Model migration method and device, electronic equipment and computer readable medium

Abstract

The embodiment of the invention discloses a model migration method and device, electronic equipment and a computer readable medium. A specific embodiment of the method comprises the following steps: classifying each piece of service user information; generating portrait information of each service user; dividing each business user customer group, combining each sub-business user information group set to obtain each business customer group fusion group, and generating each piece of matching degree information; determining the sub-service early warning model meeting a preset matching degree condition as a target sub-service early warning model; inputting the detection request information into a target sub-service early warning model to obtain early warning information; and migrating the sub-service early warning model meeting a preset model migration condition. According to the embodiment, the model can be adaptively migrated in a high-concurrency scene to reduce the load of the server, and one type of detection request information can be processed through the single target server, so that the response speed can be improved, and the downtime frequency of the server can be reduced.

Description

Technical Field

[0001] Embodiments of the present disclosure relate to the field of computer technology, and in particular to a model migration method, device, electronic device, and computer-readable medium. Background Art

[0002] When the business warning model matching the user information is used to detect the warning information of the user information, the accuracy of the detected warning information can be improved. At present, when using various business warning models to detect the warning information of the user, the method usually adopted is: first, directly use the business user information samples to train each business warning model. Then, the trained business warning models are integrated into the same server to facilitate the detection of the user's warning information.

[0003] However, in practice, it is found that when using the above method to use various business warning models to detect user warning information, the following technical problems often occur:

[0004] The trained business warning models are integrated into one server. When the number of detection request information input into the trained business warning models is large, the server load is large, resulting in a long response time of the server and a high number of server downtimes.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the inventive concept and therefore it may contain information that does not form the prior art that is already known in this country to a person of ordinary skill in the art. Summary of the invention

[0006] The content of this disclosure is used to introduce concepts in a brief form, which will be described in detail in the detailed implementation section below. The content of this disclosure is not intended to identify the key features or essential features of the technical solution claimed for protection, nor is it intended to limit the scope of the technical solution claimed for protection.

[0007] Some embodiments of the present disclosure propose a model migration method, an apparatus, an electronic device, and a computer-readable medium to solve one or more of the technical problems mentioned in the above background technology section.

[0008] In the first aspect, some embodiments of the present disclosure provide a model migration method, the method comprising: generating each business customer group fusion group based on the acquired information of each business user; in response to the detection request information of the detected business warning information, generating the matching degree information between the above detection request information and each business customer group fusion group in the above business customer group fusion group according to the above detection request information and the above business customer group fusion groups, to obtain each matching degree information; according to the above matching degree information, determining the sub-business warning model that meets the preset matching degree condition in each sub-business warning model trained by the above business customer group fusion group as the target sub-business warning model corresponding to the above detection request information, so as to adapt the sub-business warning model, wherein the sub-business warning model in the above-mentioned pre-trained sub-business warning model corresponds to the business customer group fusion group in the above-mentioned each business customer group fusion group, and the above-mentioned each sub-business warning model meets the preset matching degree condition. The sub-business warning model in the service warning model is obtained by training the corresponding business customer group fusion group; the above-mentioned detection request information is input into the above-mentioned target sub-business warning model to obtain the warning information corresponding to the above-mentioned detection request information; in response to detecting that the above-mentioned warning information meets the preset alarm conditions, the request processing detail information of each sub-business warning model in the above-mentioned sub-business warning models is obtained to obtain each request processing detail information, wherein the request processing detail information in the above-mentioned each request processing detail information includes request processing quantity information and request processing duration information; according to the above-mentioned each request processing detail information, it is determined whether there is a sub-business warning model that meets the preset model migration conditions in the above-mentioned each sub-business warning model; in response to detecting that there is a sub-business warning model that meets the preset model migration conditions in the above-mentioned each sub-business warning model, the sub-business warning model that meets the above-mentioned preset model migration conditions is migrated.

[0009] In a second aspect, some embodiments of the present disclosure provide a model migration device, including a first generating unit, configured to generate each business customer group fusion group according to each business user information obtained; a second generating unit, configured to generate, in response to detection request information of detected business warning information, matching information between the detection request information and each business customer group fusion group in the above-mentioned business customer group fusion groups according to the above-mentioned detection request information and the above-mentioned business customer group fusion groups, to obtain each matching information; a first determining unit, configured to determine, according to the above-mentioned matching information, the sub-business warning model that meets the preset matching condition in each sub-business warning model trained by the above-mentioned business customer group fusion groups as the target sub-business warning model corresponding to the above-mentioned detection request information, so as to perform adaptation processing on the sub-business warning model, wherein the sub-business warning model in the above-mentioned pre-trained sub-business warning models corresponds to the business customer group fusion group in the above-mentioned business customer group fusion groups, and the above-mentioned sub-business warning models The sub-business warning model in is obtained by training the corresponding business customer group fusion group; the input unit is configured to input the above-mentioned detection request information into the above-mentioned target sub-business warning model to obtain the warning information corresponding to the above-mentioned detection request information; the acquisition unit is configured to obtain the request processing detail information of each sub-business warning model in the above-mentioned sub-business warning models in response to detecting that the above-mentioned warning information meets the preset alarm condition, and obtain each request processing detail information, wherein the request processing detail information in the above-mentioned each request processing detail information includes request processing quantity information and request processing duration information; the second determination unit is configured to determine whether there is a sub-business warning model that meets the preset model migration condition in the above-mentioned each sub-business warning model according to the above-mentioned each request processing detail information; the migration unit is configured to migrate the sub-business warning model that meets the above-mentioned preset model migration condition in response to detecting that there is a sub-business warning model that meets the preset model migration condition in the above-mentioned each sub-business warning model.

[0010] In a third aspect, some embodiments of the present disclosure provide an electronic device comprising: one or more processors; a storage device on which one or more programs are stored, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation manner of the above-mentioned first aspect.

[0011] In a fourth aspect, some embodiments of the present disclosure provide a computer-readable medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the method described in any one of the implementations of the first aspect or the second aspect is implemented.

