Construction method and use method of detection model and related products

Abstract

The invention discloses a construction method and a use method of a detection model, and a related product. The method comprises the steps of obtaining claim settlement data; performing preprocessing and feature extraction on the claim settlement data to obtain a medical service entity serving as a root node and service features serving as child nodes; according to a relationship defined in the claim settlement data, constructing a directed edge to connect the root node and the service feature node to obtain a graph data structure; detecting whether mismatching exists between the medical entity and the service characteristics of the medical entity by using a weighted multi-tree and the graph data structure; using a density-based clustering algorithm to detect the claim behavior of the medical entity with the medical service features meeting a preset similarity standard; and based on a result obtained by detection, constructing a detection model used for identifying the non-compliance behavior in the claim settlement data. The weighted multi-tree algorithm and the density-based clustering algorithm are combined for use, so that the model can better adapt to the complexity and variability of the claim discompliance behavior.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method for constructing a detection model, a method for using the detection model, and related products. Background Art

[0002] In the field of health insurance, claim irregularity detection is a critical task that aims to identify and prevent irregular claims behaviors that are intended to illegally obtain insurance compensation. However, most current health insurance claim irregularity detection models rely mainly on supervised learning methods such as decision trees, random forests, and neural networks. The core of these methods is to train models using labeled data that clearly indicates which claims cases are irregular and which are normal.

[0003] Although supervised learning methods have achieved certain results in the field of claim non-compliance detection, they face many challenges. First, claim non-compliance is characterized by complexity and variability, which makes the pattern of claim non-compliance difficult to grasp, and the means of claim non-compliance are constantly updated, which puts a great test on the recognition ability of the model. Secondly, obtaining labeled claim non-compliance data from the medical field is an extremely difficult task. The diversity of medical data makes data collection, integration and labeling extremely complex, and the strict restrictions on medical privacy and compliance further increase the difficulty of data acquisition. These limitations not only affect the recognition accuracy and generalization ability of the model, but also limit the effectiveness and reliability of the model in actual application scenarios. Summary of the invention

[0004] Based on the above problems, the present application provides a method for constructing a detection model, a method for using the model, and related products.

[0005] The embodiments of the present application disclose the following technical solutions:

[0006] In a first aspect, an embodiment of the present application provides a method for constructing a detection model, comprising:

[0007] Obtain claims data;

[0008] Preprocessing and feature extraction are performed on the claim data to obtain a medical service entity as a root node and service features as child nodes;

[0009] According to the relationship defined in the claim data, a directed edge is constructed to connect the root node and the service feature node to obtain a graph data structure, wherein the graph data structure is used to represent the association relationship between different medical service entities and service features;

[0010] Using the weighted multi-tree and the graph data structure, detecting whether there is a mismatch between the medical entity and its service characteristics; using a density-based clustering algorithm, detecting the claim behavior of the medical entity with medical service characteristics that meet the preset similarity standard;

[0011] Based on the detection results, a detection model is built to identify non-compliant behaviors in claims data.

[0012] In a possible implementation, the detecting whether there is a mismatch between a medical entity and its service characteristics by using the weighted multi-tree and the graph data structure includes:

[0013] Calculate the frequency of occurrence of each feature node in the claims data;

[0014] Assign occurrence frequencies to edges between features;

[0015] Calculating a data set weight for each medical service entity, wherein the weight is determined based on the sum of edge frequencies of all feature nodes;

[0016] Calculating a detection threshold, where the detection threshold is the sum of the maximum weight and the minimum weight of a preset number of medical service entities divided by the total number of claims;

[0017] If the data set weight of the detected medical service entity is lower than the detection threshold, it is determined that there is a mismatch in the detected medical service entity, and the service associated with the detected medical service entity is marked as a detection case.

