vector, the third encoding vector, and the fourth encoding vector to obtain a case feature vector.

[0087] In step S201 of some embodiments, the first discrete data mainly includes diagnostic disease codes. Among them, the diagnostic disease codes are mainly code data represented by the user using the International Classification of Diseases (ICD) when diagnosing diseases. The first discrete data is generally single-valued discrete feature data. Therefore, for different disease codes, it is necessary to encode the disease names corresponding to the disease codes in sequence. Specifically, the corresponding encoding serial numbers can be preset according to the code categories of the disease codes. According to the encoding serial numbers, L1 different disease codes are encoded, and the disease names corresponding to the disease codes are sequentially encoded as 0, 1, …, L1-1, so as to obtain the first encoding vector; among them, the vector length of the first vector is 1, and the vector size of the first encoding vector is the encoded disease name, that is, 0, 1, …, L1-1.

[0088] In step S202 of some embodiments, the second discrete data mainly includes insurance type codes. Among them, each insurance product corresponds to an insurance type code (i.e., insurance type code). A user can purchase multiple insurance products, so there will be multiple insurance type codes under the user's name. The second discrete data is generally multi-valued discrete feature data. Therefore, for different insurance type codes, a multi-hot encoding method can be used for encoding to obtain the second encoding vector. Among them, the vector length of the second encoding vector is L2. For m insurance types, through multi-hot encoding, only M positions in the second encoding vector corresponding to each insurance type take the value of 1, and the other positions (L2-M positions) take the value of 0. It should be noted that the M positions taking the value of 1 in the second encoding vector can be randomly selected without limitation.

[0089] In step S203 of some embodiments, the first numerical data mainly includes the date difference between the accident time and the insurance effective period. That is, each type of insurance product corresponds to an effective date. Therefore, the difference between the effective time of each insurance type and the accident occurrence time can be calculated to obtain the date difference. This date difference is a specific value, and through this date difference, it can be judged whether the insurance product is within the grace period or whether it is a short-term claim, etc. The first numerical data is generally multi-valued numerical feature data. Therefore, for different date differences, a method of filling parameters can be used for encoding to obtain the third encoding vector. Among them, the vector length of the third encoding vector is L3. For m insurance types, through the method of filling parameters, only M positions in the third encoding vector corresponding to each insurance type take the corresponding values, and the other positions (L3-M positions) take the value of missing value (nan). It should be noted that the M positions taking the corresponding values in the third encoding vector can be randomly selected without limitation, and the positions taking the value of missing value (nan) can be numerically complemented by a tree model in subsequent steps.

[0090] In step S204 of some embodiments, the second numerical data mainly includes the policy amount, which is generally a specific value. The second numerical data is generally single-valued numerical feature data. Therefore, the numerical extraction can be directly performed on the second numerical data to obtain the target numerical feature, which is the policy amount value. The encoding process is performed on the target numerical feature to obtain the fourth encoding vector.

[0091] In step S205 of some embodiments, in order to make the semantics of the case feature vector clear, the first encoding vector, the second encoding vector, the third encoding vector, and the fourth encoding vector can be vector-concatenated in the order of diagnosis disease code - insurance type code - date difference between the accident time and the insurance effective period - policy amount to obtain the final case feature vector, where the vector length of the case feature vector is 1 + L2 + L3 + 1.

[0092] Through the above steps S201 to S205, different data can be respectively encoded according to the nature of the case information data (that is, whether the data is discretized or numerical, multi-valued or single-valued), so as to relatively completely retain the semantic information of the case information data.