[0012] The above-mentioned embodiments of the present disclosure have the following beneficial effects: through a model migration method of some embodiments of the present disclosure, the server load can be reduced, the server response time can be shortened, and the number of server downtimes can be reduced. The reason for the large server load, long server response time, and large number of server downtimes is that the trained business warning models are integrated into one server. When the number of detection request information input into the trained business warning models is large, the server load is large, resulting in a long server response time and a large number of server downtimes. Based on this, the model migration method of some embodiments of the present disclosure first generates each business customer group fusion group according to the obtained business user information. Thus, each sub-business user information group set can be obtained, which can be further used to train each business warning model. Afterwards, in response to the detection request information of the business warning information, according to the above-mentioned detection request information and the above-mentioned business customer group fusion group, the matching degree information between the above-mentioned detection request information and each business customer group fusion group in the above-mentioned business customer group fusion group is generated to obtain each matching degree information. Thus, the matching degree of the detection request information and each business customer group fusion group can be obtained. Secondly, according to the above-mentioned matching degree information, the sub-business warning model that meets the preset matching degree condition in each pre-trained sub-business warning model is determined as the target sub-business warning model corresponding to the above-mentioned detection request information, so as to adapt the sub-business warning model, wherein the sub-business warning model in each pre-trained sub-business warning model corresponds to the business customer group fusion group in the above-mentioned business customer group fusion group, and the sub-business warning model in each sub-business warning model is obtained by training the corresponding business customer group fusion group. Thus, an adapted sub-business warning model can be obtained. Afterwards, the above-mentioned detection request information is input into the above-mentioned target sub-business warning model to obtain the warning information corresponding to the above-mentioned detection request information. Thus, the warning information of the target user can be obtained. Secondly, in response to detecting that the above-mentioned warning information meets the preset alarm condition, the request processing detail information of each sub-business warning model in the above-mentioned sub-business warning model is obtained, and each request processing detail information is obtained, wherein the request processing detail information in the above-mentioned request processing detail information includes request processing quantity information and request processing duration information. Thus, the request processing detail information of each sub-business warning model can be obtained. Afterwards, according to the above-mentioned request processing details information, it is determined whether there is a sub-business warning model that meets the preset model migration conditions among the above-mentioned sub-business warning models. Thus, it can be determined whether there is a business warning model to be migrated. Finally, in response to detecting that there is a sub-business warning model that meets the preset model migration conditions among the above-mentioned sub-business warning models, migration processing is performed on the sub-business warning model that meets the preset model migration conditions. Thus, the sub-business warning model that meets the migration conditions can be migrated to reduce the load on the server.Also, when there is a business warning model that meets the above-mentioned preset migration conditions in the server, the business warning model that meets the above-mentioned preset migration conditions will be migrated, so that the model can be adaptively migrated in a high-concurrency scenario to reduce the server load, and one type of detection request information can be processed by a separate target server, thereby improving the response speed and reducing the number of server downtimes. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that components and elements are not necessarily drawn to scale.

[0014] Figure 1 is a flow chart of some embodiments of the model migration method according to the present disclosure;

[0015] Figure 2 is a schematic diagram of the structure of some embodiments of the model migration device according to the present disclosure;

[0016] Figure 3 It is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION

[0017] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0018] It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings. In the absence of conflict, the embodiments and features in the embodiments of the present disclosure can be combined with each other.

[0019] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0020] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0021] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0022] With regard to the collection, storage, and use of user personal information (such as borrowing and repayment records, user ID numbers) involved in this disclosure, relevant organizations or individuals must fulfill obligations including conducting personal information security impact assessments, fulfilling the obligation to inform the personal information subject, and obtaining the prior authorization and consent of the personal information subject before performing the corresponding operations.

[0023] The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0024] Figure 1 The flow chart 100 of some embodiments of the model migration method according to the present disclosure is shown. The model migration method comprises the following steps:

[0025] Step 101: Generate various service customer group integration groups based on the acquired service user information.

[0026] In some embodiments, the execution subject (e.g., computing device) of the model migration method may generate each business customer group fusion group according to each business user information obtained. Among them, the above-mentioned each business user information may be obtained from a database storing business user information. Each business user information in the above-mentioned each business user information may include each data information. The data information in each data information included in the above-mentioned each business user information may represent the user identification, user name, or borrowing and returning business (loan) record information. The above-mentioned user identification may represent the user's ID number. The above-mentioned user name may represent the user's name. The above-mentioned borrowing and returning business (loan) record information may include the user's loan amount (loan amount) and borrowing and returning business method (loan method). For example, the business user information may be: "Name: xx, ID number: xxxxxxxx, user borrowing and returning business (loan) record information: 50,000, cash loan". The business users represented by the business user information in the above-mentioned each business user information may be users applying for value circulation tools (credit card customer group), users handling consumer value borrowing and returning business (cash borrowing and returning business customer group), users handling vehicle value related information business (automobile related value borrowing and returning business customer group) or users of the network service platform (e-commerce customer group).

[0027] In some optional implementations of some embodiments, the execution subject may generate various service customer group integration groups according to the acquired information of various service users through the following steps:

[0028] The first step is to classify the above-mentioned business user information to obtain various business user customer groups.

[0029] The second step is to generate business user portrait information corresponding to each of the business user information according to the business user information. The business user portrait information in each of the business user portrait information corresponds to the business user information in the business user information. The business user portrait information in each of the business user portrait information can represent the portrait of the business user information.

[0030] The third step is to divide the above-mentioned various business user customer groups according to the above-mentioned various business user portrait information, and obtain various sub-business user information group sets corresponding to the above-mentioned various business user customer groups.

[0031] The fourth step is to determine the business customer group fusion groups corresponding to the obtained sub-business user information groups according to the obtained sub-business user information groups. Among them, the business customer group fusion groups in the above-mentioned business customer group fusion groups can be obtained by combining the sub-business user information groups included in the obtained sub-business user information group sets. For example, each sub-business user information group set may include: "The sub-business user information group set includes Group A1, Group A2, and Group A3; the sub-business user information group set includes Group B1, Group B2, and Group B3; the sub-business user information group set includes Group C1, Group C2, and Group C3; then the business customer group fusion groups in each business customer group fusion group may include Group A1, Group B1, and Group C2".

[0032] In some optional implementations of some embodiments, the execution subject may classify the above-mentioned various business user information through the following steps to obtain various business user customer groups:

[0033] In the first step, each of the business user information is enhanced to obtain the enhanced business user information as each of the business user information.

[0034] The second step is to classify the above-mentioned various business user information to obtain various business user customer groups. Among them, the business user customer groups in the above-mentioned various business user customer groups can represent various business user information with the same borrowing and returning business method included in the above-mentioned various business user information. In practice, the above-mentioned execution entity can determine the various business user information with the same borrowing and returning business method included in the above-mentioned various business user information as business user customer groups to obtain various business user customer groups.

