Abnormal order identification method and device of medicine diagnosis card, medium and equipment
By monitoring and extracting features from pharmacy card order information, and using an order detection model to identify abnormal orders in real time, the problem of lagging pharmacy card order identification in existing technologies is solved, ensuring the safety of insurance company assets.
Patent Information
- Application Number
- CN202511061885.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-30
- Publication Date
- 2025-11-14
AI Technical Summary
Existing methods for identifying pharmacy card orders cannot achieve real-time risk identification, are easily bypassed by malicious actors, and have delayed responses, which threatens the asset security and operational order of insurance companies.
By monitoring the order information of the medication card, target features and order-related features are extracted, and the trained order detection model is used for real-time analysis to identify abnormal orders.
It enables real-time risk identification of pharmacy card orders, timely interception of abnormal orders, and safeguards the asset security and operational order of insurance companies.
Smart Images
Figure CN120950871A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of medical insurance technology, and in particular to a method, device, medium and equipment for identifying abnormal orders for a pharmacy card. Background Technology
[0002] With the continuous innovation of service models in the insurance industry, pharmacy cards, as an important value-added service carrier for insurance products, have been widely used in scenarios such as medical consultations, drug purchases, and health consumption. However, as the scope of use of pharmacy cards expands, their potential economic value, due to their face value, has attracted the attention of criminals, leading to frequent cases of theft, cash-out, and resale, seriously threatening the asset security and operational order of insurance companies.
[0003] Currently, insurance companies typically use facial recognition verification and human customer service identity verification to manage the risk of using pharmacy cards. For example, before a user pays with a pharmacy card, the system requires facial recognition verification; in some high-risk areas, a human customer service review mechanism is introduced. In addition, some systems analyze logistics information to identify abnormal orders for pharmacy cards, such as triggering a post-event alert if a large number of identical or similar orders for the same product appear at the same address within a short period of time.
[0004] However, the above methods still have significant limitations. First, facial recognition verification is easily circumvented by malicious actors using techniques such as image synthesis and liveness detection. Second, methods based on logistics information analysis suffer from response lag due to the analysis of a large number of orders within a short period, often only detecting anomalies after the order has been shipped or even signed for, resulting in irreparable losses. Therefore, there is an urgent need to propose a method that can achieve real-time risk identification during the use of pharmacy cards. Summary of the Invention
[0005] In view of this, this application provides a method, apparatus, medium and equipment for identifying abnormal orders of a pharmacy card, which solves the problem that existing order identification methods for pharmacy cards cannot achieve real-time risk identification during the use of the pharmacy card.
[0006] According to one aspect of this application, a method for identifying abnormal orders on a pharmacy card is provided, the method comprising:
[0007] The order information of the target pharmacy card is monitored. When a payment order is generated on the target pharmacy card, the first order information of the payment order, the second order information of multiple related orders corresponding to the payment order, and the basic information of the target pharmacy card are obtained. The related orders are payment orders that occur on the target pharmacy card and other pharmacy cards within a preset time period.
[0008] Key features are extracted from the basic information of the first order information and the target medication card to obtain target features. The first order information and the second order information are analyzed to obtain order-related features.
[0009] The target features and the order-related features are input into the trained order detection model to obtain the order detection results.
[0010] Optionally, the order-related features include a first related feature, a second related feature, and a third related feature, and the step of analyzing the first order information and the second order information to obtain the order-related features includes:
[0011] The first order information and the second order information are analyzed to obtain the delivery information and the IP address of the order for each order;
[0012] Based on the delivery information, count the first number of related orders that have the same delivery address as the payment order, count the second number of related orders that have the same recipient information as the payment order, and count the third number of related orders that have the same order IP address as the payment order.
[0013] When the first quantity is greater than the first quantity threshold, feature extraction is performed on related orders with the same delivery address as the payment order to obtain the first related feature. When the second quantity is greater than the first quantity threshold, feature extraction is performed on related orders with the same recipient information as the payment order to obtain the second related feature. When the third quantity is greater than the first quantity threshold, feature extraction is performed on related orders with the same order placement IP address as the payment order to obtain the third related feature.
[0014] Optionally, the order-related features further include a fourth related feature, wherein the analysis of the first order information and the second order information to obtain the order-related features includes:
[0015] The first order information and the second order information are analyzed to obtain the drug information for each order;
[0016] Based on the drug information of each order, calculate the drug similarity between the payment order and each related order, and count the fourth number of related orders whose drug similarity is greater than the similarity threshold;
[0017] When the fourth quantity is greater than the second quantity threshold, feature extraction is performed on related orders whose drug similarity to the payment order is greater than the similarity threshold to obtain the fourth related feature.