[0093] Before step S103 of some embodiments, the claim prediction method further includes pre-training a tree model. The tree model of the embodiments of the present application is mainly constructed based on the LightGBM tree model. The tree model is trained by inputting the sample case information data and the true label corresponding to the sample case information data to obtain a tree model that meets the requirements, where the true label includes normal claims and abnormal claims. Specifically, the tree model constructed according to the Light GBM model will have multiple decision trees, each decision tree includes multiple leaf nodes, and each input case information data will be assigned to a leaf node, so that each leaf node has a unique score value, so that the tree node number corresponding to each input can be extracted through the tree model. The tree model can process the sparse numerical features in the case information data and can identify the relatively complex semantic information in the case information data, thereby improving the comprehensiveness of feature processing and the accuracy of semantic recognition.

[0094] Please refer to Figure 3 , in some embodiments, step S103 may include but is not limited to steps S301 to S303:

[0095] Step S301, generating nodes for the case feature vector through the tree model to obtain the target leaf node;

[0096] Step S302, performing a combination process on the target leaf node to obtain the target tree node data;

[0097] Step S303, perform probability calculation on the target tree node data through a preset first function to obtain a first prediction value.

[0098] In step S301 of some embodiments, since each decision tree in the tree model includes multiple leaf nodes, when the tree model extracts nodes from the case feature vector, each input case feature vector will be assigned to a leaf node, so that each leaf node has a unique score value. Therefore, the node information corresponding to each case feature vector can be extracted through the tree model to obtain the target leaf node.

[0099] In step S302 of some embodiments, the target tree node data includes a tree node number list and a comprehensive score value. Specifically, obtain the leaf node numbers of each target leaf node, splice the leaf node numbers of all decision trees to obtain the final tree node number list, and through this method, the score values predicted by all decision trees can be accumulated to obtain the comprehensive score value.

[0100] In addition, if there are 100 decision trees in the tree model, the length of the tree node number vector is 100, that is, the length of the tree node number vector is the same as the number of decision trees.

[0101] In step S303 of some embodiments, the preset first function can be a normalization function such as the sigmiod function or the softmax function. Perform probability calculation on the target tree node data through the normalization function to obtain the first prediction value, and this first prediction value is the claim prediction value of the tree model for the target case.

[0102] The above steps S301 to S303 can meet the need to process numerical features and discrete features in the case information data through the tree model, better process sparse numerical features, and perform robust processing on missing values, so as to identify relatively complex semantic information in the case information data, improving the comprehensiveness of feature processing and the accuracy of semantic recognition.

[0103] Please refer to Figure 4 , in some embodiments, step S104 may include but is not limited to steps S401 to S403:

[0104] Step S401, perform format conversion on the case description data to obtain the initial description text;

[0105] Step S402, perform segmentation processing on the initial description text according to the preset text length to obtain the target description text;

[0106] Step S403, perform encoding processing on the target description text through a preset encoding model to obtain the target semantic hidden vector.

[0107] In step S401 of some embodiments, in order to make the case description data meet the processing requirements of the text processing model, it is necessary to perform format conversion on the case description data. For example, for different types of case description data, different format templates are used for text adjustment to obtain a description text that meets the requirements of the current type, and then the description texts of multiple types are integrated to obtain an initial description text.

[0108] In step S402 of some embodiments, in order to improve the efficiency of claim settlement judgment, it is necessary to appropriately streamline the text length of the initial description text so that the length of the initial description text meets the input requirements of the text processing model (such as the BERT model). Among them, the preset text length can be set according to actual business needs. Usually, it is necessary to make the set text length cover most texts. For example, the preset text length is generally 512 text lengths. In the specific segmentation process, often the text part of the initial description text that exceeds the preset text length is truncated and deleted to obtain the target description text.

[0109] For example, if the text length of an initial description text is 700 and the preset text length is 512, then directly truncate the first 512 parts of the initial description text and use this part as the target description text, and discard the part that exceeds 512 in length at the back, thereby effectively improving the segmentation efficiency.

[0110] In step S403 of some embodiments, the preset encoding model can be the BERT model. The BERT model is used to encode the target description text to obtain the text feature hidden state of the target description text, thereby obtaining the target semantic hidden vector E.