[0035] In the process of adopting technical solutions to solve the above-mentioned technical problem one, the following technical problem two is often accompanied: when the business user information samples are directly used to train various business warning models, the business user information samples for training the business warning models are not preprocessed. When the business user information samples contain a lot of redundant and erroneous data, the training cycle of the business warning model based on the samples containing a lot of redundant and erroneous data is long, resulting in a lot of wasted computing resources in training the business warning model. For these problems, the conventional solution is generally to only delete the repeated sample user information in the business user information samples. The inventors took into account the shortcomings of only deleting the repeated sample user information in the business user information samples, and combined with the advantages of the inventor's company in machine learning, we decided to adopt the following solution:

[0036] In some optional implementations of some embodiments, the execution subject may perform enhancement processing on the above-mentioned individual service user information through the following steps to obtain the enhanced individual service user information as individual service user information:

[0037] The first step is to convert each piece of data information in the above-mentioned business user information that meets the preset conversion conditions to obtain each piece of business user information after conversion. The above-mentioned preset conversion condition may be that the data information represents the borrowing and returning business mode of the business user. In practice, the above-mentioned execution subject may convert each piece of data information in the above-mentioned business user information that meets the preset conversion conditions by a unique hot encoding method to obtain each piece of business user information after conversion.

[0038] In the second step, for each piece of data information in each business user information after the above conversion processing, perform the following steps:

[0039] The first sub-step is to determine the nearest neighbor node information corresponding to the above data information according to the pre-trained business user structure tree. The business user structure tree can be a KD-Tree constructed by a business user information set. The data structure of the business user information in the business user information set is the same as the data structure of the business user information in the above business user information. The business users represented by the business user information in the business user information set can be users who apply for value circulation tools (credit card customers), users who handle consumer value borrowing and returning services (cash borrowing and returning services customers), users who handle vehicle value related information services (automobile related value borrowing and returning services customers) or users of network service platforms (e-commerce customers). Each business user information in the above business user information can include each data information, and each data information in the included data information can be a user's borrowing and returning business (loan) record information. The nearest neighbor node information in the above nearest neighbor node information can represent the nearest neighbor node of the above data information. The nearest neighbor node information in the above nearest neighbor node information can represent each node in the business user structure tree with the smallest distance between the node corresponding to the above data information. The node in each node in the above business user structure tree can correspond to the data information in the above business user information. In practice, the execution entity may determine the nearest neighbor nodes corresponding to the data information through a pre-trained business user structure tree.

[0040] The second sub-step is to determine the quantity information corresponding to each of the nearest neighbor node information, wherein the quantity information may represent the quantity of each of the nearest neighbor node information.

[0041] In the third sub-step, in response to determining that the quantity information is greater than the preset number of adjacent points, the data point corresponding to the data information is determined as a target data point. The target data point may represent a node of the data information in the pre-trained business user structure tree. The specific value of the preset number of adjacent points is not specifically limited.

[0042] In the fourth sub-step, clustering is performed on each business user information after the conversion according to the determined target data point to obtain each business user information after the clustering. In practice, the execution subject may cluster each business user information after the conversion by a clustering algorithm to obtain each business user information after the clustering. For example, the clustering algorithm may be DBSCAN.

[0043] The fifth sub-step is to determine the noise points in the information of each service user after clustering as target noise points.

[0044] In the third step, each data information corresponding to each determined target noise point is deleted from the above-mentioned each service user information, so as to update the above-mentioned each service user information and obtain the updated each service user information as each service user information.

[0045] The fourth step is to determine the importance information of each data information included in the above-mentioned business user information to obtain each importance information. Among them, the importance information in the above-mentioned importance information can represent the importance of the data. In practice, the above-mentioned execution entity can determine the importance information of each data information included in the above-mentioned business user information through the TF-IDF algorithm.

[0046] In the fifth step, the business user information is subjected to word-memory masking according to the above-mentioned importance information to obtain the business user information after word-memory masking. In practice, the execution subject can perform word-memory masking on the business user information by using the Mask word-memory masking method to obtain the business user information after word-memory masking.

[0047] In the sixth step, the business user information after the above-mentioned word-gram masking processing is subjected to word-gram prediction processing to obtain the business user information after the word-gram prediction processing, so as to obtain the business user information after the data enhancement processing as the business user information. In practice, the execution subject can perform word-gram prediction processing on the business user information after the above-mentioned word-gram masking processing through the BERT prediction method to obtain the business user information after the word-gram prediction processing.

[0048] The above technical solution, as an inventive point of an embodiment of the present disclosure, solves the technical problem that "the cycle of training the business warning model is long, and more computing resources are wasted in training the business warning model". The factors that lead to a long cycle of training the business warning model and a large number of computing resources wasted in training the business warning model are often as follows: directly using business user information samples to train each business warning model, without preprocessing the business user information samples for training the business warning model. If the above factors are solved, the effect of reducing the computing resources wasted in training the business warning model can be achieved. In order to achieve this effect, when generating warning information, the present disclosure first performs data cleaning processing on the business user information to clear the erroneous data, and then performs data enhancement processing on the business user information after data cleaning, so as to realize the preprocessing of the business user information samples for training the business warning model, so as to reduce the redundant and erroneous data included in the business user information used for training the business warning model, thereby shortening the cycle of training the business warning model based on samples containing more redundant and erroneous data and reducing the computing resources wasted in training the business warning model.

[0049] In some optional implementations of some embodiments, the execution subject may generate each business user portrait information corresponding to each business user information according to each business user information through the following steps:

[0050] In the first step, for each piece of business user information in the above business user information, perform the following steps:

[0051] The first sub-step is to perform word segmentation processing on the business user information to obtain information about each word. The information about each word can represent each word included in the business user information. In practice, the execution subject can perform word segmentation processing on the business user information by using the Jieba word segmentation method to obtain information about each word.

[0052] The second sub-step is to perform redundancy elimination processing on each word information obtained, and obtain each word information after redundancy elimination as each target word information. In practice, first, the above-mentioned execution subject can determine the TF-IDF value corresponding to each word information in the above-mentioned word information through the TF-IDF method to obtain each TF-IDF value. Then, the word information whose corresponding TF-IDF value in each word information obtained is less than the preset value is deleted to perform redundancy elimination processing on each word information obtained. Here, the specific value of the above-mentioned preset value is not limited.

[0053] The third sub-step is to generate a vocabulary relationship table corresponding to each of the target word information according to the above-mentioned target word information. The above-mentioned vocabulary relationship table can represent a document-term matrix. In practice, the above-mentioned execution subject can create a vocabulary relationship table corresponding to each of the target word information according to the above-mentioned target word information and each TF-IDF value corresponding to each of the target word information through the TfidfVectorizer class.