[0018] In a possible implementation, the method of using a density-based clustering algorithm to detect claims behavior of medical entities having medical service characteristics that meet preset similarity standards includes:

[0019] Grouping healthcare entities based on healthcare service characteristics;

[0020] Applying a density-based clustering algorithm on each group to perform cluster analysis on claim behaviors, wherein the claim behaviors include the number of claims and the amount of claims;

[0021] Identify abnormal claim behaviors in the clustering results and mark them as detection cases.

[0022] In a possible implementation, the step of constructing a detection model based on the detection result includes:

[0023] Based on the test results, the results are compared with the preset benchmark cases;

[0024] Based on the comparison results, model parameters of the detection model are adjusted.

[0025] In a possible implementation, the preprocessing and feature extraction of the claim data to obtain a medical service entity as a root node and service features as child nodes includes:

[0026] Cleansing the claims data to remove invalid or incomplete records;

[0027] Standardize the cleaned claims data to ensure consistency in data format;

[0028] The medical service entity is extracted from the standardized claims data as the root node, and the related service features are extracted as child nodes.

[0029] In a second aspect, an embodiment of the present application provides a method for using a detection model, including:

[0030] Obtain the claims data to be tested;

[0031] The claim data to be detected is input into a detection model to obtain a detection result, and the detection model is obtained by the method described in the first aspect.

[0032] In a third aspect, an embodiment of the present application provides a system for constructing a detection model, including:

[0033] An acquisition unit, used for acquiring claim data;

[0034] An extraction unit, used for preprocessing and feature extraction of the claim data, to obtain a medical service entity as a root node and a service feature as a child node;

[0035] A first construction unit is used to construct directed edges to connect the root node and service feature nodes according to the relationship defined in the claim data to obtain a graph data structure, wherein the graph data structure is used to represent the association relationship between different medical service entities and service features;

[0036] A detection unit is used to detect whether there is a mismatch between a medical entity and its service characteristics using a weighted multi-tree and the graph data structure; and to detect the claim behavior of a medical entity having medical service characteristics that meet a preset similarity standard using a density-based clustering algorithm;

[0037] The second construction unit is used to construct a detection model for identifying non-compliant behaviors in claims data based on the detection results.

[0038] In a fourth aspect, an embodiment of the present application provides an electronic device, characterized in that it includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, it implements the method described in the first aspect above.

[0039] In a fifth aspect, an embodiment of the present application provides a computer program product. When the computer program product runs on a computer, the computer executes the method described in the first aspect.

[0040] In a sixth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein instructions are stored in the computer-readable storage medium. When the instructions are executed on a terminal device, the terminal device executes the method described in the first aspect.

[0041] Compared with the prior art, this application has the following beneficial effects:

[0042] By constructing a graph data structure, the relationship between medical service entities and service features is intuitively represented, providing a more accurate information basis for mismatch detection and cluster analysis. The combined use of weighted multi-tree algorithm and density-based clustering algorithm enables the model to better adapt to the complexity and variability of claim non-compliance behavior and improve the accuracy and robustness of detection. This method does not rely on a large amount of labeled data, but instead uses graph data structures and algorithms for detection, thereby reducing the dependence on labeled data and the difficulty of obtaining it. Through real-time monitoring and cluster analysis, the model can promptly discover and identify claim non-compliance behavior, improving the efficiency and response speed of claim non-compliance behavior detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present application 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, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0044] Figure 1 A flowchart of a method for constructing a detection model provided in an embodiment of the present application;

[0045] Figure 2 A schematic diagram of module interaction provided in an embodiment of the present application;

[0046] Figure 3 A structural diagram of a detection model construction system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0047] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0048] To facilitate understanding of the technical solutions provided by the embodiments of the present application, the background technology involved in the embodiments of the present application will be described below.

[0049] As mentioned above, most existing health insurance claim non-compliance detection models use supervised learning methods, such as decision trees, random forests, and neural networks. These methods rely on labeled data (i.e., it is known which claims are non-compliant and which are normal) for model training. However, since claim non-compliance is complex and changeable, and labeled data is difficult to obtain, the application of these models is limited.