[0018] Optionally, before monitoring the order information of the target medication card, the method for identifying abnormal orders of the medication card further includes:
[0019] Obtain the training dataset, and extract the order information and corresponding basic information of the medication card for each training order from the training dataset;
[0020] Each piece of information in the order information and the basic information of the corresponding medicine card of the training order is used as a feature variable, and order abnormality is used as the target variable. Based on each feature variable of each training order and the order detection result of each training order, an association list between the feature variable and the target variable is constructed. The cells of the association list store the actual frequency of the training order being an abnormal order and a normal order when the feature variable exists.
[0021] Based on the data in the association list, calculate the expected frequency of training orders as abnormal orders and normal orders when each feature variable exists. Based on the expected frequency and actual frequency of training orders as abnormal orders and normal orders when each feature variable exists, calculate the chi-square statistic for each feature variable.
[0022] The feature variables whose chi-square statistics are greater than a preset threshold are used as key features.
[0023] Optionally, the trained order detection model can be obtained using the following methods:
[0024] Obtain the training dataset, and extract the order information and corresponding basic information of the medication card for each training order from the training dataset;
[0025] Key features were extracted from the order information and corresponding basic information of the medication card for each training order to obtain training features;
[0026] The order information of each training order is analyzed to obtain the first relevant historical feature, the second relevant historical feature, the third relevant historical feature, and the fourth relevant historical feature;
[0027] The initial order detection model is trained based on the training features, the first relevant historical features, the second relevant historical features, the third relevant historical features, and the fourth relevant historical features to obtain the trained order detection model.
[0028] Optionally, the order detection results include normal orders, abnormal orders, and suspected orders. After obtaining the order detection results, the method for identifying abnormal orders on the medication card further includes:
[0029] Obtain the tracking information of the suspected order, and determine whether the suspected order is an abnormal order based on the tracking information;
[0030] When a suspected order is determined to be an abnormal order, the suspected order is added to the training dataset. Based on the supplemented training dataset, the order detection model is updated to obtain the updated model.
[0031] Optionally, inputting the target features and the order-related features into the trained order detection model includes:
[0032] The target features and the order-related features are input into the trained autoencoder to obtain compressed features;
[0033] The compressed features are then input into the trained order detection model.
[0034] According to another aspect of this application, an abnormal order identification device for a medication card is provided, the device comprising:
[0035] The order information acquisition module is used to monitor the order information of the target pharmacy card. When a payment order is detected to be generated on the target pharmacy card, the module acquires the first order information of the payment order, the second order information of multiple related orders corresponding to the payment order, and the basic information of the target pharmacy card. The related orders are payment orders that occurred on the target pharmacy card and other pharmacy cards within a preset time period.
[0036] The feature acquisition module is used to extract key features from the basic information of the first order information and the target drug diagnosis card respectively to obtain target features, and to analyze the first order information and the second order information to obtain order-related features;
[0037] The detection result acquisition module is used to input the target features and the order-related features into the trained order detection model to obtain the order detection results.
[0038] Optionally, the order-related features include a first related feature, a second related feature, and a third related feature, and the feature acquisition module is further configured to:
[0039] The first order information and the second order information are analyzed to obtain the delivery information and the IP address of the order for each order;
[0040] Based on the delivery information, count the first number of related orders that have the same delivery address as the payment order, count the second number of related orders that have the same recipient information as the payment order, and count the third number of related orders that have the same order IP address as the payment order.
[0041] When the first quantity is greater than the first quantity threshold, feature extraction is performed on related orders with the same delivery address as the payment order to obtain the first related feature. When the second quantity is greater than the first quantity threshold, feature extraction is performed on related orders with the same recipient information as the payment order to obtain the second related feature. When the third quantity is greater than the first quantity threshold, feature extraction is performed on related orders with the same order placement IP address as the payment order to obtain the third related feature.
[0042] Optionally, the order-related features further include a fourth related feature, and the feature acquisition module is further configured to:
[0043] The first order information and the second order information are analyzed to obtain the drug information for each order;
[0044] Based on the drug information of each order, calculate the drug similarity between the payment order and each related order, and count the fourth number of related orders whose drug similarity is greater than the similarity threshold;
[0045] When the fourth quantity is greater than the second quantity threshold, feature extraction is performed on related orders whose drug similarity to the payment order is greater than the similarity threshold to obtain the fourth related feature.
[0046] Optionally, the feature acquisition module is further configured to:
[0047] Obtain the training dataset, and extract the order information and corresponding basic information of the medication card for each training order from the training dataset;
[0048] Each piece of information in the order information and the basic information of the corresponding medicine card of the training order is used as a feature variable, and order abnormality is used as the target variable. Based on each feature variable of each training order and the order detection result of each training order, an association list between the feature variable and the target variable is constructed. The cells of the association list store the actual frequency of the training order being an abnormal order and a normal order when the feature variable exists.
[0049] Based on the data in the association list, calculate the expected frequency of training orders as abnormal orders and normal orders when each feature variable exists. Based on the expected frequency and actual frequency of training orders as abnormal orders and normal orders when each feature variable exists, calculate the chi-square statistic for each feature variable.