[0111] In the above steps S401 to S403, by using the encoding model, it is relatively convenient to model the case text of the target case, and at the same time, the complex semantic information in the case description data can be identified, improving the processing accuracy of the text information.

[0112] Please refer to Figure 5 , in some embodiments, the case description data includes first text data, second text data, and third text data. Step S401 may include but is not limited to steps S501 to S504:

[0113] Step S501, perform format conversion on the first text data according to the preset first format template to obtain the first description text;

[0114] Step S502, perform format conversion on the second text data according to the preset second format template to obtain the second description text;

[0115] Step S503: Convert the format of the third text data according to a preset third format template to obtain a third description text;

[0116] Step S504: Concatenate the first description text, the second description text, and the third description text according to a preset concatenation order to obtain an initial description text.

[0117] In step S501 of some embodiments, since there is a process of diagnosing diseases for the claim cases to be predicted, and corresponding text data will also be generated for the specific conditions of the diagnosed diseases, the first text data mainly includes description data of the diseases. For example, it is a description of the disease name of the user's medical treatment. To improve the text standardization, the preset first format template is "The diagnosed disease name of the target case is xxx". For example, when the disease name is "hypertension", after converting the disease name through the first format template, the obtained first description text T1 can be expressed as "The diagnosed disease name of the target case is hypertension".

[0118] In step S502 of some embodiments, the second text data mainly includes the agreed rules on the insurance policy. These agreed rules usually exclude certain diseases based on the pre-existing diseases informed by the user when handling the insurance. For example, the agreed rule is "Exclude liability for cervical spondylosis and gastritis", etc. It should be noted that this part is set according to the actual needs of the user, so there may be a situation where there are agreed rules on the user's insurance policy. In this regard, to improve the text standardization, the preset second format template is "The special agreement of the user's insurance policy for the target case is xxxx"; when there are no agreed rules on the user's insurance policy, the preset second format template is "There is no special agreement on the user's insurance policy for the target case". In this way, the format of the second text data can be converted more flexibly to obtain the second description text T2.

[0119] In step S503 of some embodiments, the third text data mainly includes the communication records during the insurance policy entry. Such communication records often exist in the insurance policy entry link, and in this link, the staff will enter the image information into the target system. During this process, the user can communicate with the target institution about issues such as blurred images and doubts about the insurance policy content, thus forming communication records. In this regard, to improve the text standardization, if there are communication records for the target case, the preset third format template is "The communication record during the entry link of the target case is xxxx"; if there are no communication records for the target case, the preset third format template is "There is no communication record during the entry link of the target case". In this way, the format of the third text data can be converted more flexibly to obtain the third description text T3.

[0120] In step S504 of some embodiments, the preset splicing order is disease name - the agreed rules on the insurance policy - the communication record during insurance policy entry. In some other embodiments, other splicing orders can also be adopted, and it is not limited to this. Specifically, according to the disease name - the agreed rules on the insurance policy - the communication record during insurance policy entry, the first description text, the second description text, and the third description text are spliced to obtain the initial description text T, where T = T1 + T2 + T3. This initial description text can clearly reflect the case content of the target case.

[0121] For example, the initial description text obtained through the above steps S501 to S504 is: The diagnosed disease name of the target case is hypertension. The special agreement of the user's insurance policy for the target case is that hypertension, kidney stones, and machine complications are excluded liabilities. The communication record in the entry link of the target case is "Reviewer: The hospital where the target case was treated is a non-designated hospital and there is no institutional application approval opinion. Please review whether to agree to the current treatment hospital. Institution: Pay in proportion."

