A model training method, business processing method, device and equipment
By using preset judgment rules to filter and expand historical business records when training samples are insufficient, training sample business is constructed, and an abnormal business identification model is trained. This solves the problem of model accuracy caused by insufficient sample size, and enables effective identification of abnormal business and normal operation of business processes.
Patent Information
- Application Number
- CN202011011568.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-09-23
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2040-09-23
AI Technical Summary
When the amount of training sample data is insufficient, machine learning models struggle to accurately identify abnormal business operations, especially when there are few abnormal situations. Existing technologies cannot effectively train models with high accuracy.
By acquiring historical business records, abnormal business is filtered out using preset judgment rules, the number of training samples is expanded, training sample business is constructed, and these business information are used to train an abnormal business identification model, including Bayesian classification model, support vector machine classification model, or convolutional neural network model, etc.
When the sample size is insufficient, the accuracy of the model can be ensured by expanding the training sample size, thereby effectively identifying abnormal business operations and ensuring the normal operation and effectiveness of business processes.
Smart Images

Figure CN112102049B_ABST
Abstract
Description
Technical Field
[0001] The embodiments of this specification relate to the field of computer technology, and in particular to a model training method, a business processing method, a device, and a device. Background Art
[0002] With the development of artificial intelligence technology, people can use computers to complete some tasks that require human intelligence, such as classifying target objects, or performing image recognition and processing. Machine learning technology is the core technology of artificial intelligence, which mainly improves the cognition of a certain thing by continuously acquiring new data. For example, after determining the model for identifying a certain object, it continuously obtains sample data corresponding to the object and uses the sample data for training to improve the model, thereby making the model more accurately identify the object.
[0003] In order to ensure the accuracy of the machine learning model, a large amount of training sample data is often required when training the machine learning model. However, when the amount of sample data used to train the model is too small, it may not be possible to train a model with high accuracy. For example, when the training model identifies abnormal bids, it is necessary to screen out abnormal bidding records from previous bidding records, and use the marked data to train the corresponding model. However, in real life, abnormal bidding situations such as collusion between bidding companies and bidding companies knowing the bid price information in advance are not common. The acquired bidding records may only contain a small number of abnormal bidding records. In the case of a small number of abnormal bidding records, the model trained based on abnormal bidding records may only be able to identify some special cases of abnormal bidding, which makes the model lack a certain degree of accuracy. Therefore, there is an urgent need for a method that can still achieve normal training of machine learning models when there are few training samples. Summary of the invention
[0004] The purpose of the embodiments of this specification is to provide a model training method, a business processing method, an apparatus and a device to solve the problem of how to conveniently and accurately use the trained model to identify abnormal business.
[0005] In order to solve the above technical problems, the embodiment of this specification provides a model training method, including:
[0006] Acquire at least one historically recorded service; the historically recorded service corresponds to service information;
[0007] Determine, according to the service information, a judgment result of the historically recorded service corresponding to at least one preset judgment rule; the judgment result includes normal service and abnormal service;
[0008] Using the business information of the historically recorded business including the abnormal business in the judgment result to construct at least one training sample business; the number of the training sample business is greater than the number of the historically recorded business;
[0009] An abnormal business identification model is trained based on the training sample business; the abnormal business identification model is used to identify abnormal business.
[0010] The embodiment of this specification also proposes a model training device, including:
[0011] A historical record service acquisition module, used to acquire at least one historical record service; the historical record service corresponds to service information;
[0012] A judgment result determination module, used to determine the judgment result of the historically recorded business corresponding to at least one preset judgment rule according to the business information; the judgment result includes normal business and abnormal business;
[0013] A training sample service construction module, configured to construct at least one training sample service by using service information of historically recorded services including abnormal services in the judgment result; the number of the training sample services is greater than the number of historically recorded services;
[0014] The abnormal business identification model training module is used to train the abnormal business identification model based on the training sample business; the abnormal business identification model is used to identify abnormal business.
[0015] The embodiments of the present specification also propose a model training device, including a memory and a processor; the memory is used to store computer program instructions; the processor is used to execute the computer program instructions to implement the following steps: obtaining at least one historically recorded business; the historically recorded business corresponds to business information; determining, based on the business information, a judgment result of the historically recorded business corresponding to at least one preset judgment rule; the judgment result includes normal business and abnormal business; using the business information of the historically recorded business including the abnormal business in the judgment result to construct at least one training sample business; the number of the training sample businesses is greater than the number of the historically recorded businesses; training an abnormal business identification model based on the training sample business; the abnormal business identification model is used to identify abnormal business.
[0016] In order to solve the above technical problems, the embodiment of this specification also proposes a service processing method, including:
[0017] Receiving pending business; the pending business corresponds to business information;
[0018] Based on the business information of the business to be processed, identifying whether the business to be processed is an abnormal business; wherein, including using an abnormal business identification model to identify whether the business to be processed is an abnormal business; the abnormal business identification model includes, after identifying the abnormal business in the historical record business according to a preset judgment rule, using the business information corresponding to the abnormal business to construct a training sample business, and based on the model trained by the training sample business;
[0019] In the case that the service to be processed is an abnormal service, service rejection information is fed back; the service rejection information is used to indicate that the service to be processed is rejected.
[0020] The embodiment of this specification also provides a service processing device, including:
[0021] A pending business receiving module is used to receive pending business; the pending business corresponds to business information;
[0022] An abnormal business identification module is used to identify whether the business to be processed is an abnormal business based on the business information of the business to be processed; wherein, it includes using an abnormal business identification model to identify whether the business to be processed is an abnormal business; the abnormal business identification model includes, after identifying the abnormal business in the historical record business according to a preset judgment rule, using the business information corresponding to the abnormal business to construct a training sample business, and based on the model trained by the training sample business;
[0023] The service rejection information feedback module is used to feedback service rejection information when the service to be processed is an abnormal service; the service rejection information is used to indicate that the service to be processed is rejected.