[0050] The diversity of medical data, the difficulty in obtaining labeled data from the medical field, and the lag in labeled data have led to many limitations in supervised models.

[0051] Most claims irregularity detection systems rely on records audited by professionals and rely on medical experts to determine the core feature sets and train models. However, audited medical records are very limited, and the process of relying on experts will slow down the development of claims irregularity detection systems.

[0052] Unsupervised models usually do not group homogeneous claims when analyzing medical claims, which can easily lead to false positives. Some existing models use clustering techniques and neural networks to group similar claims, but they produce intermediate results when distinguishing different claim groups, leading to false alarms. In addition, some models focus mainly on categorical data and ignore continuous variables (such as the number and amount of medical services), which are also important in identifying claims non-compliance.

[0053] In order to solve the above problems, the embodiment of the present application is divided into two stages, which analyze multivariate categorical data and continuous data respectively. The first stage uses a weighted multi-tree (WMT) model to analyze the similarities, relationships and services of medical institutions in medical data to identify false services. The second stage detects false claims by developing multiple density-based clustering (DBC) techniques on continuous data.

[0054] First, the anti-claim non-compliance behavior model proposed in this application adopts unsupervised learning technology at all stages and does not rely on any audited or labeled data, thereby reducing the workload of data labeling and expanding the scope of available data. Second, the algorithm of this application is based on a graph method, which uses a weighted multi-tree (WMT) to group similar claims, avoiding the interference of intermediate results on the final classification and improving the accuracy of the model. Third, the model of this application can perform multivariate analysis on categorical data and continuous data commonly found in medical insurance claims, and by combining the DBSCAN clustering algorithm, it improves the accuracy of detecting claims for non-compliant behavior. Fourth, since most of the service feature data of different medical service entities are the same, the construction of traditional graphs will be redundant and the construction cost is high. The model proposed in this application is developed based on the multi-tree (MT) property of a directed acyclic graph, which reduces the construction cost by avoiding the creation of redundant nodes.

[0055] It should be noted that the construction method, system, device and medium of the detection model provided in the present application can be applied to the field of computer technology. The above is only an example, and the application field of the construction method, system, device and medium of the detection model provided in the present application is not limited. In addition, the embodiments of the present application may not limit the execution subject of the construction of the detection model. For example, the construction method of the detection model of the embodiment of the present application can be applied to data processing devices such as terminal devices or servers. Among them, the terminal device can be an electronic device such as a computer, a personal digital assistant (Personal Digital Assistant, PDA). The server can be an independent server, a cloud server or a cluster server composed of multiple servers.

[0056] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0057] The following is an example of an embodiment to illustrate a method for constructing a detection model provided by the present application. Figure 1 ,Should Figure 1 A flowchart of a method for constructing a detection model provided in an embodiment of the present application, the method comprising:

[0058] S101. Obtain claims data.

[0059] Import the medical and health insurance claims data required by the model. This data usually comes from the database of the medical insurance company and contains a large number of claims records. Each record records the patient's identity information, the provision of medical services, cost details and other key information. The imported data needs to be preprocessed to ensure the accuracy and completeness of the data, providing a solid foundation for subsequent feature processing and graph data construction.

[0060] S102: Preprocess and extract features of the claim data to obtain a medical service entity as a root node and service features as child nodes.

[0061] After the data is imported, feature processing is required. Feature processing refers to extracting features from the raw data that are useful for detecting claims non-compliance behavior. In this application, feature processing focuses on the medical service provider entities (such as hospitals, clinics, etc.) and the service characteristics they provide (such as qualifications, service locations, service types, etc.). These features are regarded as nodes in the dataset, with the medical service provider entity as the root node and other feature nodes entered in the same order. The purpose of feature processing is to convert the raw data into a form that the model can understand and process for subsequent claims non-compliance detection.

[0062] S103: According to the relationship defined in the claim data, a directed edge is constructed to connect the root node and the service feature node to obtain a graph data structure. The graph data structure is used to represent the association relationship between different medical service entities and service features.