[0122] Before step S105 of some embodiments, the claim prediction method further includes pre-training a claim prediction model. The claim prediction model includes an encoding layer and a prediction layer. The encoding layer is used to encode the input list of tree node numbers to obtain a tree node encoding vector. The prediction layer is mainly used to perform prediction processing on the tree node encoding vector and the input semantic vector to obtain a claim prediction value. When training the claim prediction model, the gradient descent method can be used to optimize the loss function. Among them, the loss function Loss can be expressed as shown in formula (1):

[0123] Loss = -(Ylog(P)+(1 - Y)log(1 - P)) Formula (1)

[0124] Among them, Y represents the true label of the sample case. This true label is 0 or 1. When the true label is 1, it means the sample case is a normal claim. When the true label is 0, it means the sample case is an abnormal claim. P represents the claim prediction value obtained through the claim prediction model. This claim prediction value is the predicted probability that the sample case is a normal claim obtained through the claim prediction model.

[0125] Please refer to Figure 6 , in some embodiments, the claim prediction model includes an encoding layer and a prediction layer. Step S105 includes but is not limited to steps S601 to S602:

[0126] Step S601, encoding the target tree node data through the encoding layer to obtain a tree node encoding vector;

[0127] Step S602, performing claim prediction on the tree node encoding vector and the target semantic hidden vector through the prediction layer to obtain a second prediction value.

[0128] In step S601 of some embodiments, the target tree node data includes a tree node number list TreeList, the length of the tree node number list is K, and TreeList i is the leaf node number hit on the i-th decision tree. If the i-th decision tree has Leave i leaf nodes, the tree nodes in the tree node number list are encoded by the encoding layer to obtain a tree node encoding vector TreeVector. Among them, the length of the tree node encoding vector is Among them, in the tree node encoding vector, the values of K positions are 1, and the values of the other L T -K positions are 0.

[0129] In step S602 of some embodiments, first, the tree node encoding vector and the target semantic hidden vector are fused to obtain a target case hidden vector. The target case hidden vector can be used not only for claim prediction in the embodiments of the present application but also for other application scenarios such as clustering or anomaly analysis of the target case. Further, the claim rate of the target case hidden vector is calculated by the second function of the prediction layer, so as to obtain a second prediction value, which is used to represent the judgment result of the claim prediction model on the claim situation of the target case.

[0130] Please refer to Figure 7 , in some embodiments, step S602 may include but is not limited to steps S701 to S702:

[0131] Step S701: Add the tree node encoding vector and the target semantic hidden vector according to the preset weight parameter to obtain a target case hidden vector;

[0132] Step S702: Calculate the claim rate of the target case hidden vector through the second function of the prediction layer to obtain a second prediction value.

[0133] In step S701 of some embodiments, the preset weight parameter can be set according to actual business requirements without limitation. The weight parameter can be represented as W1, and W1 is a two-dimensional trainable vector. According to the preset weight parameter W1, the tree node encoding vector TreeVector and the target semantic hidden vector E are added vectorially to obtain a target case hidden vector M, where the target case hidden vector can be represented as

[0134] In step S702 of some embodiments, the second function is the sigmoid function. The process of calculating the claim rate of the target case hidden vector M through the sigmoid function to obtain the second prediction value P2 can be expressed as shown in formula (2):

[0135]

[0136] Among them, W2 is a one-dimensional trainable vector, b is a trainable vector of a single value, and the specific sizes of W1, W2, and b can be determined according to the actual business scenario.

[0137] In step S106 of some embodiments, if the first prediction value and the second prediction value represent the same claim settlement situation of the target case, then the claim settlement situation reflected by the first prediction value or the second prediction value is used as the target claim settlement data. If the first prediction value and the second prediction value represent different claim settlement situations of the target case, then the claim settlement situation reflected by the second prediction value is used as the target claim settlement data.

[0138] Specifically, compare the first prediction value with a preset prediction threshold. If the first prediction value is greater than the prediction threshold, it indicates that the first prediction value presumes that the target case is a normal claim settlement. If the first prediction value is less than or equal to the prediction threshold, it indicates that the first prediction value presumes that the target case is an abnormal claim settlement. Similarly, compare the second prediction value with the preset prediction threshold. If the second prediction value is greater than the prediction threshold, it indicates that the second prediction value presumes that the target case is a normal claim settlement. If the second prediction value is less than or equal to the prediction threshold, it indicates that the second prediction value presumes that the target case is an abnormal claim settlement.