[0024] The embodiment of this specification also proposes a business processing device, including a memory and a processor; the memory is used to store computer program instructions; the processor is used to execute the computer program instructions to implement the following steps: receiving a business to be processed; the business to be processed corresponds to business information; identifying whether the business to be processed is an abnormal business based on the business information of the business to be processed; wherein, including using an abnormal business identification model to identify whether the business to be processed is an abnormal business; the abnormal business identification model includes, after identifying the abnormal business in the historical recorded business according to a preset judgment rule, using the business information corresponding to the abnormal business to construct a training sample business, and a model obtained by training the training sample business; in the case that the business to be processed is an abnormal business, feedback business rejection information; the business rejection information is used to indicate a refusal to process the business to be processed.
[0025] It can be seen from the technical solutions provided by the above embodiments of this specification that after obtaining the historical record samples, the model training method described in the embodiments of this specification screens out abnormal business in the historical record samples according to the preset judgment conditions, and expands the business information corresponding to the abnormal business, thereby obtaining a training sample business with a number greater than the abnormal business, so that the business identification model can be trained using the training sample business. When the sample size used for training the model is insufficient, the above method expands the training sample size according to the information of the pre-screened business, thereby ensuring the accuracy of the trained model, which is conducive to using the model to complete the identification of abnormal business in subsequent steps and ensuring the normal progress of the business process. After obtaining the business to be processed, the business processing method determines whether the business to be processed is an abnormal business through the abnormal business identification model. In the case where the abnormal business identification model is a model obtained by training with a sufficient number of samples, the accuracy of identifying abnormal business using the abnormal business identification model is ensured, so that in the subsequent process, the occurrence of abnormal business processing or related process damage caused by processing abnormal business can be avoided, thereby improving the effectiveness of business processing. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] In order to more clearly illustrate the embodiments of this specification 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 recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0027] Figure 1 This is a flow chart of a model training method according to an embodiment of this specification;
[0028] Figure 2 This is a flow chart of a business processing method according to an embodiment of this specification;
[0029] Figure 3 This is a module diagram of a model training device according to an embodiment of this specification;
[0030] Figure 4 This is a module diagram of a service processing device according to an embodiment of this specification;
[0031] Figure 5 This is a structural diagram of a model training device according to an embodiment of this specification;
[0032] Figure 6 This is a structural diagram of a business processing device according to an embodiment of this specification. DETAILED DESCRIPTION
[0033] The following will be combined with the drawings in the embodiments of this specification to clearly and completely describe the technical solutions in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.
[0034] The following is combined with Figure 1 A model training method according to an embodiment of this specification is introduced. The execution subject of the method is a model training device, which includes but is not limited to a server, an industrial computer, a PC, etc. The model training method includes the following specific implementation steps.
[0035] S110: Acquire at least one historically recorded service; the historically recorded service corresponds to service information.
[0036] The historically recorded business may be a business recorded in a database or other storage system and corresponding to the model trained in the embodiment of this specification. The historically recorded business corresponds to corresponding business information, and the historically recorded business can be analyzed and utilized.
[0037] Since the embodiment of this specification is a model for training and identifying abnormal business, the historical record business may include normal business and abnormal business. The abnormal business may be a business that interferes with the normal business process or does not comply with business-related regulations. The specific criteria for distinguishing between normal business and abnormal business can be set according to the needs of actual application, and will not be repeated here.
[0038] Business information is information corresponding to the historical record business. Specifically, the business information may be information corresponding to the historical record business itself, or may be identity information of the user who submitted the historical record business. When the historical record business corresponds to an institution, organization or company, the business information may also include registration information of the corresponding institution, organization or company and relevant identity information of the employed employees. In actual applications, the corresponding business information may be obtained according to the specific circumstances of the historical record business, which is not limited to the above examples and will not be elaborated here.
[0039] In some embodiments, the model trained by the embodiments of this specification can be a model for identifying abnormal bids. When a seller needs to sell an item, a bidding method can be used, that is, after multiple buyers make bids for the above item, the seller decides the object to which the item is sold based on the buyers' bids, for example, it can be the buyer with the highest bid. However, in the bidding process, abnormal bidding may occur. These abnormal bidding phenomena may be collusion between buyers to lower the bid as much as possible, or buyers with bad credit maliciously raise the bid price, thereby affecting the normal progress of the bidding.
[0040] Accordingly, when the model trained in the embodiment of this specification is a model for identifying abnormal bids, the historical record business may be a previous bidding record. The business information corresponding to these bidding records may be the bidding amount, bidding company information, bidding company employee information, etc.
[0041] S120: Determine, according to the service information, a judgment result of the historically recorded service corresponding to at least one preset judgment rule; the judgment result includes normal service and abnormal service.
[0042] After obtaining the historically recorded business, the historically recorded business can be analyzed and judged to obtain the judgment result corresponding to the historically recorded business. The judgment result is used to determine the category corresponding to the historically recorded business, so as to select the business that can be used to implement model training in subsequent steps from the historically recorded business.
[0043] In some implementations, the judgment result includes normal services and abnormal services. In the case where the model targeted by the embodiments of this specification is used to identify abnormal services, the corresponding model can be trained based on the historical records of services containing abnormal services in the judgment result.