[0063] After completing feature processing, it is necessary to prepare the edge nodes of the graph data. This step mainly connects different feature nodes to construct directed edges based on the relationships defined in the claim data. In this application, directed edges represent the association between different feature nodes, such as the association between the medical service provider entity and its service features, the similarity between service features, etc. By constructing directed edges, a complete graph data structure can be formed, which facilitates subsequent analysis and detection.

[0064] After feature processing and graph data edge node preparation are completed, a data set (π) can be constructed. In this application, the data set (π) consists of different medical service provider entities (p) and their corresponding service feature nodes. Each medical service provider entity is used as a root node in the data set, and other feature nodes are input in the same order. Since feature nodes can be shared by different root nodes, the data set (π) can more comprehensively reflect the relationship between medical service providers and their service features.

[0065] In the implementation scenario, the above process can be performed by the data module, which is the basis and premise of the entire claim non-compliance behavior detection model. Through data import, feature processing, graph data edge node preparation and data set (π) construction, the data module provides the necessary input and support for the subsequent grouping recognition module and claim non-compliance behavior detection module. The quality of the data module directly determines the accuracy and reliability of the entire model.

[0066] S104. Using the weighted multi-tree and the graph data structure, detect whether there is a mismatch between the medical entity and its service characteristics; using a density-based clustering algorithm, detect the claim behavior of the medical entity with medical service characteristics that meet the preset similarity standards.

[0067] The weighted multi-tree (WMT) method is used to analyze the frequent patterns of services submitted by medical service providers with specific profile information, with the aim of observing the relationship between the provider's profile and the services they provide. This process is based on a directed multi-tree structure, which implements weighted analysis by assigning frequencies to edges between feature nodes in the dataset.

[0068] The weighted multi-tree (WMT) method is an analysis method based on the multi-tree (MT) structure and combined with the weight characteristics of the data. The multi-tree (MT) itself is a tree-shaped data structure that can clearly display the hierarchy and association relationship between nodes. On this basis, the weighted multi-tree further considers the importance (i.e., weight) of each node (or edge), thereby more accurately reflecting the actual distribution of the data.

[0069] For the data set compiled based on the information obtained in the above process, each medical service entity is used as a root node, and its related service features are used as child nodes or edges. According to the number of occurrences of each feature node in the claims data, a frequency value is assigned to the edge between the features. With the medical service entity as the root node, a weighted multi-tree is constructed based on the service features it provides. Each path from the root node to the leaf node represents a set of details for a specific provider type. This step helps determine the relative importance of the feature nodes on each path from the root node to the leaf node.

[0070] Assume that πk is a dataset with k features and πkp is a dataset of medical service entity p. In this module, the edge frequency is defined as f. Then the sum of the edge frequencies of all feature nodes of the dataset πkp of medical service entity p is as shown in the following equation:

[0071]

[0072] The weight of the dataset of each institution p can be expressed as:

[0073]

[0074] The threshold is calculated as the sum of the maximum weight and the minimum weight of the dataset of N medical service entities p divided by the total number of claims, as shown in the following equation:

[0075]

[0076] By calculating the threshold using the sum of the maximum and minimum weights, the need to detect potential claim non-compliance can be balanced with the need to avoid false positives. The threshold calculated in this way balances the need to detect potential claim non-compliance with the need to avoid false positives. This approach ensures that the threshold is sensitive enough to capture potential claim non-compliance, but not so low that normal differences are incorrectly marked as claim non-compliance. If the weight of the dataset of a medical service entity p is lower than the threshold calculated above, it is considered that there is a mismatch between the entity and its service characteristics, and the relevant service can be defined as a claim non-compliant behavior service. This practice helps to identify behaviors that do not conform to normal service patterns, thereby improving the accuracy of claim non-compliance detection.