[0139] Further, when the first prediction value presumes that the target case is a normal claim settlement and the second prediction value presumes that the target case is a normal claim settlement, the target claim settlement data is: determining that the target case is a normal claim settlement; or, when the first prediction value presumes that the target case is an abnormal claim settlement and the second prediction value presumes that the target case is an abnormal claim settlement, the target claim settlement data is: determining that the target case is an abnormal claim settlement.

[0140] Further, when the first prediction value presumes that the target case is a normal claim settlement and the second prediction value presumes that the target case is an abnormal claim settlement, the target claim settlement data is: determining that the target case is an abnormal claim settlement; or, when the first prediction value presumes that the target case is an abnormal claim settlement and the second prediction value presumes that the target case is a normal claim settlement, the target claim settlement data is: determining that the target case is a normal claim settlement.

[0141] The claim prediction method according to the embodiment of the present application obtains the case information data and case description data of the target case. Among them, the case information data is used to record the basic case information of the target case, and the case description data is text data, which is used to describe the case situation of the target case, and can perform claim prediction based on the basic case information and the case situation. Further, feature extraction is performed on the case information data to obtain a case feature vector, and node prediction is performed on the case feature vector through a preset tree model to obtain target tree node data and a first prediction value. The tree model can meet the requirements of processing numerical features and discrete features in the case information data, and can identify relatively complex semantic information in the case information data, improving the comprehensiveness of feature processing and the accuracy of semantic recognition. Further, encoding processing is performed on the case description data to obtain a target semantic hidden vector, and prediction processing is performed on the target tree node data and the target semantic hidden vector through a pre-trained claim prediction model to obtain a second prediction value, which can combine the basic case information and the case situation of the target case for prediction, improving the diversity and comprehensiveness of the feature information used for claim prediction. Finally, claim calculation is performed according to the first prediction value and the second prediction value to obtain target claim data, which can perform claim analysis by comprehensively considering the prediction situations of the tree model and the claim prediction model, making the target claim data more reasonable and objective, thereby improving the accuracy of judging the claim situation.

[0142] Please refer to Figure 8 , the embodiment of the present application also provides a claim prediction device, which can implement the above claim prediction method. The device includes:

[0143] A data acquisition module 801, configured to acquire the case information data and case description data of the target case. Among them, the case information data is used to record the basic case information of the target case, and the case description data is used to describe the case situation of the target case;

[0144] A feature extraction module 802, configured to perform feature extraction on the case information data to obtain a case feature vector;

[0145] A node prediction module 803, configured to perform node prediction on the case feature vector through a preset tree model to obtain target tree node data and a first prediction value;

[0146] An encoding module 804, configured to perform encoding processing on the case description data to obtain a target semantic hidden vector;

[0147] A claim prediction module 805, configured to perform prediction processing on the target tree node data and the target semantic hidden vector through a pre-trained claim prediction model to obtain a second prediction value;

[0148] A calculation module 806, configured to perform claim calculation according to the first prediction value and the second prediction value to obtain target claim data.

[0149] In some embodiments, the case information data includes first discrete data, second discrete data, first numerical data, and second numerical data. The feature extraction module 802 includes:

[0150] A first encoding unit for encoding the first discrete data according to a preset encoding sequence number to obtain a first encoding vector;

[0151] A second encoding unit for performing multi-hot encoding on the second discrete data to obtain a second encoding vector;

[0152] A third encoding unit for encoding the first numerical data according to preset padding parameters to obtain a third encoding vector;

[0153] A fourth encoding unit for extracting numerical features from the second numerical data to obtain target numerical features, and encoding the target numerical features to obtain a fourth encoding vector;

[0154] A vector splicing unit for splicing the first encoding vector, the second encoding vector, the third encoding vector, and the fourth encoding vector to obtain a case feature vector.

