Link generation method and apparatus

Abstract

Embodiments of the present specification provide a link generation method and device, wherein the link generation method comprises calling a third-party platform, recording calling data generated by calling the third-party platform to a calling log, and preprocessing calling data of the calling log; performing feature labeling on the preprocessed calling data to obtain initial features of the calling data, and selecting target features from the initial features of the calling data based on a preset selection rule; and determining a calling link of the third-party platform based on the initial features and the target features.

Description

[0001] This application is a divisional application of application No. 202110688119.2, titled "Link generation method and device", filed on June 22, 2021. TECHNICAL FIELD

[0002] Embodiments of the present specification relate to the technical field of computer technology, and particularly relate to a link generation method. One or more embodiments of the present specification also relate to a link generation device, a computing device, and a computer-readable storage medium. BACKGROUND

[0003] In the prior art, in the process of interaction between an enterprise and an external cooperation agency, the other party provides one or more interfaces (APIs) to provide related services, such as bank card fast payment related debiting, lending, balance inquiry, and signing. The external cooperation agency generally adopts a mainstream distributed architecture to split different services or functions according to the dimensions of the system, such as a special authentication system for signing, an account system for accounting, a lending and repayment system for personal loan business, and a lending and debiting system for enterprise loan business. Different systems provide services to the outside through the same interface, but their stability, throughput, and maintenance period are different. Since the internal system design of the external cooperation agency is often a black box, the service calling party (enterprise) is not informed of the design method of the internal system of the external cooperation agency, so the monitoring and operation strategy adopted by the enterprise is generally based on the interface dimension. If a service of the external cooperation agency fails, the entire interface for interaction between the enterprise and the external cooperation agency needs to be stopped, so the enterprise cannot call other services of the external cooperation agency for interaction, and the flexibility is poor. SUMMARY

[0004] Therefore, embodiments of the present specification provide a link generation method in the technical field of computer technology, and particularly relate to a link generation method. One or more embodiments of the present specification also relate to a link generation device, a computing device, and a computer-readable storage medium to solve the technical defects in the prior art.

[0005] According to a first aspect of embodiments of the present specification, a link generation method is provided, including:

[0006] calling a third-party platform, and recording calling data generated by calling the third-party platform to a calling log;

[0007] preprocessing the calling data of the calling log;

[0008] performing feature labeling on the preprocessed calling data to obtain initial features of the calling data, and selecting target features from the initial features of the calling data based on a preset selection rule;

[0009] determine the calling link of the third-party platform based on the initial feature and the target feature.

[0010] According to a second aspect of the embodiments of the present specification, a link generation apparatus is provided, comprising:

[0011] a data processing module configured to call a third-party platform, and record calling data generated by calling the third-party platform to a calling log, and pre-process the calling data of the calling log;

[0012] a feature analysis module configured to perform feature labeling on the pre-processed calling data, obtain initial features of the calling data, select target features from the initial features of the calling data based on preset selection rules, and determine the calling link of the third-party platform based on the initial features and the target features.

[0013] According to a third aspect of the embodiments of the present specification, a computing device is provided, comprising:

[0014] a memory and a processor;

[0015] The memory is configured to store computer executable instructions, and the processor is configured to execute the computer executable instructions, which implement the steps of the above link generation method when executed by the processor.

[0016] According to a fourth aspect of the embodiments of the present specification, a computer readable storage medium is provided, which stores computer executable instructions, which implement the steps of the above link generation method when executed by the processor.