[0077] By introducing the weighted multi-tree method, potential non-compliant claims in medical services can be more effectively identified. This method not only considers the frequency distribution of service characteristics, but also enhances the sensitivity and accuracy of the model through weighted analysis. In addition, by calculating the threshold to balance the need for detecting non-compliant claims and the need to avoid false positives, the model has higher reliability and practicality in practical applications.

[0078] DBSCAN is a density-based clustering algorithm that can find clusters of arbitrary shapes in noisy spatial data. The algorithm identifies clusters by calculating the density connectivity between data points and marks those data points that are not dense enough to form clusters as noise. This gives DBSCAN a significant advantage in processing data with complex distribution characteristics.

[0079] Choose appropriate parameters (such as ε and MinPts) to define the density connectivity between data points. Among them, ε represents the maximum distance between data points, and MinPts represents the minimum number of data points required for a cluster. Iterate through the data set and calculate the number of neighbors for each data point based on ε and MinPts.

[0080] Data points are divided into core points, boundary points, and noise points according to the number of neighbors. Core points are those data points whose number of neighbors is greater than or equal to MinPts, boundary points are those data points whose number of neighbors is less than MinPts but are neighbors of a core point, and noise points are those data points that are neither core points nor boundary points. Clusters are constructed based on core points and boundary points. For each core point, find all its density-reachable data points and classify them into the same cluster.

[0081] The DBSCAN algorithm can identify clusters of data points that are significantly different from other normal service patterns. This includes possible unnecessary services or abnormally high costs. Even if the number of services and claim amounts are distributed in distributions of different shapes, DBSCAN can accurately cluster similar claims. Once the clusters are formed, it is possible to determine whether there is claim irregularity based on the characteristics of the data within each cluster. For example, if a cluster contains a large number of high-value claims but a relatively single type of service, this may be a sign of claim irregularity.

[0082] S105. Based on the detection results, a detection model is constructed to identify non-compliant behaviors in the claims data.

[0083] Obtain a dataset of known real claim non-compliance cases that have been verified to be actual claim non-compliance cases. This can be obtained through historical records, expert review, or other reliable sources. Compare the potential claim non-compliance cases detected in both phases with the collected real claim non-compliance cases. Through comparative analysis, determine how many of the detected potential claim non-compliance cases are real claim non-compliance cases and how many are false positives or negatives.

[0084] Other relevant performance indicators can be calculated as needed, such as specificity, confusion matrix, etc. The following are several performance indicators that can be used:

[0085] Calculate the accuracy: the ratio of correctly classified samples to the total number of samples.

[0086] Calculate the recall rate: the proportion of all actual claim non-compliance cases that are correctly identified as claim non-compliance.

[0087] Calculate the precision: the proportion of all cases marked as claim non-compliance that are actually claim non-compliance.

[0088] Calculate the F1 score: the harmonic mean of precision and recall, which is used to comprehensively evaluate the performance of the model.

[0089] Based on the above performance indicators, comprehensively evaluate the overall performance of the model. Identify the strengths and weaknesses of the model in different aspects, such as it performs well in detecting certain types of claim non-compliance behaviors, while other types may need improvement.

[0090] In one possible implementation, the model parameters can be fine-tuned based on the evaluation results to improve the model's accuracy, recall, and other key performance indicators. Cross-validation and other methods can also be used to ensure that the model has good generalization capabilities and avoid overfitting.

[0091] In one possible implementation, model performance can be re-evaluated periodically to account for new claims non-compliance measures and changing data distributions. Based on the latest claims non-compliance cases and feedback, the model can be updated to maintain its effectiveness.