[0155] In some embodiments, the node prediction module 803 includes:

[0156] A node generation unit for generating nodes from the case feature vector through a tree model to obtain target leaf nodes;

[0157] A combination unit for combining the target leaf nodes to obtain target tree node data;

[0158] A probability calculation unit for calculating the probability of the target tree node data through a preset first function to obtain a first prediction value.

[0159] In some embodiments, the encoding module 804 includes:

[0160] A format conversion unit for converting the format of the case description data to obtain an initial description text;

[0161] A segmentation unit for segmenting the initial description text according to a preset text length to obtain a target description text;

[0162] A fifth encoding unit for encoding the target description text through a preset encoding model to obtain a target semantic hidden vector.

[0163] In some embodiments, the case description data includes first text data, second text data, and third text data. The format conversion unit includes:

[0164] The first conversion subunit is configured to perform format conversion on the first text data according to a preset first format template to obtain a first description text;

[0165] The second conversion subunit is configured to perform format conversion on the second text data according to a preset second format template to obtain a second description text;

[0166] The third conversion subunit is configured to perform format conversion on the third text data according to a preset third format template to obtain a third description text;

[0167] The text splicing unit is configured to splice and process the first description text, the second description text, and the third description text according to a preset splicing order to obtain an initial description text.

[0168] In some embodiments, the claim prediction model includes an encoding layer and a prediction layer. The claim prediction module 805 includes:

[0169] The sixth encoding unit is configured to perform encoding processing on the target tree node data through the encoding layer to obtain a tree node encoding vector;

[0170] The prediction unit is configured to perform claim prediction on the tree node encoding vector and the target semantic hidden vector through the prediction layer to obtain a second prediction value.

[0171] In some embodiments, the prediction unit includes:

[0172] The vector calculation subunit is configured to perform vector addition on the tree node encoding vector and the target semantic hidden vector according to preset weight parameters to obtain a target case hidden vector;

[0173] The claim rate calculation subunit is configured to calculate the claim rate of the target case hidden vector through a second function of the prediction layer to obtain a second prediction value.

[0174] The specific implementation manner of this claim prediction device is basically the same as the specific embodiments of the above claim prediction method, and will not be elaborated here.

[0175] An embodiment of this application further provides an electronic device, which includes: a memory, a processor, a program stored on the memory and executable on the processor, and a data bus for realizing connection communication between the processor and the memory. When the program is executed by the processor, the above claim prediction method is implemented. This electronic device can be any intelligent terminal including a tablet computer, a vehicle-mounted computer, etc.

[0176] Please refer to Figure 9 , Figure 9 which illustrates the hardware structure of an electronic device in another embodiment. The electronic device includes:

[0177] The processor 901 can be implemented in the form of a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, etc., and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0178] The memory 902 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM), etc. The memory 902 can store an operating system and other application programs. When implementing the technical solutions provided in the embodiments of this specification through software or firmware, the relevant program codes are stored in the memory 902 and are called by the processor 901 to execute the claim prediction method of the embodiments of the present application;

[0179] The input / output interface 903 is used to implement information input and output;

[0180] The communication interface 904 is used to implement communication interaction between this device and other devices, and can implement communication through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.);

[0181] The bus 905 transmits information between the various components of the device (such as the processor 901, the memory 902, the input / output interface 903, and the communication interface 904);

[0182] Among them, the processor 901, the memory 902, the input / output interface 903, and the communication interface 904 achieve communication connections with each other inside the device through the bus 905.

[0183] The embodiments of the present application also provide a storage medium. The storage medium is a computer-readable storage medium for computer-readable storage. The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the above-mentioned claim prediction method.