[0017] One embodiment of the present specification implements a link generation method and apparatus, wherein the link generation method comprises calling a third-party platform, recording calling data generated by calling the third-party platform to a calling log, and pre-processing the calling data of the calling log; performing feature labeling on the pre-processed calling data, obtaining initial features of the calling data, and selecting target features from the initial features of the calling data based on preset selection rules; and determining the calling link of the third-party platform based on the initial features and the target features. Specifically, the link generation method generates the calling link associated with the third-party platform by pre-processing, feature analysis, link rule analysis, and learning, etc. of the calling data generated when calling the third-party platform. When the third-party platform fails or abnormally, the calling link of the third-party platform can be analyzed to determine a certain fault or abnormal link, and only the service of the link is stopped or limited, etc., without affecting other calling links of the third-party platform to process the project, thereby realizing fine control of the calling link of the third-party platform, and having strong flexibility. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 is a flow chart of a link generation method provided by one embodiment of the present specification;

[0019] Figure 2 is a visual link view in a link generation method provided by one embodiment of the present specification;

[0020] Figure 3 is a fault processing schematic diagram of a visual link view in a link generation method provided by one embodiment of the present specification;

[0021] Figure 4 is a process flow chart of a link generation method provided by one embodiment of the present specification;

[0022] Figure 5 is a structural schematic diagram of a link generation device provided by one embodiment of the present specification;

[0023] Figure 6 is a structural block diagram of a computing device provided by one embodiment of the present specification. DETAILED DESCRIPTION

[0024] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present specification. However, the present specification can be practiced without the specific details, other than in the examples provided herein, and it is understood that the scope of the present specification is not limited to the details below.

[0025] The terminology used in one or more embodiments of the present specification is for the purpose of describing particular embodiments only and is not intended to be limiting of one or more embodiments of the present specification. As used in one or more embodiments of the present specification and the accompanying claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in one or more embodiments of the present specification, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0026] It will be understood that, although the terms first, second, etc. can be used herein to describe various information, these terms are not intended to denote a temporal or chronological order. Rather, these terms are used solely to distinguish one from another only. For example, without departing from the scope of one or more embodiments of the present specification, first can be termed second, and similarly, second can be termed first. The term "if' as used herein, can be interpreted as meaning "when" or "in response to determining" depending on the context.

[0027] First, the terms used in the description of one or more embodiments of the present specification are explained.

[0028] Stream computing: can be well analyzed in real time for large-scale streaming data in the process of continuous change, capture information that may be useful, and send the results to the next computing node.

[0029] Interface (API): API (Application Programming Interface) is a set of pre-defined interfaces (such as functions, HTTP interfaces), or refers to the convention for the connection of different components of a software system; used to provide a set of routines that application programs can access based on a certain software or hardware, without accessing the source code or understanding the details of the internal working mechanism.

[0030] Fast payment: fast payment refers to that when the user purchases goods, he does not need to open online banking, only needs to provide bank card number, name, mobile phone number and other information, the bank verifies the correctness of the mobile phone number, and the third-party payment sends the mobile dynamic password to the user's mobile phone. The user inputs the correct mobile dynamic password, and the payment is completed. If the user chooses to save the card information, the user only needs to input the payment password of the third-party payment or the payment password and the mobile dynamic password to complete the payment next time.

[0031] Dynamic construction: without pre-setting known parameters or variables, relevant parameters or variables are constructed through automatic rule discovery, machine learning, etc.

[0032] Link view: a visual graphical interface or command line that can intuitively reflect important nodes in the system link calling process, such as scene, traffic, busy, importance, etc.

[0033] Machine learning: machine learning is a multi-disciplinary subject, involving probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory, etc. It is a subject that studies how computers simulate or implement human learning behavior to obtain new knowledge or skills, and reorganize existing knowledge structure to continuously improve their performance.

[0034] Cold start: a start mode of the computer. That is, cut off the power of the computer, restart, once the cold start, the contents of the memory are lost, the hardware is detected again, and then the operating system is started.

[0035] K-means: K-means clustering algorithm is an iterative solution of clustering analysis algorithm, the steps are, pre-data into K groups, then randomly select K objects as the initial cluster center, then calculate the distance between each object and each seed cluster center, assign each object to the nearest cluster center.

[0036] MeanShift algorithm: MeanShift clustering algorithm is a non-parametric clustering algorithm based on density, the algorithm idea is to assume that different cluster data sets conform to different probability density distribution, find the fastest direction of the density increase of any sample point, the sample density high area corresponds to the maximum value of the distribution, these sample points will eventually converge to the local maximum density, and the points converging to the same local maximum are considered to be members of the same cluster.