[0044] Specifically determining the judgment result corresponding to the historically recorded business may be to discriminate the historically recorded business using a preset judgment rule. The preset judgment rule may be a pre-summarized judgment rule corresponding to various characteristics of abnormal business. Since the preset judgment rules are often used to judge the details of the historically recorded business and often have a large number, in actual applications, directly using the preset judgment rules to identify abnormal business has high requirements for the completeness of business information and requires more time and resources. Therefore, the preset judgment rules are generally only used to screen training samples before training the model.
[0045] In some embodiments, the preset judgment rule may be a correlation judgment rule. The correlation judgment rule is used to judge whether there is correlation between historical record services for the same object. Judging whether there is correlation between historical record services may be after obtaining the business information of the historical record services, judging whether there is correlation between historical record services corresponding to the same target object according to the business information. If it is judged that there is correlation between the business information corresponding to different historical record services, the historical record services with correlation between these business information may be identified as abnormal businesses.
[0046] Using a specific example to illustrate, when the historically recorded business is a bidding business, the target object may be a bidding item. After obtaining the historically recorded business, the bidding items corresponding to each historically recorded business may be determined by analyzing the business information of each historically recorded business, and the historically recorded business may be classified into businesses corresponding to different bidding items. For historically recorded businesses corresponding to the same bidding item, whether different historically recorded businesses are associated may be determined by analyzing whether the business information between the historically recorded businesses is associated.
[0047] For example, when the historical record business is a bidding business, it can be determined whether the corresponding bidding companies employ the same employees, or the actual controllers of the bidding companies are the same company, or the bank accounts used by the bidding companies are the same; it can also be determined whether there is a relationship between the employees of the bidding companies, for example, by obtaining the identity information of employees of different bidding companies, and analyzing whether there are former colleagues or relatives between the employees of different bidding companies, or whether there are frequent or large transfers between employees of different bidding companies near the bidding date, and then determining whether there is an overly close relationship between the employees of different bidding companies that affects the normal progress of the bidding.
[0048] When it is determined through analysis that there are correlations between different bidding businesses, these bidding businesses may be businesses proposed by corresponding users after collusion. These bidding businesses can be determined as abnormal businesses and thus used as training samples to perform subsequent steps.
[0049] In some implementations, the preset judgment rule may also be a bid amount similarity rule. When the historically recorded business is a business corresponding to a bidding project, the historically recorded business is the corresponding bidding business, and the bidding business corresponds to a certain bidding amount. By comparing the bidding amounts for the same bidding project, if the difference between the bidding amounts of the bidding businesses is greater than the maximum difference in bidding amounts or less than the minimum difference in bidding amounts, the corresponding bidding business may be a business proposed by the corresponding user in advance communication, and the corresponding bidding business may also be identified as an abnormal business.
[0050] To illustrate with a specific example, assuming that for a certain bidding project, the maximum difference in the corresponding bid amounts is 500,000, and the minimum difference in the bid amounts is 10,000. After obtaining the historically recorded business corresponding to the bidding project, the corresponding bid amounts are obtained based on the business information corresponding to each historically recorded business, and the differences between the bid amounts are analyzed. If the difference between any two bid amounts is greater than 500,000 or less than 10,000, then the two bid amounts may be the bid amounts determined by the initiating users of the business after collusion, and the historically recorded business corresponding to these two bid amounts can be determined as abnormal business.
[0051] In some implementations, the preset judgment rule may also be a bid price leakage rule. The historically recorded business may correspond to a bidding project, and the bidding project may correspond to a budget amount. The budget amount may be the bid price of the bidding project, that is, the lowest bid for the bidding project. After the historically recorded business is acquired, the corresponding bid amount may be extracted from the business information of the historically recorded business, and the difference between the bid amount and the budget amount may be determined, and when the difference is less than the bid difference threshold, the historically recorded business may be identified as an abnormal business.
[0052] The bid difference threshold may be a fixed value or a fixed ratio. When the bid difference threshold is a fixed ratio, the fixed ratio may correspond to the budget amount, and a corresponding value is calculated based on the budget amount and the fixed ratio and then compared with the difference.
[0053] The above implementation is described using a specific example. Assuming that the budget amount of a bidding project is 100,000 and the corresponding bid difference threshold is 5,000, after obtaining the historical recorded business corresponding to the bidding project, if there is a historical recorded business with a bid amount in the range of 100,000 to 100,500, the historical recorded business is judged to be an abnormal business.
[0054] In some implementations, the preset judgment rule may also be a dishonest judgment rule. Judging whether a historically recorded business is an abnormal business according to the preset judgment rule may be to determine whether there is a dishonest record in the business information of the historically recorded business after obtaining the historically recorded business. The dishonest record may be a list of arrears, a list of tax arrears, a list of abnormal taxpayers, a list of restricted high consumption, a list of restricted exit, a dishonest old defaulter, a list of civil and commercial judgment documents, a list of criminals and suspects, a list of administrative violations, and public execution information. If there is a dishonest record in the business information, the corresponding historically recorded business may be identified as an abnormal business.
[0055] In order to prevent users who submit historical record services from failing to add corresponding default records in the service information, the method for obtaining default records may also be the user identity or institution identification of the user who submits the historical record service, and using the user identity or institution identification to query whether there is a default record in the corresponding historical record.
[0056] Using a specific example to illustrate, when a historical business is found, it is determined that the submitter of the historical business is on a list of people restricted from high consumption. Then, executing the historical business may affect the normal processing of the business, and therefore the historical business can be identified as an abnormal business.
[0057] In actual applications, the method of identifying abnormal services in historical record services may not be limited to the above implementation mode, and the specific method may be adjusted according to the actual application situation, which will not be described in detail here.
[0058] It should be noted that the acquired historically recorded services may have certain processing results, such as normal processing or rejection processing, but the processing results do not affect the determination of abnormal services. For example, if a historically recorded service was processed as a normal service in the previous processing process, but it is determined to be an abnormal service when judging by using the preset judgment rules, the historically recorded service is still determined to be an abnormal service, and the previous processing results have no effect on it.