[0050] The feature variables whose chi-square statistics are greater than a preset threshold are used as key features.
[0051] Optionally, the abnormal order identification device for the medication card further includes:
[0052] The model training module is used to acquire the training dataset and obtain the order information and corresponding basic information of the medication card for each training order from the training dataset.
[0053] Key features were extracted from the order information and corresponding basic information of the medication card for each training order to obtain training features;
[0054] The order information of each training order is analyzed to obtain the first relevant historical feature, the second relevant historical feature, the third relevant historical feature, and the fourth relevant historical feature;
[0055] The initial order detection model is trained based on the training features, the first relevant historical features, the second relevant historical features, the third relevant historical features, and the fourth relevant historical features to obtain the trained order detection model.
[0056] Optionally, the order detection results include normal orders, abnormal orders, and suspected orders, and the abnormal order identification device for the medication card further includes:
[0057] The model update module is used to obtain the tracking information of the suspected order and determine whether the suspected order is an abnormal order based on the tracking information.
[0058] When a suspected order is determined to be an abnormal order, the suspected order is added to the training dataset. Based on the supplemented training dataset, the order detection model is updated to obtain the updated model.
[0059] Optionally, the detection result acquisition module is further configured to:
[0060] The target features and the order-related features are input into the trained autoencoder to obtain compressed features;
[0061] The compressed features are then input into the trained order detection model.
[0062] According to another aspect of this application, a storage medium is provided that stores a computer program thereon, which, when executed by a processor, implements the above-described method for identifying abnormal orders of a medical card.
[0063] According to another aspect of this application, a computer device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor executes the program to implement the above-mentioned abnormal order identification method for a medical card.
[0064] By utilizing the above technical solution, this application provides a method, apparatus, medium, and device for identifying abnormal orders on a medication card. When a payment order is detected on a target medication card, the method acquires the first order information of the payment order, the second order information of multiple related orders corresponding to the payment order, and the basic information of the target medication card. Target features are extracted from the first order information and the basic information of the target medication card. The first and second order information are analyzed to obtain order-related features. The target features and order-related features are input into a trained order detection model to obtain order detection results. At the moment the payment order occurs, the method acquires the target features involved in the order and the order association features similar to the payment order within a preset time period. The order detection model performs order result detection based on the target features and order association features, promptly determining the order detection result so that staff can intercept orders based on the detection results, ensuring the asset security and operational order of the insurance company.
[0065] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description
[0066] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:
[0067] Figure 1 A flowchart illustrating an abnormal order identification method for a medication card provided in an embodiment of this application is shown.
[0068] Figure 2 This paper illustrates another flowchart of a method for identifying abnormal orders for a medication card provided in an embodiment of this application.
[0069] Figure 3 This illustration shows another flowchart of an abnormal order identification method for a pharmacy card provided in an embodiment of this application;
[0070] Figure 4 This illustration shows a structural schematic diagram of an abnormal order identification device for a medicine card provided in an embodiment of this application;
[0071] Figure 5 A schematic diagram of the device structure of a computer device provided in an embodiment of this application is shown.
[0072] in,
[0073] Figure 4In Chinese: 402 - Order Information Acquisition Module; 404 - Feature Acquisition Module; 406 - Detection Result Acquisition Module. Detailed Implementation
[0074] The present application will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of the present application can be combined with each other.
[0075] This embodiment provides a method for identifying abnormal orders on a medication clinic card, such as... Figure 1 As shown, the method includes:
[0076] 102: Monitor the order information of the target drug card. When a payment order is generated on the target drug card, obtain the first order information of the payment order, the second order information of multiple related orders corresponding to the payment order, and the basic information of the target drug card. Among them, the related orders are the payment orders that occurred on the target drug card and other drug cards within a preset time period.
[0077] 104: Extract key features from the basic information of the first order and the target medication card to obtain target features; analyze the first order and the second order to obtain order-related features.
[0078] 106: Input the target features and order-related features into the trained order detection model to obtain the order detection results.
[0079] A pharmacy card is an electronic credential or physical card that integrates medical diagnosis, drug purchase, and personal health information management. It can be used in scenarios such as pharmaceutical retail and medical services. The pharmacy card can be used on multiple platforms, and a payment order will be generated when it is used.
[0080] Monitor the order information of the target pharmacy card. When a payment order is generated on the target pharmacy card, promptly check the payment order to determine if it is an abnormal order so that it can be intercepted. Remind the operations staff that the order is abnormal and handle it in time to avoid losses.
[0081] When an order is detected to be generated on the target drug card, the first order information of the payment order and the basic information of the target drug card are obtained. Key features are extracted from the first order information and the basic information of the target drug card to obtain the target features. The target features are significant features that have a significant impact on judging whether the order is abnormal.