[0054] The fourth sub-step is to input the above-mentioned vocabulary relationship table and the preset topic quantity information into the pre-trained topic information generation model to obtain each topic information corresponding to the above-mentioned each target word information. Among them, the topic information in the above-mentioned each topic information may include a topic name and a topic probability. The above-mentioned topic probability may represent the probability that the topic information is the above-mentioned each target word information. The above-mentioned preset topic quantity information may represent the preset number of topics. For example, the preset topic quantity information may be 5. For example, the subject information may be: "Topic name: Automobile-related value lending and returning business; Topic probability: 50%". The above-mentioned pre-trained topic information generation model may be a model that takes the vocabulary relationship table and the preset topic quantity information as input and each topic information as output. For example, the pre-trained topic information generation model may be an LDA model. The training data of the above-mentioned pre-trained topic information generation model may be each target vocabulary relationship table and the number of topics and each topic information corresponding to each target vocabulary relationship table in each target vocabulary relationship table. The target vocabulary relationship table in each target vocabulary relationship table has the same data structure as the above-mentioned vocabulary relationship table. The training method of the above-mentioned pre-trained topic information generation model may be trained by a batch training method.

[0055] The fifth sub-step is to generate business user portrait information corresponding to the business user information based on the above-mentioned subject information and the above-mentioned target word information. In practice, the above-mentioned execution subject can generate business user portrait information corresponding to the above-mentioned business user information based on the above-mentioned subject information and the above-mentioned target word information through a support vector machine (SVM).

[0056] In some optional implementations of some embodiments, the execution subject may divide the above-mentioned various business user customer groups according to the above-mentioned various business user portrait information through the following steps to obtain various sub-business user information group sets corresponding to the above-mentioned various business user customer groups:

[0057] The first step is to perform feature extraction processing on each of the above business user portrait information to obtain each portrait feature information. Among them, the portrait feature information in each of the above portrait feature information can correspond to the business user portrait information in the above each business user portrait information. In practice, the above execution subject can perform feature extraction processing on each of the above business user portrait information through a feature extraction algorithm to obtain each portrait feature information. For example, the feature extraction algorithm can be a principal component analysis (PCA) algorithm.

[0058] The second step is to perform standardization processing on the above-mentioned image feature information, and obtain the standardized image feature information as the image feature information of each target. In practice, the execution subject can perform standardization processing on the above-mentioned image feature information by Min-Max standardization method, and obtain the standardized image feature information as the image feature information of each target.

[0059] Step 3: For each of the above business user groups, perform the following steps:

[0060] In the first step, for each target portrait feature information corresponding to the above business user customer group, perform the following steps:

[0061] In the first sub-step, each target portrait feature information that is different from the above target portrait feature information among each target portrait feature information corresponding to the above business user group is determined as a target portrait feature information set.

[0062] The second sub-step is to determine the similarity information between the target portrait feature information and each target portrait feature information in the determined target portrait feature information set, and obtain each similarity information. The similarity information in each similarity information can represent the similarity between the target portrait feature information and the target portrait feature information in the target portrait feature information set. For example, the similarity information can be 50%. In practice, the execution subject can determine the similarity information between the target portrait feature information and each target portrait feature information in the determined target portrait feature information set by the cosine similarity method.

[0063] In the third sub-step, based on the above-mentioned similarity information, each target portrait feature information in the determined target portrait feature information set that meets the preset selection condition and the above-mentioned target portrait feature information are determined as a target portrait feature information group. The above-mentioned preset selection condition may be that the similarity between the target portrait feature information in the determined target portrait feature information set and the above-mentioned target portrait feature information is greater than a preset similarity value. Here, the specific value of the above-mentioned preset similarity value is not limited.

[0064] The fourth sub-step is to determine each business user information corresponding to the determined target portrait feature information group as a sub-business user information group.

[0065] The second step is to determine the determined sub-service user information groups as a sub-service user information group set.

[0066] However, it is found in practice that in the process of adopting technical solutions to solve the above technical problem 1, the following technical problems are often accompanied:

[0067] When training various business warning models directly using business user information samples as is commonly done, the following technical problems often occur: directly using business user information samples to train various business warning models, without classifying and storing the business user information samples used to train each business warning model, resulting in the types of business user information samples used to train the business warning models being relatively complex, leading to low pertinence of the trained business warning models, and low accuracy of the business warning models that match the corresponding detection request information from the trained business warning models. To address these issues, conventional solutions are generally to classify and store only based on the type of business user information samples. The inventors took into account the shortcomings of classifying and storing only based on the type of business user information samples, and combined with the advantages of the inventor's company in machine learning, we decided to adopt the following solution:

[0068] In some optional implementations of some embodiments, the execution subject may determine each service customer group fusion group corresponding to each obtained sub-service user information group according to each obtained sub-service user information group through the following steps:

[0069] The first step is to determine the fused portrait feature information corresponding to each sub-business user information group in the above-mentioned sub-business user information group sets, and obtain each fused portrait feature information. Among them, the fused portrait feature information in the above-mentioned fused portrait feature information can represent the features obtained after the feature splicing processing of each portrait feature information of the business user information group. In practice, the above-mentioned execution subject can perform feature splicing processing on each target portrait feature information corresponding to each sub-business user information group in the above-mentioned sub-business user information group sets to obtain fused portrait feature information.

[0070] In the second step, for each sub-service user information group set in the above sub-service user information group sets, the following steps are performed:

[0071] In the first sub-step, based on the above-mentioned fusion portrait feature information, each sub-service user information group in the above-mentioned sub-service user information group set that meets the preset fusion condition is determined as a fusion sub-service user information group, and each fusion sub-service user information group is obtained. Among them, the above-mentioned preset fusion condition can be that the similarity between each fusion portrait feature information corresponding to two sub-service user information groups is greater than the preset similarity. Here, the specific value of the above-mentioned similarity is not limited. In practice, the above-mentioned execution subject can determine each sub-service user information group in the sub-service user information group set that meets the preset fusion condition by the cosine similarity method.

[0072] The second sub-step is to perform the following steps for each fusion sub-service user information group in each fusion sub-service user information group obtained:

[0073] The first execution step is to determine the above-mentioned sub-service user information group sets and the sub-service user information group sets that are different from the above-mentioned sub-service user information group sets as different sub-service user information group sets.

[0074] The second execution step is to determine each different sub-service user information group that meets the preset selection condition in each different sub-service user information group set in the above-mentioned different sub-service user information group sets as a target different sub-service user information group set, and obtain each target different sub-service user information group set. Among them, the above-mentioned preset selection condition can be that the similarity between the different sub-service user information group and the above-mentioned fusion sub-service user information group is greater than the preset similarity. Here, the specific value of the above-mentioned preset similarity is not limited.