[0059] S130: construct at least one training sample business using the business information of the historically recorded business including the abnormal business in the judgment result; the number of the training sample business is greater than the number of the abnormal business.
[0060] The training sample business is the sample data for training the model targeted by the embodiments of this specification. After obtaining the historically recorded business that contains abnormal business in the judgment result, the business information of the historically recorded business can be used to construct the training sample business. In actual applications, due to the differences in the application conditions of the historically recorded business, the number of the historically recorded businesses may be small. For example, when the historically recorded business is a bidding business, there may not be many abnormal bidding businesses in real life, and the number of historically recorded businesses finally obtained may be small. Therefore, it is necessary to expand the number of the historically recorded businesses, that is, to construct training sample businesses using historically recorded businesses, so that the number of training sample businesses is greater than the number of abnormal businesses, thereby completing the expansion of the training samples.
[0061] It should be noted that, when there are multiple preset judgment rules corresponding to the historically recorded business, there may be multiple judgment results corresponding to the same historically recorded business. As long as there is a judgment result that identifies the historically recorded business as an abnormal business among these judgment results, the business information of the historically recorded business is used to construct the training sample business in this step.
[0062] After obtaining the historically recorded business that includes abnormal business in the judgment result, the training sample business can be constructed using the historically recorded business. Specifically, the historically recorded business that includes abnormal business in the judgment result can be marked as the initial abnormal business, and then the business information of the initial abnormal business is divided into business feature information corresponding to different business features, and then the business feature information of different initial abnormal businesses is respectively combined to obtain candidate sample businesses. Then, the candidate sample business is divided into at least two candidate sample business sets according to the similarity between the candidate sample businesses. The candidate sample business set is analyzed, and if the candidate sample business includes the initial abnormal business, the candidate sample business in the candidate sample business set is used as the training sample business.
[0063] Business features can be different types of business information. For example, when the business information contains the identity information of the user submitting the business, the business data of the business itself, and the information of the organization submitting the business, the corresponding business feature information can be separated from the business information based on the business features such as the business information itself, the business owner information, and the business organization information. In actual applications, the definition of the business features and the separation of business feature information can be set according to the needs of actual applications, and are not limited to the above examples. For example, the business information can also be derived from features through the Featuretools method, which will not be repeated here.
[0064] After determining the characteristic information of each business, the business characteristic information corresponding to different business characteristics can be combined respectively to obtain the combined candidate sample business. Since business information can be divided into multiple business characteristics, the obtained candidate sample business can be increased exponentially by combining business characteristic information. For example, when 10 initial abnormal businesses are obtained, three business characteristics are divided for the initial abnormal businesses, and 1,000 candidate sample businesses can be obtained by combination. Even if there may be duplicate businesses among these candidate sample businesses, the number of candidate sample businesses is greatly increased.
[0065] Since the number of services is exponentially magnified by combining different service feature information, it is easy to obtain too many candidate sample services. Training the model using candidate sample services will have a high computational cost, thus consuming more time and resources. Therefore, after obtaining the candidate sample services, certain services can be selected from the candidate sample services as training sample services to train the model.
[0066] In some implementations, the candidate sample services may be divided into a plurality of candidate sample service sets according to the similarity between the candidate sample services. For example, the candidate sample services may be classified by a KNN classification algorithm, thereby dividing the candidate sample services into different candidate sample service sets. The specific classification categories and the number of categories may be set according to the needs of the actual application, and will not be described in detail.
[0067] After the candidate sample services are classified, the candidate sample services in each candidate sample service set have high similarity. By analyzing the candidate sample services in different candidate sample service sets, if there are initial abnormal services in these candidate sample services, then due to the similarity between services, the possibility that other services in the candidate sample service set are abnormal services is also higher, so the candidate sample services in the candidate sample service set containing the initial abnormal services can be used as training sample services.
[0068] By constructing candidate sample businesses and screening out corresponding training sample businesses from the candidate sample businesses, the number of samples is expanded, overcoming the problem of insufficient number of samples for training the model. After screening the training sample businesses, it not only ensures that the screened samples have a strong correlation with the model, but also reduces the time and resources required for training the model, which is conducive to the training of the model in subsequent steps.
[0069] S140: Training an abnormal service identification model based on the training sample services; the abnormal service identification model is used to identify abnormal services.
[0070] After the training sample services are acquired, the training sample services can be used to train the abnormal service identification model. The abnormal service identification model can be used to identify abnormal services. By inputting the service information of a service into the abnormal service identification model, it can be determined whether the service is an abnormal service.
[0071] In the embodiments of the present specification, the abnormal business identification model may be a mathematical model, such as a Bayesian classification model, a support vector machine classification model (Support Vector Machine, SVM), or a convolutional neural network classification model (Convolutional Neural Networks, CNN), etc.
[0072] In some embodiments, the abnormal business identification model may be a model trained based on a KNN classification algorithm. The KNN classification algorithm is the K nearest neighbor method. When most of the K closest samples in the feature space belong to a certain category, it can be determined that the sample belongs to the category. The KNN classification algorithm determines the distance between each object as a non-similar indicator between objects to avoid mutual matching between objects. The distance between objects can be Euclidean distance or Manhattan distance. The Euclidean distance can be calculated using the formula Calculate, where d(x,y) is the Euclidean distance between point x and point y, and x k and k are the coordinates of point x and point y in dimension k respectively; Manhattan distance can be calculated using the formula To calculate, where d′(x,y) is the Manhattan distance between point x and point y, x k and k are the coordinates of point x and point y in the k dimension respectively.