[0082] Since some abnormal orders are multiple payment orders with common characteristics that occurred within a certain time period, multiple orders for the target drug card and other drug cards within the preset time period when the payment order occurred are considered as related orders. The first order information of the payment order and the second order information of multiple related orders are analyzed to extract their common features as order-related features.
[0083] The target features and order-related features are input into the trained order detection model. The order detection model performs order detection and outputs the detection results of payment orders. When the detection result of a payment order is a normal order, no action is taken. When the detection result of a payment order is an abnormal order, an alarm message is output to the operations personnel so that the operations personnel can intercept and handle it in a timely manner.
[0084] In a preferred embodiment of this example, inputting the target features and order-related features into the trained order detection model includes: inputting the target features and order-related features into the trained autoencoder to obtain compressed features, and inputting the compressed features into the trained order detection model.
[0085] Specifically, target features and order-related features are high-dimensional data. To improve the model's detection efficiency, dimensionality reduction is performed on these features. The trained autoencoder is used to "compress" these high-dimensional features—by learning the inherent patterns between features, the most crucial information for order detection is retained, while redundant or noisy information (such as repetitive formatted data and minor features with little impact) is removed. This makes the features more refined, avoids irrelevant information interfering with the detection results, and reduces the computational load of subsequent models, thus improving detection efficiency.
[0086] The service platforms used by the target drug diagnosis card in this application include, but are not limited to, medical insurance service platforms, public service platforms, and digital healthcare platforms. The abnormal order identification methods for the drug diagnosis card include, but are not limited to, medical insurance, public service, digital healthcare, and insurance business application scenarios.
[0087] This application provides a method for identifying abnormal orders on pharmacy cards. Compared with existing technologies, when a payment order is detected on a target pharmacy card, the method obtains the first order information of the payment order, the second order information of multiple related orders corresponding to the payment order, and the basic information of the target pharmacy card. Target features are extracted from the first order information and the basic information of the target pharmacy card. The first and second order information are analyzed to obtain order-related features. The target features and order-related features are input into a trained order detection model to obtain the order detection result. When the payment order just occurs, the method obtains the target features involved in the order and the order association features similar to the payment order within a preset time period. The order detection model performs order result detection based on the target features and order association features, and promptly determines the order detection result so that staff can intercept the order based on the detection result, ensuring the asset security and operational order of the insurance company.
[0088] In one embodiment, the order-related features include a first related feature, a second related feature, a third related feature, and a fourth related feature, such as... Figure 2 As shown, the first and second order information are analyzed to obtain order-related features, including:
[0089] 202: Analyze the information of the first and second orders to obtain the receipt information, order IP address, and drug information for each order;
[0090] 204: Based on the shipping information, count the first number of related orders with the same shipping address as the payment order, count the second number of related orders with the same recipient information as the payment order, and count the third number of related orders with the same order IP address as the payment order;
[0091] 206: When the first quantity is greater than the first quantity threshold, feature extraction is performed on related orders with the same delivery address as the payment order to obtain the first related feature; when the second quantity is greater than the first quantity threshold, feature extraction is performed on related orders with the same recipient information as the payment order to obtain the second related feature.
[0092] 208: When the third quantity is greater than the first quantity threshold, feature extraction is performed on related orders with the same order IP address as the payment order to obtain the third related feature;
[0093] 210: Based on the drug information of each order, calculate the drug similarity between the payment order and each related order, and count the fourth number of related orders whose drug similarity is greater than the similarity threshold;
[0094] 212: When the fourth quantity is greater than the second quantity threshold, feature extraction is performed on related orders whose drug similarity to the payment order is greater than the similarity threshold to obtain the fourth related feature.
[0095] Specifically, some abnormal orders share commonalities. For example, some orders have the same delivery address, some have the same recipient information, some have the same ordering IP address, and some contain similar medications. Therefore, it is necessary to obtain related orders that occurred on the target pharmacy card and other pharmacy cards within a preset time period when the payment order was placed. The delivery address, recipient information, ordering IP address, and medication similarity of the payment order and related orders are analyzed. When the number of related orders with the same delivery address as the payment order exceeds a first threshold, it is determined that the payment order and these related orders with the same delivery address share commonalities and require further analysis. Therefore, the features of the payment order and these related orders with the same delivery address are extracted as the first relevant features.
[0096] When the number of related orders with the same recipient information as the payment order exceeds the first quantity threshold, it is determined that the payment order and these related orders with the same recipient information have commonalities and need to be analyzed in detail. Therefore, the features of the payment order and these related orders with the same recipient information are extracted as the second related features.