[0075] The third execution step is to randomly select a target different sub-service user information group set from the above target different sub-service user information group sets.

[0076] The fourth execution step is to delete the randomly selected target different sub-service user information group set from the above-mentioned target different sub-service user information group sets to obtain updated target different sub-service user information group sets.

[0077] A fifth execution step is to determine each target different sub-service user information group in the randomly selected target different sub-service user information group set and the above-mentioned fused sub-service user information group as a first service user information group to obtain each first service user information group;

[0078] The sixth execution step is to randomly select a preset number of target different sub-service user information from each target different sub-service user information group set in each updated target different sub-service user information group set to obtain each randomly selected target different sub-service user information group.

[0079] The seventh execution step is to determine each of the randomly selected target different sub-service user information groups and each of the first service user information in the above-mentioned first service user information groups as a second service user information group to obtain each second service user information group.

[0080] In the eighth execution step, each obtained second service user information group is subjected to redundancy elimination processing, and the obtained second service user information groups after redundancy elimination are used as each service customer group fusion group. In practice, the above-mentioned execution subject can perform redundancy elimination processing on each obtained second service user information group through a distributed deduplication algorithm, and the obtained second service user information groups after redundancy elimination are used as each service customer group fusion group.

[0081] The ninth execution step is to store the obtained fusion groups of each business customer group into a related database.

[0082] The above technical solution, as an inventive point of an embodiment of the present disclosure, solves the technical problem that "the targetedness of each trained business warning model is low, and the accuracy of the business warning model matching the corresponding detection request information from each trained business warning model is low". The factors that lead to the low targetedness of each trained business warning model and the low accuracy of the business warning model matching the corresponding detection request information from each trained business warning model are often as follows: directly using business user information samples to train each business warning model, without dividing and processing the business user information samples for training each business warning model. If the above factors are solved, the effect of improving the targetedness of each trained business warning model and improving the accuracy of the business warning model matching the corresponding detection request information from each trained business warning model can be achieved. In order to achieve this effect, the present disclosure generates warning information, first, classifies the business user information according to the portrait features, and then combines the business user information with high similarity, so that each business user information sample can be divided into fine-grained, further improving the targetedness of each trained business warning model and improving the accuracy of the business warning model matching the corresponding detection request information from each trained business warning model.

[0083] Step 102, in response to the detection request information of the business warning information, generating matching degree information between the detection request information and each business customer group fusion group according to the detection request information and each business customer group fusion group to obtain each matching degree information.

[0084] In some embodiments, in response to the detection request information of the business warning information, the execution subject may generate the matching degree information between the detection request information and each of the business customer group fusion groups according to the detection request information and the business customer group fusion groups, and obtain each matching degree information. Among them, the matching degree information in the above-mentioned each matching degree information may characterize the matching degree between the detection request information and the business customer group fusion group. The detection request information may be a request for characterizing the warning information of the detection target user. The detection request information may include the business user information of the target user. The warning information may characterize the risk level of the target user. For example, the warning information may be that the risk level of the business user is 50%.

[0085] In some optional implementations of some embodiments, for each of the above-mentioned business customer groups, the above-mentioned execution subject can generate a model based on the above-mentioned business customer group, the business user portrait information corresponding to the above-mentioned detection request information, and the above-mentioned preset fitness information to obtain the matching degree information between the above-mentioned business customer group and the above-mentioned detection request information. Among them, the above-mentioned preset fitness information generation model can be a model that takes the fusion portrait feature information corresponding to the business customer group and the business user portrait information corresponding to the detection request information as input, and takes the matching degree between the above-mentioned business customer group and the above-mentioned detection request information as output. For example, the preset fitness information generation model can be a collaborative filtering model. It should be noted that the training data for training the above-mentioned preset fitness information generation model can be each group of training data. Among them, each group of training data in the above-mentioned groups of training data can include the business user portrait information corresponding to the detection request information, the fusion portrait feature information corresponding to the business customer group, and the matching degree information. The included matching degree information can characterize the matching degree between the business customer group and the business user portrait information. The number of the above-mentioned groups of training data is not limited here. The method for training the above-mentioned preset fitness information generation model can be a batch training method. In practice, first, the execution subject may determine the business user profile information of the target user represented by the corresponding detection request information. Then, for each of the above business customer groups, the fusion portrait feature information corresponding to the above business customer group and the business user profile information corresponding to the above detection request information are input into the above preset adaptation degree information generation model to obtain the matching degree corresponding to each business customer group.

[0086] Step 103, according to each matching degree information, the sub-business warning model that meets the preset matching degree condition among the sub-business warning models trained by each business customer group fusion group is determined as the target sub-business warning model corresponding to the detection request information, so as to adapt the sub-business warning model.

[0087] In some embodiments, according to the above-mentioned various matching degree information, the above-mentioned execution subject can determine the sub-business warning model that meets the preset matching degree condition in each sub-business warning model trained by the above-mentioned various business customer group fusion groups as the target sub-business warning model corresponding to the above-mentioned detection request information, so as to adapt the sub-business warning model. Among them, the sub-business warning model in the above-mentioned pre-trained sub-business warning model corresponds to the business customer group fusion group in the above-mentioned various business customer group fusion groups. The sub-business warning model in the above-mentioned various sub-business warning models is obtained by training the corresponding business customer group fusion group. The above-mentioned preset matching degree condition can be that the matching degree between the business customer group fusion group corresponding to the sub-business warning model and the detection request information is the largest matching degree in the above-mentioned various matching degree information. The above-mentioned sub-business warning model can be a model with detection request information as input and warning information as output. For example, the sub-business warning model can be a logistic regression model. The training data of the sub-business warning model in the above-mentioned various sub-business warning models can be trained with the warning information corresponding to each detection request information corresponding to the business customer group fusion group and each detection request information in each detection request information. Each of the above-mentioned sub-business early warning models may be trained by a batch training method.

[0088] Step 104: input the detection request information into the target sub-business warning model to obtain warning information corresponding to the detection request information.

[0089] In some embodiments, the execution entity may input the detection request information into the target sub-business warning model to obtain warning information corresponding to the detection request information.

[0090] Step 105 , in response to detecting that the warning information meets the preset warning condition, obtaining request processing detail information of each sub-business warning model in each sub-business warning model, and obtaining each request processing detail information.