[0184] The memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. In addition, the memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory optionally includes a memory remotely disposed relative to the processor, and these remote memories can be connected to the processor through a network. Examples of the above networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0185] The claim prediction method, claim prediction device, electronic device, and storage medium provided by the embodiments of the present application obtain case information data and case description data of a target case, wherein the case information data is used to record the basic case information of the target case, and the case description data is text data used to describe the case situation of the target case, and can perform claim prediction based on the basic case information and the case situation. Further, feature extraction is performed on the case information data to obtain a case feature vector, and node prediction is performed on the case feature vector through a preset tree model to obtain target tree node data and a first prediction value. The tree model can meet the requirements of processing numerical features and discrete features in the case information data, can better handle sparse features and missing values, and can identify relatively complex semantic information in the case information data, improving the comprehensiveness of feature processing and the accuracy of semantic recognition, making the overall solution have better robustness. Further, encoding processing is performed on the case description data to obtain a target semantic hidden vector, and prediction processing is performed on the target tree node data and the target semantic hidden vector through a pre-trained claim prediction model to obtain a second prediction value, which can perform prediction by combining the basic case information and the case situation of the target case, improving the diversity and comprehensiveness of the feature information for claim prediction. Finally, claim calculation is performed according to the first prediction value and the second prediction value to obtain target claim data, which can perform claim analysis by comprehensively considering the prediction situations of the tree model and the claim prediction model, process discrete features and numerical features through the tree model, and encode text features through a preset encoding model, thereby realizing the modeling of case text information and basic case information, and using different models to model different types of case data, making the obtained target claim data have better rationality and objectivity, and thus improving the accuracy of judging claim situations.

[0186] The embodiments described in the embodiments of the present application are for more clearly illustrating the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those skilled in the art know that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of the present application are equally applicable to similar technical problems.

[0187] Those skilled in the art can understand that Figure 1-7 the technical solutions shown in

[0188] do not constitute a limitation to the embodiments of the present application, and may include more or fewer steps than those illustrated, or combine certain steps, or different steps.

[0189] Those of ordinary skill in the art can understand that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, and their appropriate combinations.

[0190] The terms "first", "second", "third", "fourth", etc. (if any) in the description of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices.

[0191] It should be understood that in the present application, "at least one (item)" means one or more, and "a plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships can exist. For example, "A and / or B" can represent: only A exists, only B exists, and both A and B exist at the same time. Among them, A and B can be singular or plural. The character " / " generally represents an "or" relationship between the associated objects before and after. "At least one (one) of the following" or its similar expression refers to any combination of these items, including any combination of single item (one) or plural items (ones). For example, at least one (one) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0192] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the above-mentioned units is only a logical function division. In actual implementation, there may be other division methods. For example, 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 displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of devices or units can be in electrical, mechanical or other forms.

[0193] The units described above as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0194] In addition, in each embodiment of the present application, the functional units can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0195] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in each embodiment of the present application. The foregoing storage medium includes: various media that can store programs such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs.

[0196] The preferred embodiments of the embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of rights of the embodiments of the present application. Any modifications, equivalent replacements, and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of the present application shall be within the scope of rights of the embodiments of the present application.

Claims

1. A claim prediction method, characterized in that, The method includes: Obtaining case information data and case description data of a target case, where the case information data is used to record the basic case information of the target case, and the case description data is used to describe the details of the target case; Performing feature extraction on the case information data to obtain a case feature vector; Performing node prediction on the case feature vector through a preset tree model to obtain target tree node data and a first prediction value; Performing encoding processing on the case description data to obtain a target semantic hidden vector; Performing prediction processing on the target tree node data and the target semantic hidden vector through a pre-trained claim prediction model to obtain a second prediction value; Performing claim calculation based on the first prediction value and the second prediction value to obtain target claim data; The claim prediction model includes an encoding layer and a prediction layer. The step of performing prediction processing on the target tree node data and the target semantic hidden vector through the pre-trained claim prediction model to obtain a second prediction value includes: Performing encoding processing on the target tree node data through the encoding layer to obtain a tree node encoding vector; Performing vector addition on the tree node encoding vector and the target semantic hidden vector according to preset weight parameters to obtain a target case hidden vector; Calculating the claim rate of the target case hidden vector through a second function of the prediction layer to obtain the second prediction value.