[0097] When the number of related orders with the same IP address as the payment order exceeds a first threshold, it is determined that the payment order and these related orders with the same IP address share commonalities and require further analysis. Therefore, the characteristics of the payment order and these related orders with the same IP address are extracted as the third relevant feature. Alternatively, commonalities can be determined based on information such as the registered account or registered mobile phone number. For example, when the number of related orders with the same registered account as the payment order exceeds a first threshold, it is determined that the payment order and these related orders with the same registered account share commonalities, and the third relevant feature is extracted; similarly, when the number of related orders with the same registered mobile phone number as the payment order exceeds a first threshold, it is determined that the payment order and these related orders with the same registered mobile phone number share commonalities, and the third relevant feature is extracted.
[0098] When the number of related orders with high similarity to the drugs in the payment order exceeds the second quantity threshold, it is determined that the payment order and these related orders with high similarity to the drugs have commonalities and need to be analyzed in detail. Therefore, the features of the payment order and the related orders with the same ordering IP address are extracted as the fourth related feature.
[0099] In one embodiment, such as Figure 3 As shown, before monitoring the order information of the target medication card, the method for identifying abnormal orders of the medication card also includes:
[0100] 302: Obtain the training dataset. Extract the order information and corresponding basic information of the medication card for each training order from the training dataset.
[0101] 304: Take each piece of information in the order information of the training order and the basic information of the corresponding medication card as a feature variable, and take order anomalies as the target variable;
[0102] 306: Based on each feature variable of each training order and the order detection result of each training order, construct an association list between the feature variable and the target variable. The cells of the association list store the actual frequency of training orders being abnormal and normal orders when the feature variable exists.
[0103] 308: Based on the data in the association list, calculate the expected frequency of training orders as abnormal orders and normal orders when each feature variable exists. Based on the expected frequency and actual frequency of training orders as abnormal orders and normal orders when each feature variable exists, calculate the chi-square statistic for each feature variable.
[0104] 310: Feature variables with chi-square statistics greater than a preset threshold are used as key features.
[0105] In this embodiment, since the order information and the basic information of the medication card contain a lot of information, such as the order information including: the mobile phone model, device ID, order time, order IP address, registered account, registered mobile phone number, ordered product information, product type, recipient name, recipient mobile phone number, and recipient address, and the basic information of the medication card including province, city, district, medication card type, policy number, and insurance institution, it is necessary to extract the information that has the greatest impact on the judgment of order anomalies from the order information and the basic information of the medication card as key features.
[0106] Obtain the training dataset, which contains a large number of past order cases (which may include normal orders and abnormal orders). Treat each specific field in the order information and the basic information of the medication card (such as "order amount", "user age", "quantity of medicine purchased") as an independent feature variable. Treat the detection result of the order as the target variable, such as the target variable including normal order and abnormal order.
[0107] Association lists, also known as contingency tables, are used to count the actual frequency of associations between each feature variable and the target variable.
[0108] Assuming the feature variable is "whether the quantity of medicine purchased exceeds the regular dosage" (yes / no) and the target variable is "whether the order is abnormal", the contingency table might look like this:
[0109]
[0110] This table shows:
[0111] Of the orders that "the quantity of medicine purchased exceeded the regular dosage," 120 were actually abnormal orders, and 80 were normal orders.
[0112] Among the orders where "the quantity of medicine purchased did not exceed the regular dosage", there were 50 abnormal orders and 250 normal orders.
[0113] The expected frequency of each cell is calculated using the following formula: Expected Frequency = (Row Total × Column Total) / Total Sample Size
[0114] For example: when "the quantity of medicine purchased exceeds the usual dosage",
[0115] The expected frequency of abnormal orders = (200 × 170) / 500 = 68;
[0116] The expected frequency of normal orders = (200 × 330) / 500 = 132;
[0117] The chi-square statistic is used to measure the degree of deviation between the actual frequency and the expected frequency. The greater the deviation, the more likely there is a correlation between the feature and the target variable.
[0118] Formula for calculating the degree of deviation: χ² 2 =Σ[(actual frequency - expected frequency)2 / expected frequency]
[0119] Chi-square statistic = [(120-68)] 2 / 68+(80-132) 2 / 132+(50-102) 2 / 102+(250-198) 2 / 198]≈84.7
[0120] A larger chi-square value indicates a stronger correlation between the feature variable and the target variable. If the chi-square statistic of a feature variable is greater than a preset threshold, the feature variable is considered to be significantly correlated with order anomalies and is thus designated as a key feature. The preset threshold is set based on business needs or the level of statistical significance.
[0121] In one embodiment, the trained order detection model is obtained using the following method:
[0122] Obtain the training dataset, and extract the order information and corresponding basic information of the medication card for each training order from the training dataset;
[0123] Key features were extracted from the order information and corresponding basic information of the medication card for each training order to obtain training features;
[0124] The order information of each training order is analyzed to obtain the first relevant historical feature, the second relevant historical feature, the third relevant historical feature, and the fourth relevant historical feature;
[0125] The initial order detection model is trained based on the training features, the first relevant historical features, the second relevant historical features, the third relevant historical features, and the fourth relevant historical features to obtain the trained order detection model.