[0091] In some embodiments, in response to detecting that the above-mentioned warning information meets the preset warning condition, the above-mentioned execution subject can obtain the request processing detail information of each sub-business warning model in the above-mentioned each sub-business warning model, and obtain each request processing detail information. Among them, the above-mentioned preset warning condition can represent the order of the above-mentioned warning information as a preset order. The order of the above-mentioned preset information can represent the order of the above-mentioned warning information in each warning information generated by each sub-business warning model. For example, the order of the warning information can be the tenth. The request processing detail information in the above-mentioned each request processing detail information may include request processing quantity information and request processing duration information. The above-mentioned request processing detail information can represent the details of the sub-business warning model processing the detection request information. The above-mentioned request processing quantity information can represent the number of detection request information processed by the sub-business warning model. The above-mentioned request processing duration information can represent the time consuming of the sub-business warning model processing the detection request information. In practice, the above-mentioned execution subject can obtain each request processing detail information from a database storing each request processing detail information.

[0092] Step 106: Determine whether there is a sub-business warning model that meets the preset model migration condition among the sub-business warning models according to the detailed information of each request processing.

[0093] In some embodiments, the execution entity may determine whether there is a sub-business warning model that meets the preset model migration condition in each of the sub-business warning models based on the request processing details information. The preset model migration condition may be that the request processing quantity information corresponding to the sub-business warning model is greater than the preset request processing quantity and the request processing duration corresponding to the sub-business warning model is greater than the preset request processing duration. Here, no specific limitation is made to the value of the preset request processing quantity. No specific limitation is made to the value of the preset request processing duration.

[0094] Step 107 : in response to detecting that there is a sub-business warning model that meets the preset model migration condition among the sub-business warning models, migration processing is performed on the sub-business warning model that meets the preset model migration condition.

[0095] In some embodiments, in response to detecting that there is a sub-business warning model that meets the preset model migration conditions among the above-mentioned sub-business warning models, the above-mentioned execution entity may perform migration processing on the sub-business warning model that meets the above-mentioned preset model migration conditions. In practice, the above-mentioned execution entity may send the model file corresponding to the sub-business warning model that meets the above-mentioned preset model migration conditions to the target server to store the model file in the above-mentioned target server. The above-mentioned target server may be a pre-set server for receiving the migrated sub-business warning model.

[0096] Optionally, after step 107, first, the execution subject may determine the warning level information corresponding to the warning information according to the warning information. In practice, the execution subject may determine the warning level information corresponding to the warning information from the preset level information corresponding to the warning information in the preset level information set. The preset level information in the preset level information set may represent the correspondence between the warning information and the warning level information.

[0097] Afterwards, the warning method corresponding to the above warning level information can be determined based on the above warning level information. In practice, the above execution entity can determine the preset warning method information corresponding to the above preset level information from the preset warning method information corresponding to the above preset level information in the preset warning method information set. The preset warning method information in the above preset warning method information set can characterize the correspondence between the warning level information and the preset warning method. For example, the preset warning method information can be a warning level 1 warning, and the preset warning method can be a telephone warning.

[0098] Secondly, the above-mentioned detection request information can be processed for early warning according to the above-mentioned early warning method.

[0099] Afterwards, the warning information, the warning level information and the warning method corresponding to the detection request information may be determined as the warning record information corresponding to the detection request information.

[0100] Then, the warning record information may be serialized to obtain the serialized warning record information. In practice, the execution subject may serialize the warning record information using a serialization algorithm to obtain the serialized warning record information. For example, the serialization algorithm may be a JSON serialization algorithm.

[0101] Finally, the serialized warning record information is stored in the relevant database.

[0102] The above technical solution, as an inventive point of an embodiment of the present disclosure, solves the technical problem: "Directly storing the warning record information in an associated database to record the warning information without preprocessing the warning record information. When a large amount of warning record information is stored, the stored warning record information occupies a large storage space." The factors that cause the stored warning record information to occupy a large storage space are often as follows: directly storing the warning record information in an associated database without preprocessing the warning record information. If the above factors are solved, the effect of reducing the storage space occupied by the stored warning record information can be achieved. In order to achieve this effect, when adapting the business model, the present disclosure first serializes the warning record information to be stored, and then stores the serialized warning record information in an associated database. Thereby reducing the storage space occupied by storing the warning record information.

[0103] The above-mentioned embodiments of the present disclosure have the following beneficial effects: through a model migration method of some embodiments of the present disclosure, the server load can be reduced, the server response time can be shortened, and the number of server downtimes can be reduced. The reason for the large server load, long server response time, and large number of server downtimes is that the trained business warning models are integrated into one server. When the number of detection request information input into the trained business warning models is large, the server load is large, resulting in a long server response time and a large number of server downtimes. Based on this, the model migration method of some embodiments of the present disclosure first generates each business customer group fusion group according to the obtained business user information. Thus, each sub-business user information group set can be obtained, which can be further used to train each business warning model. Afterwards, in response to the detection request information of the business warning information, according to the above-mentioned detection request information and the above-mentioned business customer group fusion group, the matching degree information between the above-mentioned detection request information and each business customer group fusion group in the above-mentioned business customer group fusion group is generated to obtain each matching degree information. Thus, the matching degree of the detection request information and each business customer group fusion group can be obtained. Secondly, according to the above-mentioned matching degree information, the sub-business warning model that meets the preset matching degree condition in each pre-trained sub-business warning model is determined as the target sub-business warning model corresponding to the above-mentioned detection request information, so as to adapt the sub-business warning model, wherein the sub-business warning model in each pre-trained sub-business warning model corresponds to the business customer group fusion group in the above-mentioned business customer group fusion group, and the sub-business warning model in each sub-business warning model is obtained by training the corresponding business customer group fusion group. Thus, an adapted sub-business warning model can be obtained. Afterwards, the above-mentioned detection request information is input into the above-mentioned target sub-business warning model to obtain the warning information corresponding to the above-mentioned detection request information. Thus, the warning information of the target user can be obtained. Secondly, in response to detecting that the above-mentioned warning information meets the preset alarm condition, the request processing detail information of each sub-business warning model in the above-mentioned sub-business warning model is obtained, and each request processing detail information is obtained, wherein the request processing detail information in the above-mentioned request processing detail information includes request processing quantity information and request processing duration information. Thus, the request processing detail information of each sub-business warning model can be obtained. Afterwards, according to the above-mentioned request processing details information, it is determined whether there is a sub-business warning model that meets the preset model migration conditions among the above-mentioned sub-business warning models. Thus, it can be determined whether there is a business warning model to be migrated. Finally, in response to detecting that there is a sub-business warning model that meets the preset model migration conditions among the above-mentioned sub-business warning models, migration processing is performed on the sub-business warning model that meets the preset model migration conditions. Thus, the sub-business warning model that meets the migration conditions can be migrated to reduce the load on the server.Also, when there is a business warning model that meets the above-mentioned preset migration conditions in the server, the business warning model that meets the above-mentioned preset migration conditions will be migrated, so that the model can be adaptively migrated in a high-concurrency scenario to reduce the server load, and one type of detection request information can be processed by a separate target server, thereby improving the response speed and reducing the number of server downtimes.