[0073] Specifically, the process of using the KNN algorithm to calculate can be to calculate the distance between a certain training data and other training data, and sort them according to the increasing relationship of the distance. From these distances, select K points with the smallest distance, and determine the frequency of occurrence of the category of these K points, and use the category with the highest frequency as the predicted classification category of the training data. Among them, K can be set by the user.
[0074] The above implementation is only an introduction to a specific example of the abnormal business identification model. In practical applications, any machine learning model that can identify corresponding abnormal business based on the business data of sample business can be used as the abnormal business identification model for training, which will not be described in detail.
[0075] Through the introduction of the embodiments of the above method, it can be seen that after obtaining the historical record samples, the model training method screens out abnormal businesses in the historical record samples according to the preset judgment conditions, and uses the business information corresponding to the abnormal business to expand, thereby obtaining a training sample business whose number is greater than the abnormal business, so that the business identification model can be trained using the training sample business. When the sample size used to train the model is insufficient, the above method expands the training sample size based on the information of the pre-screened business, thereby ensuring the accuracy of the trained model, which is conducive to using the model in subsequent steps to complete the identification of abnormal business and ensure the normal operation of the business process.
[0076] The following is combined with Figure 2 The present specification describes a service processing method according to an embodiment of the present specification. The execution subject of the method is a service processing device, which includes but is not limited to a server, an industrial computer, a PC, etc. The service processing method includes the following specific implementation steps.
[0077] S210: Receive a pending business; the pending business corresponds to business information.
[0078] The pending business may be a business that has not been processed yet, but needs to be analyzed for abnormal conditions before being processed. For example, when the pending business is a bidding business, that is, a business of bidding for a certain bidding project, it can be first determined whether the bidding business is normal to avoid the situation where the bidding business is maliciously obtaining bidding items or maliciously raising the price of bidding items.
[0079] The pending business corresponds to business information. Business information is the information corresponding to the pending business. Specifically, the business information can be the information corresponding to the pending business itself, or it can be the identity information of the user who submitted the pending business. When the pending business corresponds to an institution, organization or company, the business information can also include the registration information of the corresponding institution, organization or company and the relevant identity information of the employed employees. In actual applications, the corresponding business information can be obtained according to the specific circumstances of the pending business, which is not limited to the above examples and will not be repeated here. Based on the business information, it can be identified whether the pending business is an abnormal business.
[0080] S220: Identify whether the to-be-processed business is an abnormal business based on the business information; wherein, including, using an abnormal business identification model to identify whether the to-be-processed business is an abnormal business; the abnormal business identification model includes, after identifying the abnormal business in the historical recorded business according to a preset judgment rule, using the business information corresponding to the abnormal business to construct a training sample business, and based on the model obtained by training the training sample business.
[0081] After obtaining the business information of the pending business, it is possible to identify whether the pending business is an abnormal business through the business information. Specifically, it is possible to identify whether the pending business is an abnormal business using an abnormal business identification model. The abnormal business identification model may be a model trained based on the business information corresponding to the sample training business, and it is possible to determine whether the pending business is an abnormal business by analyzing the business information of the pending business.
[0082] In the embodiments of the present specification, the abnormal business identification model may be a mathematical model, such as a Bayesian classification model, a support vector machine classification model (Support Vector Machine, SVM), or a convolutional neural network classification model (Convolutional Neural Networks, CNN), etc.
[0083] In some embodiments, the abnormal business identification model may be a model trained based on a KNN classification algorithm. The KNN classification algorithm is the K nearest neighbor method. When most of the K closest samples in the feature space belong to a certain category, it can be determined that the sample belongs to the category. The KNN classification algorithm determines the distance between each object as a non-similar indicator between objects to avoid mutual matching between objects. The distance between objects can be Euclidean distance or Manhattan distance. The Euclidean distance can be calculated using the formula Calculate, where d(x,y) is the Euclidean distance between point x and point y, and x k and k are the coordinates of point x and point y in dimension k respectively; Manhattan distance can be calculated using the formula To calculate, where d′(x,y) is the Manhattan distance between point x and point y, x k and k are the coordinates of point x and point y in the k dimension respectively.
[0084] Specifically, the process of using the KNN algorithm to calculate can be to calculate the distance between a certain training data and other training data, and sort them according to the increasing relationship of the distance. From these distances, select K points with the smallest distance, and determine the frequency of occurrence of the category of these K points, and use the category with the highest frequency as the predicted classification category of the training data. Among them, K can be set by the user.
[0085] In some embodiments, the model trained by the embodiments of this specification can be a model for identifying abnormal bids. When a seller needs to sell an item, a bidding method can be used, that is, after multiple buyers make bids for the above item, the seller decides the object to which the item is sold based on the buyers' bids, for example, it can be the buyer with the highest bid. However, in the bidding process, abnormal bidding may occur. These abnormal bidding phenomena may be collusion between buyers to lower the bid as much as possible, or buyers with bad credit maliciously raise the bid price, thereby affecting the normal progress of the bidding.
[0086] The specific method for obtaining the business allocation model can be referred to Figure 1 The introduction in the corresponding embodiments will not be repeated here.
[0087] S230: If the service to be processed is an abnormal service, feedback service rejection information is provided; the service rejection information is used to indicate a rejection of processing the service to be processed.
[0088] If the pending business is identified as an abnormal business through the business identification model, then accepting or executing the pending business may affect the normal progress of the business execution process. In this case, business rejection information can be fed back to the user who submitted the pending business. The business rejection information is used to indicate the rejection of the pending business, that is, to notify the submitting user that the pending business submitted by the submitting user will not be processed.