[0092] In a possible implementation, the above steps may be executed and implemented by different modules. Figure 2 As shown, Figure 2 A schematic diagram of module interaction provided for an embodiment of the present application, including a database icon on the left and a flowchart box on the right, connected by arrows. There are multiple rectangular boxes in the flowchart box, representing different steps or modules, and the arrows indicate the order of the process. The following is an explanation and analysis based on the content in the figure:

[0093] The process starts with a cylindrical icon, which represents the data source. This usually represents a repository for claims data, such as a database or data warehouse. Data flows from the data source to the flowchart on the right, which consists of multiple rectangular boxes and arrows, representing the various steps of data processing. The data module is the cornerstone of the entire anti-fraud model. It is mainly responsible for importing claims data for medical and health insurance, and performing feature processing and preparation of graph data edge nodes. This module receives claims data from different medical service entities, such as "Pathology" and "Internal Medicine", which are regarded as the root nodes of the data set. Under each root node, there is an associated series of feature nodes, such as qualifications, service locations, etc. These feature nodes are entered in the same order and can be shared by different root nodes. In the data preparation stage, the module connects these feature nodes through directed edges according to the relationships defined in the claims data to build a complete graph data structure.

[0094] The group identification module is a key part of the model. It introduces the weighted multi-tree (WMT) method to analyze the frequent patterns of services provided by medical service providers with specific profile information. This step aims to observe the relationship between the provider profile and the services provided to detect potential fraud. In specific implementation, the module assigns corresponding weights (i.e., frequency of occurrence) to the edges between features based on the frequency of occurrence of feature nodes in the claims data. Each path from the root node to the leaf node represents a detailed feature set of a provider type. By calculating the weight of each medical service entity data set and setting a threshold to balance the sensitivity and false alarm rate of fraud detection, the module is able to identify service features that do not match specific medical entities, thereby defining them as fraudulent services.

[0095] The fraud detection module further analyzes the number of claims and the amount of claims for medical entities with similar medical service characteristics. This module applies the density-based clustering algorithm DBSCAN to perform cluster analysis on groups with similar characteristics. The DBSCAN method can more effectively define the correlation between data points, thereby identifying unnecessary services and excessive charges. Even if the number of services and the amount of claims are distributed in any shape of distribution, DBSCAN can cluster similar claims to help identify potential fraud.

[0096] Finally, the fraud cases detected by the two stages (grouping recognition module and fraud detection module) are comprehensively compared with the real fraud cases; the performance indicators such as accuracy, recall rate, precision, etc. of the evaluation model are evaluated by comparison; the model parameters are adjusted according to the evaluation results to optimize the model performance.

[0097] The above are some specific implementations of the detection model construction method provided in the embodiment of the present application. Based on this, the present application also provides a corresponding detection model construction system. The system provided in the embodiment of the present application will be introduced from the perspective of functional modularization. Figure 3 A structural diagram of a detection model construction system provided in an embodiment of the present application.

[0098] The system comprises:

[0099] An acquisition unit 110, used to acquire claim data;

[0100] An extraction unit 111 is used to pre-process and extract features from the claim data to obtain a medical service entity as a root node and service features as child nodes;

[0101] A first construction unit 112 is used to construct directed edges to connect the root node and service feature nodes according to the relationship defined in the claim data to obtain a graph data structure, wherein the graph data structure is used to represent the association relationship between different medical service entities and service features;

[0102] The detection unit 113 is used to detect whether there is a mismatch between the medical entity and its service characteristics using the weighted multi-tree and the graph data structure; and to detect the claim behavior of the medical entity having medical service characteristics that meet the preset similarity standard using a density-based clustering algorithm;

[0103] The second construction unit 114 is used to construct a detection model for identifying non-compliant behaviors in the claims data based on the detection results.

[0104] The embodiments of the present application also provide corresponding devices and computer storage media for implementing the solutions provided by the embodiments of the present application.

[0105] The device includes a memory and a processor, the memory is used to store instructions or codes, and the processor is used to execute the instructions or codes so that the device executes the method described in any embodiment of the present application.

[0106] The computer storage medium stores codes, and when the codes are executed, a device executing the codes implements the method described in any embodiment of the present application.

[0107] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the various embodiments can be referred to each other. For the system or device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0108] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: 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.