[0126] Specifically, the training dataset contains a large number of past order cases (which may include normal and abnormal orders, used to teach the model the difference between "normal" and "abnormal").
[0127] Order information includes the raw data for each order, such as: order number, device ID of the device that placed the order, IP address of the device that placed the order, order time, type / quantity / price of medicines purchased, delivery address, recipient information, etc.
[0128] The basic information of the medication card includes: user identity information (age, gender), past medical records (such as diagnosed symptoms), historical medication purchase records (such as commonly used drug types), medical insurance status, insurance company, medication card level, etc.
[0129] Obtain the order information and corresponding basic information of the medication card for each training order in the training dataset. Extract key features from the order information and corresponding basic information of the medication card for each training order to obtain training features.
[0130] Some abnormal orders have multiple orders with the same shipping address, some have multiple orders with the same recipient information, and some have multiple orders with the same ordering IP address. Therefore, it is necessary to obtain the shipping address, recipient information, and ordering IP address of these abnormal orders, determine whether there are many orders with the same shipping address, the same recipient information, or the same ordering IP address, and extract the features of these abnormal orders. The model then analyzes these abnormal orders and traces the correlation between them.
[0131] Each training order is analyzed to obtain the delivery information, order IP address, and drug information. For training orders with the same delivery address, features are extracted to obtain the first relevant historical features. For training orders with the same recipient information, features are extracted to obtain the second relevant historical features. For training orders with the same order IP address, features are extracted to obtain the third relevant historical features. For training orders with drug similarity greater than the similarity threshold, features are extracted to obtain the fourth relevant historical features.
[0132] The target features, first relevant historical features, second relevant historical features, third relevant historical features, fourth relevant historical features, and detection results of each training order are input into the initial order detection model. The initial order detection model is then trained to obtain the trained order detection model.
[0133] In one embodiment, the order detection results include normal orders, abnormal orders, and suspected orders. After obtaining the order detection results, the method for identifying abnormal orders on the medication card further includes:
[0134] Obtain tracking information for suspected orders and determine whether the suspected orders are abnormal based on the tracking information;
[0135] When a suspected order is identified as an abnormal order, the suspected order is added to the training dataset. Based on the supplemented training dataset, the order detection model is updated to obtain the updated model.
[0136] Specifically, order detection results include normal orders, abnormal orders, and suspected orders. Order detection is typically completed within a preset timeframe after payment. If operations personnel do not receive abnormal order detection results within the preset timeframe, they will proceed with subsequent processing as normal. If an abnormal order is identified, the abnormal order information is output to the operations personnel's terminal, for example, via email or SMS. The operations personnel mark and intercept the abnormal order, for example, by preventing it from leaving the warehouse, or by intercepting logistics for orders that have already left the warehouse. If a suspected order is identified, the suspected order information is output to the operations personnel. After receiving the suspected order information, the operations personnel will track the suspected order and obtain tracking information.
[0137] After obtaining the final tracking information input by the operations staff, when the suspected order is found to be an abnormal order, the suspected order is added to the training dataset. The order detection model is then updated based on the supplemented training dataset to obtain the updated order detection model. Subsequent payment orders are then detected based on the updated order detection model.
[0138] In one embodiment, big data and artificial intelligence technologies are combined to continuously monitor and optimize the order anomaly detection process, dynamically adjust strategies, and form a closed-loop feedback mechanism.
[0139] Furthermore, as Figure 1 In terms of specific implementation, this application provides an abnormal order identification device for a medication card, such as... Figure 4 As shown, the device includes:
[0140] The order information acquisition module 402 is used to monitor the order information of the target drug card. When a payment order is detected to be generated on the target drug card, the module acquires the first order information of the payment order, the second order information of multiple related orders corresponding to the payment order, and the basic information of the target drug card. The related orders are the payment orders that occurred on the target drug card and other drug cards within a preset time period.
[0141] The feature acquisition module 404 is used to extract key features from the basic information of the first order information and the target drug diagnosis card respectively to obtain target features, and to analyze the first order information and the second order information to obtain order-related features;
[0142] The detection result acquisition module 406 is used to input the target features and order-related features into the trained order detection model to obtain the order detection results.
[0143] This application provides an abnormal order identification device for pharmacy cards. Compared with the prior art, when a payment order is detected on a target pharmacy card, the device acquires the first order information of the payment order, the second order information of multiple related orders corresponding to the payment order, and the basic information of the target pharmacy card. Target features are extracted from the first order information and the basic information of the target pharmacy card. The first and second order information are analyzed to obtain order-related features. The target features and order-related features are input into a trained order detection model to obtain the order detection result. When a payment order is first generated, the device acquires the target features involved in the order and the order association features similar to the payment order within a preset time period. The order detection model performs order result detection based on the target features and order association features, promptly determining the order detection result so that staff can intercept orders based on the detection results, ensuring the asset security and operational order of the insurance company.