[0089] The abnormal business may be used to indicate a business that affects the processing of the business. For example, when the business is a bidding business, the abnormal business may be an abnormal bid that interferes with the bidding process or bidding results. For a specific definition of the abnormal business, please refer to Figure 1 The description of step S120 in the corresponding embodiment will not be repeated here.
[0090] In some implementations, when the business to be processed is a bidding business, that is, a bidding for a certain bidding project, the feedback business rejection information may be that the bid corresponding to the business to be processed is not accepted. In actual applications, after rejecting the bidding business, other methods may be used to process it, for example, the user corresponding to the abnormal bid may be marked as a key monitoring user, etc. Other implementations are not described in detail here.
[0091] In some implementations, if the business to be processed is identified by the business identification model and it is determined that the business to be processed is not an abnormal business, the business to be processed can be processed. The specific processing method can be to process the business to be processed based on the needs of the business to be processed in actual applications. For example, if the business to be processed is a bidding business, that is, bidding for a certain bidding project, the bidding business can be reviewed and analyzed as a bid that can be accepted.
[0092] Through the introduction of the above-mentioned business processing method embodiment, it can be seen that after obtaining the business to be processed, the business processing method uses the abnormal business identification model to determine whether the business to be processed is an abnormal business. In the case where the abnormal business identification model is a model obtained by training with a sufficient number of samples, the accuracy of identifying abnormal business using the abnormal business identification model is ensured, so that the occurrence of business processing abnormalities or related process damage caused by handling abnormal business can be avoided in the subsequent process, thereby improving the effectiveness of business processing.
[0093] based on Figure 1 The corresponding model training method introduces a model training device in the embodiment of this specification. The model training device is set in the model training device. Figure 3 As shown, the model training device may include the following modules.
[0094] A historical record service acquisition module 310 is used to acquire at least one historical record service; the historical record service corresponds to service information;
[0095] A judgment result determination module 320, configured to determine, based on the service information, a judgment result of the historically recorded service corresponding to at least one preset judgment rule; the judgment result includes normal service and abnormal service;
[0096] A training sample service construction module 330 is used to construct at least one training sample service by using service information of historically recorded services including abnormal services in the judgment result; the number of the training sample services is greater than the number of historically recorded services;
[0097] The abnormal business identification model training module 340 is used to train an abnormal business identification model based on the training sample business; the abnormal business identification model is used to identify abnormal business.
[0098] based on Figure 2 The corresponding service processing method introduces a service processing device in the embodiment of this specification. The service processing device is set in the service processing device. Figure 4 As shown, the business processing device includes the following modules.
[0099] The to-be-processed business receiving module 410 is used to receive the to-be-processed business; the to-be-processed business corresponds to business information;
[0100] The abnormal business identification module 420 is used to identify whether the to-be-processed business is an abnormal business based on the business information; wherein, it includes using an abnormal business identification model to identify whether the to-be-processed business is an abnormal business; the abnormal business identification model includes, after identifying the abnormal business in the historical record business according to a preset judgment rule, using the business information corresponding to the abnormal business to construct a training sample business, and based on the model trained by the training sample business;
[0101] The service rejection information feedback module 430 is used to feedback service rejection information when the service to be processed is an abnormal service; the service rejection information is used to indicate that the service to be processed is rejected.
[0102] based on Figure 1 Corresponding to the model training method, this specification embodiment provides a model training device. Figure 5 As shown, the model training device may include a memory and a processor.
[0103] In this embodiment, the memory may be implemented in any appropriate manner. For example, the memory may be a read-only memory, a mechanical hard disk, a solid-state hard disk, or a USB flash drive, etc. The memory may be used to store computer program instructions.
[0104] In this embodiment, the processor can be implemented in any appropriate manner. For example, the processor can take the form of a microprocessor or processor and a computer-readable medium storing a computer-readable program code (such as software or firmware) that can be executed by the (micro)processor, a logic gate, a switch, an application-specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller, etc. The processor can execute the computer program instructions to implement the following steps: obtaining at least one historical record business; the historical record business corresponds to business information; determining the judgment result of the historical record business corresponding to at least one preset judgment rule according to the business information; the judgment result includes normal business and abnormal business; constructing at least one training sample business using the business information of the historical record business including the abnormal business in the judgment result; the number of the training sample business is greater than the number of the historical record business; training an abnormal business identification model based on the training sample business; the abnormal business identification model is used to identify abnormal business.
[0105] based on Figure 2 Corresponding to the service processing method, this specification embodiment provides a service processing device. Figure 6 As shown, the business processing device may include a memory and a processor.
[0106] In this embodiment, the memory may be implemented in any appropriate manner. For example, the memory may be a read-only memory, a mechanical hard disk, a solid-state hard disk, or a USB flash drive, etc. The memory may be used to store computer program instructions.
[0107] In this embodiment, the processor can be implemented in any appropriate manner. For example, the processor can take the form of a microprocessor or processor and a computer-readable medium storing a computer-readable program code (such as software or firmware) that can be executed by the (micro)processor, a logic gate, a switch, an application-specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller, etc. The processor can execute the computer program instructions to implement the following steps: receiving a pending business; the pending business corresponds to business information; based on the business information of the pending business, identifying whether the pending business is an abnormal business; including, using an abnormal business identification model to identify whether the pending business is an abnormal business; the abnormal business identification model includes, after identifying an abnormal business in the historical record business according to a preset judgment rule, using the business information corresponding to the abnormal business to construct a training sample business, and based on the model obtained by training the training sample business; in the case that the pending business is an abnormal business, feedback business rejection information; the business rejection information is used to indicate the refusal to process the pending business.