2. The claim settlement prediction method according to claim 1, wherein The step of performing node prediction on the case feature vector through the preset tree model to obtain target tree node data and a first prediction value includes: Generating nodes for the case feature vector through the tree model to obtain target leaf nodes; Performing combination processing on the target leaf nodes to obtain the target tree node data; Calculating the probability of the target tree node data through a preset first function to obtain the first prediction value.

3. The claim settlement prediction method according to claim 1, wherein The case information data includes first discrete data, second discrete data, first numerical data, and second numerical data. The step of performing feature extraction on the case information data to obtain a case feature vector includes: Performing encoding processing on the first discrete data according to a preset encoding sequence number to obtain a first encoding vector; Performing multi-hot encoding processing on the second discrete data to obtain a second encoding vector; Performing encoding processing on the first numerical data according to preset padding parameters to obtain a third encoding vector; Extracting numerical features of the second numerical data to obtain target numerical features, and performing encoding processing on the target numerical features to obtain a fourth encoding vector; Performing vector splicing on the first encoding vector, the second encoding vector, the third encoding vector, and the fourth encoding vector to obtain the case feature vector.

4. The claim settlement prediction method according to any one of claims 1 to 3, characterized in that, The step of performing encoding processing on the case description data to obtain a target semantic hidden vector includes: Performing format conversion on the case description data to obtain an initial description text; Performing segmentation processing on the initial description text according to a preset text length to obtain a target description text; Performing encoding processing on the target description text through a preset encoding model to obtain the target semantic hidden vector.

5. The claim settlement prediction method according to claim 4, wherein The case description data includes first text data, second text data, and third text data. The step of performing format conversion on the case description data to obtain an initial description text includes: Performing format conversion on the first text data according to a preset first format template to obtain a first description text; Performing format conversion on the second text data according to a preset second format template to obtain a second description text; Performing format conversion on the third text data according to a preset third format template to obtain a third description text; Performing splicing processing on the first description text, the second description text, and the third description text according to a preset splicing order to obtain the initial description text.

6. A claims settlement prediction device, characterized in that, The device includes: A data acquisition module for acquiring case information data and case description data of a target case, where the case information data is used to record the basic case information of the target case, and the case description data is used to describe the case situation of the target case; A feature extraction module for extracting features from the case information data to obtain a case feature vector; A node prediction module for performing node prediction on the case feature vector through a preset tree model to obtain target tree node data and a first prediction value; An encoding module for performing encoding processing on the case description data to obtain a target semantic hidden vector; A claim prediction module for performing prediction processing on the target tree node data and the target semantic hidden vector through a pre-trained claim prediction model to obtain a second prediction value; A calculation module for performing claim calculation based on the first prediction value and the second prediction value to obtain target claim data; The claim prediction model includes an encoding layer and a prediction layer. The step of performing prediction processing on the target tree node data and the target semantic hidden vector through the pre-trained claim prediction model to obtain a second prediction value includes: Performing encoding processing on the target tree node data through the encoding layer to obtain a tree node encoding vector; Performing vector addition on the tree node encoding vector and the target semantic hidden vector according to preset weight parameters to obtain a target case hidden vector; Calculating a claim rate for the target case hidden vector through a second function of the prediction layer to obtain the second prediction value.

7. An electronic device, characterized in that, The electronic device includes a memory, a processor, a program stored on the memory and executable on the processor, and a data bus for implementing connection communication between the processor and the memory. When the program is executed by the processor, the steps of the claim prediction method according to any one of claims 1 to 5 are implemented.

8. A storage medium, the storage medium being a computer-readable storage medium for computer-readable storage, characterized in that, The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of the claim prediction method according to any one of claims 1 to 5.

Images