[0109] It should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0110] It should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0111] It should also be noted that, in this application, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants 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, the elements defined by the statement "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0112] The steps of the method or algorithm described in conjunction with the embodiments disclosed in this application can be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module can be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0113] The above description of the disclosed embodiments enables professionals and technicians in the field to implement or use the present application. Various modifications to these embodiments will be apparent to professionals and technicians in the field, and the general principles defined in this application can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown in the present application, but will conform to the widest range consistent with the principles and novel features disclosed in the present application.

Claims

1. A method for constructing a detection model, characterized in that: include: Obtain claims data; Preprocessing and feature extraction are performed on the claim data to obtain a medical service entity as a root node and service features as child nodes; According to the relationship defined in the claim data, a directed edge is constructed to connect the root node and the service feature node to obtain a graph data structure, wherein the graph data structure is used to represent the association relationship between different medical service entities and service features; Using the weighted multi-tree and the graph data structure, detecting whether there is a mismatch between the medical entity and its service characteristics; using a density-based clustering algorithm, detecting the claim behavior of the medical entity with medical service characteristics that meet the preset similarity standard; Based on the detection results, a detection model is built to identify non-compliant behaviors in claims data.

2. The method according to claim 1, characterized in that The method of using the weighted multi-tree and the graph data structure to detect whether there is a mismatch between a medical entity and its service characteristics includes: Calculate the frequency of occurrence of each feature node in the claims data; Assign occurrence frequencies to edges between features; Calculating a data set weight for each medical service entity, wherein the weight is determined based on the sum of edge frequencies of all feature nodes; Calculating a detection threshold, where the detection threshold is the sum of the maximum weight and the minimum weight of a preset number of medical service entities divided by the total number of claims; If the data set weight of the detected medical service entity is lower than the detection threshold, it is determined that there is a mismatch in the detected medical service entity, and the service associated with the detected medical service entity is marked as a detection case.

3. The method according to claim 1, characterized in that The density-based clustering algorithm is used to detect the claim behavior of medical entities with medical service characteristics that meet preset similarity standards, including: Grouping healthcare entities based on healthcare service characteristics; Applying a density-based clustering algorithm on each group to perform cluster analysis on claim behaviors, wherein the claim behaviors include the number of claims and the amount of claims; Identify abnormal claim behaviors in the clustering results and mark them as detection cases.

4. The method according to claim 1, characterized in that: The construction of a detection model based on the detection results includes: Based on the test results, the results are compared with the preset benchmark cases; Based on the comparison results, model parameters of the detection model are adjusted.

5. The method according to claim 1, characterized in that The preprocessing and feature extraction of the claim data to obtain the medical service entity as the root node and the service features as the child nodes include: Cleansing the claims data to remove invalid or incomplete records; Standardize the cleaned claims data to ensure consistency in data format; The medical service entity is extracted from the standardized claims data as the root node, and the related service features are extracted as child nodes.

6. A method for using a detection model, characterized in that: include: Obtain the claims data to be tested; The claim data to be detected is input into a detection model to obtain a detection result, and the detection model is obtained by any method as claimed in claim 1 to claim 5.

7. A system for constructing a detection model, characterized in that: include: An acquisition unit, used for acquiring claim data; An extraction unit, used for preprocessing and feature extraction of the claim data, to obtain a medical service entity as a root node and a service feature as a child node; A first construction unit is used to construct directed edges to connect the root node and service feature nodes according to the relationship defined in the claim data to obtain a graph data structure, wherein the graph data structure is used to represent the association relationship between different medical service entities and service features; A detection unit is used to detect whether there is a mismatch between a medical entity and its service characteristics using a weighted multi-tree and the graph data structure; and to detect the claim behavior of a medical entity having medical service characteristics that meet a preset similarity standard using a density-based clustering algorithm; The second construction unit is used to construct a detection model for identifying non-compliant behaviors in claims data based on the detection results.

8. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method according to any one of claims 1 to 5 is implemented.

9. A computer program product, characterized in that When the computer program product runs on a computer, the computer executes the method according to any one of claims 1 to 5.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed on a terminal device, the terminal device executes the method according to any one of claims 1 to 5.

Images