[0108] In the 1990s, improvements to a technology could be clearly distinguished as hardware improvements (for example, improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the method flow). However, with the development of technology, many improvements to the method flow today can be regarded as direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved method flow into the hardware circuit. Therefore, it cannot be said that an improvement in a method flow cannot be implemented using a hardware entity module. For example, a programmable logic device (PLD) (such as a field programmable gate array (FPGA)) is such an integrated circuit whose logical function is determined by the user's programming of the device. Designers can "integrate" a digital system on a PLD by programming it themselves, without having to ask a chip manufacturer to design and produce a dedicated integrated circuit chip2. Moreover, nowadays, instead of manually making integrated circuit chips, this kind of programming is mostly implemented by "logic compiler" software, which is similar to the software compiler used when developing and writing programs, and the original code before compilation must also be written in a specific programming language, which is called hardware description language (HDL). There is not only one HDL, but many kinds, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. The most commonly used ones are VHDL (Very-High-Speed Integrated Circuit Hardware Description Language) and Verilog2. Those skilled in the art should also know that it is only necessary to program the method flow slightly in the above-mentioned hardware description languages and program it into the integrated circuit, and then it is easy to obtain the hardware circuit that implements the logic method flow.
[0109] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.
[0110] It can be known from the above description of the implementation mode that the technicians in this field can clearly understand that the present specification can be implemented by means of software plus the necessary first hardware platform. Based on such an understanding, the technical solution of the present specification can be essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present specification or some parts of the embodiments.
[0111] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.
[0112] This specification can be used in many first or special computer system environments or configurations, such as personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc.
[0113] This specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. This specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.
[0114] Although the present specification is described through embodiments, those skilled in the art will appreciate that there are many modifications and changes to the present specification without departing from the spirit of the present specification, and it is intended that the appended claims include these modifications and changes without departing from the spirit of the present specification.
Claims
1. A method for training an abnormal business identification model, characterized in that: include: Acquire at least one historically recorded service; the historically recorded service corresponds to service information; Determine, according to the service information, a judgment result of the historically recorded service corresponding to at least one preset judgment rule; the judgment result includes normal service and abnormal service; Using the business information of the historically recorded business including the abnormal business in the judgment result to construct at least one training sample business; the number of the training sample business is greater than the number of the historically recorded business; An abnormal business identification model is trained based on the training sample business; the abnormal business identification model is used to identify abnormal business; Among them, using the business information of the historically recorded business including the abnormal business in the judgment result to construct at least one training sample business; the number of the training sample business is greater than the number of the historically recorded business, including: Marking the historical record business containing abnormal business in the judgment result as the initial abnormal business; dividing the service information of the initial abnormal service into service feature information corresponding to different service features; Combining the service feature information of different initial abnormal services respectively to obtain candidate sample services; Dividing the candidate sample services into at least two candidate sample service sets according to the similarities between the candidate sample services; Analyze the candidate sample service set, and if the candidate sample service includes the initial abnormal service, use the candidate sample service in the candidate sample service set as a training sample service; The business information includes the identity information of the user who submitted the business, the business data of the business itself, and the information of the organization that submitted the business; Accordingly, dividing the business information of the initial abnormal business into business characteristic information corresponding to different business characteristics includes: separating corresponding business characteristic information from the business information according to business characteristics of the business itself, business owner information, and business organization information.
2. The method according to claim 1, characterized in that The historically recorded business includes bidding business; accordingly, the business information includes at least one of bidding company information, bidding company employee information, and bidding amount.
3. The method according to claim 1, characterized in that The historically recorded business corresponds to a target object; and the determining, according to the business information, that the historically recorded business corresponds to a judgment result of at least one preset judgment rule includes: Determining whether there is a correlation between business information of at least two historically recorded businesses corresponding to the same target object; If so, the historical recorded business with correlation between the business information is marked as abnormal business.
4. The method according to claim 1, characterized in that The historically recorded business corresponds to a bidding item; the business information includes a bidding amount; and the determination of a judgment result that the historically recorded business corresponds to at least one preset judgment rule according to the business information includes: Obtaining bid amounts of at least two historically recorded businesses corresponding to the same bidding project; When the difference between different bid amounts is greater than the maximum bid amount difference or less than the minimum bid amount difference, the historically recorded business corresponding to the bid amount is determined to be an abnormal business.
5. The method according to claim 1, characterized in that The historically recorded business corresponds to a bidding project; the bidding project corresponds to a budget amount; the business information of the historically recorded business includes the bidding amount; and the determination of the judgment result of the historically recorded business corresponding to at least one preset judgment rule according to the business information includes: Obtaining the budget amount of the bidding project corresponding to the historically recorded business; When the difference between the bid amount and the budget amount is smaller than a bid difference threshold, the historically recorded business is determined to be an abnormal business.
6. The method according to claim 1, characterized in that The determining, according to the service information, that the historically recorded service corresponds to a judgment result of at least one preset judgment rule comprises: Determine whether the business information contains a dishonest record; the dishonest record is used to indicate that the user corresponding to the historically recorded business has abnormal credit data; If so, it is determined that the historically recorded business corresponding to the business information is an abnormal business.
7. The method according to claim 1, characterized in that The abnormal business identification model is trained based on the training sample business, including: The abnormal business identification model is trained using the KNN algorithm based on the training sample business.