[0144] In one embodiment, the order-related features include a first related feature, a second related feature, and a third related feature, and the feature acquisition module is further configured to:
[0145] Analyze the information of the first and second orders to obtain the delivery information and the IP address where the order was placed for each order;
[0146] Based on the delivery information, count the first number of related orders with the same delivery address as the payment order, count the second number of related orders with the same recipient information as the payment order, and count the third number of related orders with the same order IP address as the payment order.
[0147] When the first quantity is greater than the first quantity threshold, feature extraction is performed on related orders with the same delivery address as the payment order to obtain the first related feature. When the second quantity is greater than the first quantity threshold, feature extraction is performed on related orders with the same recipient information as the payment order to obtain the second related feature. When the third quantity is greater than the first quantity threshold, feature extraction is performed on related orders with the same order placement IP address as the payment order to obtain the third related feature.
[0148] In one embodiment, the order-related features further include a fourth related feature, and the feature acquisition module is further configured to:
[0149] Analyze the information from the first and second orders to obtain the drug information for each order;
[0150] Based on the drug information of each order, calculate the drug similarity between the payment order and each related order, and count the fourth number of related orders whose drug similarity is greater than the similarity threshold;
[0151] When the fourth quantity is greater than the second quantity threshold, feature extraction is performed on related orders whose drug similarity to the payment order is greater than the similarity threshold to obtain the fourth related feature.
[0152] In one embodiment, the feature acquisition module is further configured to:
[0153] Obtain the training dataset, and extract the order information and corresponding basic information of the medication card for each training order from the training dataset;
[0154] Each piece of information in the order information and the corresponding basic information of the medical card of the training order is used as a feature variable, and the order abnormality is used as the target variable. Based on each feature variable of each training order and the order detection result of each training order, an association list between the feature variable and the target variable is constructed. The cells of the association list store the actual frequency of the training order being an abnormal order and a normal order when the feature variable exists.
[0155] Based on the data in the association list, calculate the expected frequency of training orders as abnormal orders and normal orders when each feature variable exists. Based on the expected frequency and actual frequency of training orders as abnormal orders and normal orders when each feature variable exists, calculate the chi-square statistic for each feature variable.
[0156] Feature variables with chi-square statistics greater than a preset threshold are used as key features.
[0157] In one embodiment, the abnormal order identification device for the medication card further includes:
[0158] The model training module is used to obtain the training dataset and extract the order information and corresponding basic information of the medication card for each training order from the training dataset.
[0159] Key features were extracted from the order information and corresponding basic information of the medication card for each training order to obtain training features;
[0160] The order information of each training order is analyzed to obtain the first relevant historical feature, the second relevant historical feature, the third relevant historical feature, and the fourth relevant historical feature;
[0161] The initial order detection model is trained based on the training features, the first relevant historical features, the second relevant historical features, the third relevant historical features, and the fourth relevant historical features to obtain the trained order detection model.
[0162] In one embodiment, the order detection results include normal orders, abnormal orders, and suspected orders. The abnormal order identification device for the medication card further includes:
[0163] The model update module is used to obtain tracking information of suspected orders and determine whether the suspected orders are abnormal orders based on the tracking information.
[0164] When a suspected order is identified as an abnormal order, the suspected order is added to the training dataset. Based on the supplemented training dataset, the order detection model is updated to obtain the updated model.
[0165] In one embodiment, the detection result acquisition module is further configured to:
[0166] The target features and order-related features are input into the trained autoencoder to obtain compressed features;
[0167] The compressed features are then input into the trained order detection model.
[0168] It should be noted that other corresponding descriptions of the functional units involved in the abnormal order identification device for a medication card provided in this application embodiment can be found in the following references. Figures 1 to 3 The corresponding descriptions in the method will not be repeated here.
[0169] This application also provides a computer device, which may specifically be a personal computer, a server, a network device, etc. Figure 5 As shown, the computer device includes a bus, a processor, memory, and a communication interface, and may also include an input / output interface and a display device. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores location information. The network interface allows communication with external terminals via a network connection. When the computer program is executed by the processor, it implements the steps in the various method embodiments.
[0170] Those skilled in the art will understand that Figure 5The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0171] In one embodiment, a computer-readable storage medium is provided, which may be non-volatile or volatile, having stored thereon a computer program that, when executed by a processor, implements the steps in the above method embodiments.
[0172] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.
[0173] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.
[0174] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.
[0175] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0176] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A method for identifying abnormal orders on a medication prescription card, characterized in that, The method includes: The order information of the target pharmacy card is monitored. When a payment order is generated on the target pharmacy card, the first order information of the payment order, the second order information of multiple related orders corresponding to the payment order, and the basic information of the target pharmacy card are obtained. The related orders are payment orders that occur on the target pharmacy card and other pharmacy cards within a preset time period. Key features are extracted from the basic information of the first order information and the target medication card to obtain target features. The first order information and the second order information are analyzed to obtain order-related features. The target features and the order-related features are input into the trained order detection model to obtain the order detection results.