[0104] Further references Figure 2 As an implementation of the methods shown in the above figures, the present disclosure provides some embodiments of a model migration method. These device embodiments are similar to Figure 1 Corresponding to the method embodiments shown, the device can be specifically applied to various electronic devices.

[0105] like Figure 2As shown, the model migration device 200 of some embodiments includes: a first generating unit 201 , a second generating unit 202 , a first determining unit 203 , an input unit 204 , an acquiring unit 205 , a second determining unit 206 and a migration unit 207 . Among them, the first generating unit 201 is configured to generate each business customer group fusion group according to the acquired business user information; the second generating unit 202 is configured to generate the matching degree information between the above-mentioned detection request information and each business customer group fusion group in the above-mentioned business customer group fusion group in response to the detection request information of the detected business warning information, according to the above-mentioned detection request information and the above-mentioned business customer group fusion groups, to obtain each matching degree information; the first determining unit 203 is configured to determine the sub-business warning model that meets the preset matching degree condition in each sub-business warning model trained by the above-mentioned business customer group fusion group as the target sub-business warning model corresponding to the above-mentioned detection request information according to the above-mentioned matching degree information, so as to adapt the sub-business warning model, wherein the sub-business warning model in the above-mentioned pre-trained sub-business warning model corresponds to the business customer group fusion group in the above-mentioned business customer group fusion group, and the sub-business warning model in the above-mentioned sub-business warning model is determined by corresponding obtained by training the business customer group fusion group; the input unit 204 is configured to input the above-mentioned detection request information into the above-mentioned target sub-business warning model to obtain the warning information corresponding to the above-mentioned detection request information; the acquisition unit 205 is configured to obtain the request processing detail information of each sub-business warning model in the above-mentioned each sub-business warning model in response to detecting that the above-mentioned warning information meets the preset alarm condition, and obtain each request processing detail information, wherein the request processing detail information in the above-mentioned each request processing detail information includes request processing quantity information and request processing duration information; the second determination unit 206 is configured to determine whether there is a sub-business warning model that meets the preset model migration condition in the above-mentioned each sub-business warning model according to the above-mentioned each request processing detail information; the migration unit 207 is configured to migrate the sub-business warning model that meets the above-mentioned preset model migration condition in response to detecting that there is a sub-business warning model that meets the preset model migration condition in the above-mentioned each sub-business warning model.

[0106] It is understood that the units described in the device 200 are similar to those described in the reference Figure 1 Therefore, the operations, features and beneficial effects described above for the method are also applicable to the device 200 and the units included therein, and will not be described in detail here.

[0107] Reference below Figure 3 , which shows a schematic structural diagram of an electronic device 300 (eg, a computing device) suitable for implementing some embodiments of the present disclosure. Figure 3The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0108] like Figure 3 As shown, the electronic device 300 may include a processing device 301 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. In the RAM 303, various programs and data required for the operation of the electronic device 300 are also stored. The processing device 301, the ROM 302, and the RAM 303 are connected to each other via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0109] Typically, the following devices may be connected to the I / O interface 305: input devices 306 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; output devices 307 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 308 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 309. The communication devices 309 may allow the electronic device 300 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 3 The electronic device 300 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead. Figure 3 Each block shown in the figure may represent one device, or may represent multiple devices as required.

[0110] In particular, according to some embodiments of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, some embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In some such embodiments, the computer program can be downloaded and installed from the network through the communication device 309, or installed from the storage device 308, or installed from the ROM 302. When the computer program is executed by the processing device 301, the above-mentioned functions defined in the method of some embodiments of the present disclosure are executed.

[0111] It should be noted that the computer-readable medium recorded in some embodiments of the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In some embodiments of the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In some embodiments of the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0112] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0113] The above-mentioned computer-readable medium may be included in the above-mentioned electronic device; or it may exist independently without being installed in the electronic device. The above-mentioned computer-readable medium carries one or more programs. When the above-mentioned one or more programs are executed by the electronic device, the electronic device: generates each business customer group fusion group according to the acquired business user information; in response to the detection request information of the business warning information detected, generates the matching degree information between the above-mentioned detection request information and each business customer group fusion group in the above-mentioned business customer group fusion group according to the above-mentioned detection request information and the above-mentioned business customer group fusion group, and obtains each matching degree information; according to the above-mentioned matching degree information, determines the sub-business warning model that meets the preset matching degree condition in each sub-business warning model trained by the above-mentioned business customer group fusion group as the target sub-business warning model corresponding to the above-mentioned detection request information, so as to adapt the sub-business warning model, wherein the sub-business warning model in the above-mentioned pre-trained sub-business warning model corresponds to the business customer group fusion in the above-mentioned business customer group fusion group. The sub-business warning models in the above-mentioned sub-business warning models are obtained by training the corresponding business customer group fusion group; the above-mentioned detection request information is input into the above-mentioned target sub-business warning model to obtain the warning information corresponding to the above-mentioned detection request information; in response to detecting that the above-mentioned warning information meets the preset alarm conditions, the request processing detail information of each sub-business warning model in the above-mentioned sub-business warning models is obtained to obtain each request processing detail information, wherein the request processing detail information in the above-mentioned each request processing detail information includes request processing quantity information and request processing duration information; according to the above-mentioned each request processing detail information, it is determined whether there is a sub-business warning model that meets the preset model migration conditions in the above-mentioned each sub-business warning model; in response to detecting that there is a sub-business warning model that meets the preset model migration conditions in the above-mentioned each sub-business warning model, the sub-business warning model that meets the preset model migration conditions is migrated.

[0114] Computer program code for performing the operations of some embodiments of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0115] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0116] The units described in some embodiments of the present disclosure may be implemented by software or by hardware. The described units may also be arranged in a processor, for example, they may be described as: a first generation unit, a second generation unit, a first determination unit, an input unit, an acquisition unit, a second determination unit, and a migration unit. Among them, the names of these units do not constitute a limitation on the units themselves under certain circumstances. For example, the migration unit may also be described as "a unit that performs migration processing on the sub-business warning models that meet the preset model migration conditions in response to detecting that there are sub-business warning models that meet the preset model migration conditions in the above-mentioned sub-business warning models."