8. A model training device, characterized in that: include: A historical record service acquisition module, used to acquire at least one historical record service; the historical record service corresponds to service information; A judgment result determination module, used to determine the judgment result of the historically recorded business corresponding to at least one preset judgment rule according to the business information; the judgment result includes normal business and abnormal business; A training sample service construction module, configured to construct at least one training sample service by using service information of historically recorded services including abnormal services in the judgment result; the number of the training sample services is greater than the number of the historically recorded services; An abnormal business identification model training module is used to train an abnormal business identification model based on the training sample business; the abnormal business identification model is used to identify abnormal business; Among them, using the business information of the historically recorded business including the abnormal business in the judgment result to construct at least one training sample business; the number of the training sample business is greater than the number of the historically recorded business, including: Marking the historical record business containing abnormal business in the judgment result as the initial abnormal business; dividing the service information of the initial abnormal service into service feature information corresponding to different service features; Combining the service feature information of different initial abnormal services respectively to obtain candidate sample services; Dividing the candidate sample services into at least two candidate sample service sets according to the similarities between the candidate sample services; Analyze the candidate sample service set, and if the candidate sample service includes the initial abnormal service, use the candidate sample service in the candidate sample service set as a training sample service; The business information includes the identity information of the user who submitted the business, the business data of the business itself, and the information of the organization that submitted the business; Accordingly, dividing the business information of the initial abnormal business into business characteristic information corresponding to different business characteristics includes: separating corresponding business characteristic information from the business information according to business characteristics of the business itself, business owner information, and business organization information.
9. A model training device comprising a memory and a processor; The memory is used to store computer program instructions; The processor is used to execute the computer program instructions to implement the following steps: obtaining at least one historically recorded business; the historically recorded business corresponds to business information; determining, according to the business information, a judgment result of the historically recorded business corresponding to at least one preset judgment rule; the judgment result includes normal business and abnormal business; constructing at least one training sample business using the business information of the historically recorded business including the abnormal business in the judgment result; the number of the training sample business is greater than the number of the historically recorded business; training an abnormal business identification model based on the training sample business; the abnormal business identification model is used to identify abnormal business; Constructing at least one training sample business using business information of historically recorded businesses including abnormal businesses in the judgment results; The number of the training sample services is greater than the number of the historically recorded services, including: Marking the historical record business containing abnormal business in the judgment result as the initial abnormal business; dividing the service information of the initial abnormal service into service feature information corresponding to different service features; Combining the service feature information of different initial abnormal services respectively to obtain candidate sample services; Dividing the candidate sample services into at least two candidate sample service sets according to the similarities between the candidate sample services; Analyze the candidate sample service set, and if the candidate sample service includes the initial abnormal service, use the candidate sample service in the candidate sample service set as a training sample service; The business information includes the identity information of the user who submitted the business, the business data of the business itself, and the information of the organization that submitted the business; Accordingly, dividing the business information of the initial abnormal business into business characteristic information corresponding to different business characteristics includes: separating corresponding business characteristic information from the business information according to business characteristics of the business itself, business owner information, and business organization information.
10. A business processing method, characterized in that: include: Receiving pending business; the pending business corresponds to business information; Identify whether the to-be-processed business is an abnormal business based on the business information of the to-be-processed business; Among them, including, using the abnormal business identification model established by the abnormal business identification model training method of any one of claims 1 to 7 to identify whether the to-be-processed business is an abnormal business; the abnormal business identification model includes, after identifying the abnormal business in the historical record business according to the preset judgment rule, using the business information corresponding to the abnormal business to construct a training sample business, and based on the model obtained by training the training sample business; If the service to be processed is an abnormal service, feeding back service rejection information; The service rejection information is used to indicate a rejection of processing the service to be processed.
11. The method according to claim 10, characterized in that The to-be-processed business includes a bidding business; and when the to-be-processed business is an abnormal business, feeding back business rejection information includes: In the case that the business to be processed is an abnormal business, the bid corresponding to the business to be processed will not be accepted.
12. The method according to claim 10, characterized in that After identifying whether the to-be-processed service is an abnormal service based on the service information, the method further includes: If the to-be-processed business is not an abnormal business, the to-be-processed business is processed.
13. A service processing device, characterized in that: include: A pending business receiving module is used to receive pending business; the pending business corresponds to business information; An abnormal business identification module, used to identify whether the to-be-processed business is an abnormal business based on the business information of the to-be-processed business; Wherein, including, using an abnormal business identification model to identify whether the to-be-processed business is an abnormal business; the abnormal business identification model includes, after identifying the abnormal business in the historical recorded business according to a preset judgment rule, using the business information corresponding to the abnormal business to construct a training sample business, and a model obtained by training based on the training sample business, wherein the abnormal business identification model is established according to the abnormal business identification model training method according to any one of claims 1 to 7; The service rejection information feedback module is used to feedback service rejection information when the service to be processed is an abnormal service; the service rejection information is used to indicate that the service to be processed is rejected.
14. A business processing device comprising a memory and a processor; The memory is used to store computer program instructions; The processor is used to execute the computer program instructions to implement the following steps: receiving a pending business; the pending business corresponds to business information; Identify whether the to-be-processed business is an abnormal business based on the business information of the to-be-processed business; Among them, including, using the abnormal business identification model established by the abnormal business identification model training method of any one of claims 1 to 7 to identify whether the to-be-processed business is an abnormal business; the abnormal business identification model includes, after identifying the abnormal business in the historical recorded business according to the preset judgment rule, using the business information corresponding to the abnormal business to construct a training sample business, and based on the model obtained by training the training sample business; in the case that the to-be-processed business is an abnormal business, feedback of business rejection information; The service rejection information is used to indicate a rejection of processing the service to be processed.
Citation Information
Patent Citations
Training method and device for active risk real-time identification model
CN109034209A
Service system monitoring method and device
CN110275814A
Abnormal bidding behavior detection and evaluation method and system based on data driving
CN111028050A
Risk prediction model training method, risk prediction method and related device
CN111062563A
Electronic bid inviting purchase abnormal behavior recognition method, system and device, and storage medium
CN111091259A