2. The method for identifying abnormal orders on a medication card according to claim 1, characterized in that, The order-related features include a first related feature, a second related feature, and a third related feature. The step of analyzing the first order information and the second order information to obtain the order-related features includes: The first order information and the second order information are analyzed to obtain the delivery information and the IP address of the order for each order; Based on the delivery information, count the first number of related orders that have the same delivery address as the payment order, count the second number of related orders that have the same recipient information as the payment order, and count the third number of related orders that have the same order IP address as the payment order. When the first quantity is greater than the first quantity threshold, feature extraction is performed on related orders with the same delivery address as the payment order to obtain the first related feature. When the second quantity is greater than the first quantity threshold, feature extraction is performed on related orders with the same recipient information as the payment order to obtain the second related feature. When the third quantity is greater than the first quantity threshold, feature extraction is performed on related orders with the same order placement IP address as the payment order to obtain the third related feature.
3. The method for identifying abnormal orders on a medication card according to claim 2, characterized in that, The order-related features also include a fourth related feature. The analysis of the first order information and the second order information to obtain the order-related features includes: The first order information and the second order information are analyzed to obtain the drug information for each order; Based on the drug information of each order, calculate the drug similarity between the payment order and each related order, and count the fourth number of related orders whose drug similarity is greater than the similarity threshold; When the fourth quantity is greater than the second quantity threshold, feature extraction is performed on related orders whose drug similarity to the payment order is greater than the similarity threshold to obtain the fourth related feature.
4. The method for identifying abnormal orders on a medication card according to claim 1, characterized in that, Before monitoring the order information of the target medication card, the method for identifying abnormal orders of the medication card further includes: Obtain the training dataset, and extract the order information and corresponding basic information of the medication card for each training order from the training dataset; Each piece of information in the order information and the basic information of the corresponding medicine card of the training order is used as a feature variable, and order abnormality is used as the target variable. Based on each feature variable of each training order and the order detection result of each training order, an association list between the feature variable and the target variable is constructed. The cells of the association list store the actual frequency of the training order being an abnormal order and a normal order when the feature variable exists. Based on the data in the association list, calculate the expected frequency of training orders as abnormal orders and normal orders when each feature variable exists. Based on the expected frequency and actual frequency of training orders as abnormal orders and normal orders when each feature variable exists, calculate the chi-square statistic for each feature variable. The feature variables whose chi-square statistics are greater than a preset threshold are used as key features.
5. The method for identifying abnormal orders on a medication clinic card according to claim 4, characterized in that, The trained order detection model is obtained using the following method: Obtain the training dataset, and extract the order information and corresponding basic information of the medication card for each training order from the training dataset; Key features were extracted from the order information and corresponding basic information of the medication card for each training order to obtain training features; The order information of each training order is analyzed to obtain the first relevant historical feature, the second relevant historical feature, the third relevant historical feature, and the fourth relevant historical feature; The initial order detection model is trained based on the training features, the first relevant historical features, the second relevant historical features, the third relevant historical features, and the fourth relevant historical features to obtain the trained order detection model.
6. The method for identifying abnormal orders on a medication clinic card according to claim 5, characterized in that, The order detection results include normal orders, abnormal orders, and suspected orders. After obtaining the order detection results, the method for identifying abnormal orders on the medication card further includes: Obtain the tracking information of the suspected order, and determine whether the suspected order is an abnormal order based on the tracking information; When a suspected order is determined to be an abnormal order, the suspected order is added to the training dataset. Based on the supplemented training dataset, the order detection model is updated to obtain the updated model.
7. The method for identifying abnormal orders of a medication card according to any one of claims 1-6, characterized in that, The step of inputting the target features and the order-related features into the trained order detection model includes: The target features and the order-related features are input into the trained autoencoder to obtain compressed features; The compressed features are then input into the trained order detection model.
8. An abnormal order identification device for a medication clinic card, characterized in that, The device includes: The order information acquisition module is used to monitor the order information of the target pharmacy card. When a payment order is detected to be generated on the target pharmacy card, the module acquires the first order information of the payment order, the second order information of multiple related orders corresponding to the payment order, and the basic information of the target pharmacy card. The related orders are payment orders that occurred on the target pharmacy card and other pharmacy cards within a preset time period. The feature acquisition module is used to extract key features from the basic information of the first order information and the target drug diagnosis card respectively to obtain target features, and to analyze the first order information and the second order information to obtain order-related features; The detection result acquisition module is used to input the target features and the order-related features into the trained order detection model to obtain the order detection results.
9. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.
10. A computer device, comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.