[0117] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0118] The above descriptions are only some preferred embodiments of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the above-mentioned inventive concept. For example, the above-mentioned features are replaced with the technical features with similar functions disclosed in the embodiments of the present disclosure (but not limited to) and the technical solutions formed.

Claims

1. A model migration method, comprising: Generate various business customer group integration groups based on the acquired business user information; In response to detecting the detection request information of the service warning information, generating, according to the detection request information and the respective service customer group fusion groups, matching degree information between the detection request information and each service customer group fusion group, to obtain respective matching degree information; According to the respective matching degree information, the sub-business warning model that meets the preset matching degree condition in the respective sub-business warning models trained by the respective business customer group fusion groups is determined as the target sub-business warning model corresponding to the detection request information, so as to perform adaptation processing on the sub-business warning model, wherein the sub-business warning model in the respective pre-trained sub-business warning models corresponds to the business customer group fusion group in the respective business customer group fusion groups, and the sub-business warning model in the respective sub-business warning models is obtained by training the corresponding business customer group fusion group; Inputting the detection request information into the target sub-business warning model to obtain warning information corresponding to the detection request information; In response to detecting that the warning information meets the preset warning condition, obtaining request processing detail information of each sub-business warning model in each sub-business warning model to obtain each request processing detail information, wherein the request processing detail information in each request processing detail information includes request processing quantity information and request processing duration information; Determining, according to the processing details of each request, whether there is a sub-business warning model that meets a preset model migration condition among the sub-business warning models; In response to detecting that there is a sub-business warning model that meets the preset model migration condition among the various sub-business warning models, migration processing is performed on the sub-business warning model that meets the preset model migration condition.

2. The method according to claim 1, wherein: The method further comprises: According to the warning information, determining warning level information corresponding to the warning information; According to the warning level information, determining a warning method corresponding to the warning level information; According to the warning method, performing warning processing on the detection request information; Determine the warning information, the warning level information, and the warning method corresponding to the detection request information as the warning record information corresponding to the detection request information; Serializing the warning record information to obtain the serialized warning record information; The serialized warning record information is stored in the associated database.

3. The method according to claim 1, wherein: The migrating process of the sub-business early warning model that meets the preset model migration condition includes: The model file corresponding to the sub-business early warning model that meets the preset model migration condition is sent to the target server to store the model file in the target server.

4. The method according to claim 1, wherein: The generating of each service customer group fusion group according to the acquired service user information includes: Classify the information of each business user to obtain each business user customer group; Generate, according to each business user information, each business user portrait information corresponding to each business user information, wherein the business user portrait information in each business user portrait information corresponds to the business user information in each business user information; According to the business user portrait information, the business user customer groups are divided and processed to obtain each sub-business user information group set corresponding to each business user customer group; According to each obtained sub-service user information group, each service customer group fusion group corresponding to each obtained sub-service user information group is determined.

5. The method according to claim 4, wherein: The classifying and processing the information of each business user to obtain each business user customer group includes: Performing enhancement processing on the respective business user information to obtain the enhanced business user information as the respective business user information; The information of each business user is classified and processed to obtain each business user customer group.

6. The method according to claim 4, wherein: The generating, according to the respective business user information, respective business user portrait information corresponding to the respective business user information comprises: For each piece of business user information, perform the following steps: Perform word segmentation processing on the business user information to obtain information of each word; Perform redundancy elimination processing on each word information obtained, and obtain each word information after redundancy elimination as each target word information; According to the information of each target word, a vocabulary relationship table corresponding to the information of each target word is generated; Inputting the vocabulary relationship table and the preset topic quantity information into a pre-trained topic information generation model to obtain each topic information corresponding to each target word information; Business user portrait information corresponding to the business user information is generated according to the respective subject information and the respective target word information.

7. The method according to claim 4, wherein: The process of dividing each business user group according to the business user portrait information to obtain each sub-business user information group set corresponding to each business user group includes: Performing feature extraction processing on each of the business user portrait information to obtain each of the portrait feature information, wherein the portrait feature information in each of the portrait feature information corresponds to the business user portrait information in the each of the business user portrait information; Performing standardization processing on each of the image feature information to obtain each of the image feature information after standardization processing as each of the target image feature information; For each of the business user customer groups, perform the following steps: For each target portrait feature information corresponding to the business user customer group, perform the following steps: Determine each target portrait feature information different from the target portrait feature information in each target portrait feature information corresponding to the business user group as a target portrait feature information set; Determine similarity information between the target portrait feature information and each target portrait feature information in the determined target portrait feature information set to obtain respective similarity information; According to the respective similarity information, each target portrait feature information satisfying a preset selection condition in the determined target portrait feature information set and the target portrait feature information are determined as a target portrait feature information group; Determine each business user information corresponding to the determined target portrait feature information group as a sub-business user information group; The determined sub-service user information groups are determined as a sub-service user information group set.

8. A model migration device, comprising: The first generating unit is configured to generate each service customer group fusion group according to the acquired information of each service user; The second generating unit is configured to generate, in response to the detection request information of the detected service warning information, matching degree information between the detection request information and each of the service customer group fusion groups according to the detection request information and the respective service customer group fusion groups, to obtain respective matching degree information; A first determination unit is configured to determine, according to the respective matching degree information, a sub-business warning model that satisfies a preset matching degree condition among the respective sub-business warning models trained by the respective business customer group fusion groups as a target sub-business warning model corresponding to the detection request information, so as to perform adaptation processing on the sub-business warning model, wherein the sub-business warning model in the respective pre-trained sub-business warning models corresponds to the business customer group fusion group in the respective business customer group fusion groups, and the sub-business warning model in the respective sub-business warning models is obtained by training the corresponding business customer group fusion group; An input unit, configured to input the detection request information into the target sub-service warning model to obtain warning information corresponding to the detection request information; an acquisition unit, configured to, in response to detecting that the warning information satisfies a preset warning condition, acquire request processing detail information of each sub-business warning model in each sub-business warning model, and obtain each request processing detail information, wherein the request processing detail information in each request processing detail information includes request processing quantity information and request processing duration information; A second determining unit is configured to determine whether there is a sub-business warning model that meets a preset model migration condition among the sub-business warning models according to the detailed information of each request processing; The migration unit is configured to, in response to detecting that there is a sub-business warning model that meets the preset model migration condition among the various sub-business warning models, perform migration processing on the sub-business warning model that meets the preset model migration condition.

9. An electronic device, comprising: one or more processors; a storage device having one or more programs stored thereon; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

10. A computer readable medium having a computer program stored thereon, wherein: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images