[0037] DBSCAN: (Density-Based Spatial Clustering of Applications with Noise, density-based clustering method with noise) is a density-based spatial clustering algorithm. The algorithm divides the area with sufficient density into clusters and discovers clusters of arbitrary shape in a spatial database with noise. It defines a cluster as the maximum set of density-connected points.

[0038] Gaussian Mixture Models (GMMs): is to use Gaussian probability density function (normal distribution curve) to accurately quantify things, it is a model that decomposes things into several Gaussian probability density function (normal distribution curve) based models.

[0039] In the process of interacting with external cooperation agencies, external cooperation agencies will provide one or more interfaces (APIs) for related services, such as bank card fast payment related deduction, loan, balance inquiry, subscription and other services. External cooperation agencies generally also use mainstream distributed architecture to split different services or functions according to the dimensions of the system, such as a dedicated authentication system for subscription, an account system for accounting, a loan and repayment system for personal loan business, and a loan and deduction system for enterprise loan business. Different systems provide services through the same interface, but their stability, throughput, maintenance cycle, etc. are different. The monitoring and operation strategy adopted by the calling party (i.e. the enterprise calling the services of the external cooperation agency) is generally based on the interface dimension (coarse granularity), i.e. the loan and deduction dimension, rather than the personal loan, personal deduction, enterprise loan and enterprise deduction dimension.

[0040] The reason for such coarse-grained control is that the internal system design of the external cooperation organization is often a black box, which does not inform the calling party of its own internal system design, and the internal system architecture gradually changes due to the optimization and upgrading of the system of the external cooperation organization. Moreover, as the number of external systems accessed by the calling party increases, it is impossible to manually maintain the internal system design of various external cooperation organizations in real time, so the interface dimension can only be managed in a coarse-grained manner.

[0041] Of course, in order to solve the technical problems in the background art, one solution can be to white-box the view, that is, after detailed communication with the external cooperation organization and sufficient understanding of the application architecture design (internal system design) of the external cooperation organization, the link view corresponding to all services calling the external cooperation organization is manually constructed through parameters and configuration. However, this solution must communicate in detail with each external cooperation organization and obtain the application architecture design of the other party, which has a very high communication cost. Moreover, after the application architecture of the other party changes, the entire calling link needs to be manually adjusted, which has a high maintenance cost. In addition, when some external cooperation organizations do not cooperate or the information is incorrect, the white-box view will be missing or incorrect, which will cause greater problems, such as making incorrect judgments and issuing incorrect instructions.

[0042] Another solution is to perform link analysis through non-machine learning data statistics and construct corresponding link parameters or configurations, which can also achieve relatively simple functions. However, this solution must have some understanding of the data and meet fixed rules. If the data characteristics change, it is easy to cause data omission. Moreover, different statistical rules need to be set for the data of different external cooperation organizations, which has poor flexibility and scalability.

[0043] Therefore, the embodiments of the present specification provide a link generation method, which provides a solution that can dynamically construct a black box (internal system design) link of an external cooperation organization. Based on a dynamic strategy, the internal system architecture of the external cooperation organization can be automatically identified in the case that the internal system of the external cooperation organization is a black box and cannot be seen. Through machine learning and other methods, a set of calling links calling the external cooperation organization is dynamically constructed, which provides fine guidance and help for production emergency, link stress testing, capacity evaluation and other operation and maintenance methods, and improves flexibility. For example, when a service of the external cooperation organization fails, only the calling link corresponding to the failed service can be suspended, instead of shutting down the entire interface connected with the external cooperation organization, or through multi-project staggered calling, mutual interference is reduced, the calling of the service of the external cooperation organization is realized, and the faulty service of the external cooperation organization is avoided, thereby improving user experience.

[0044] In the present specification, a link generation method is provided. One or more embodiments of the present specification also relate to a link generation apparatus, a computing device, and a computer readable storage medium, which are described in detail in the following embodiments.

[0045] Referring to Figure 1 , Figure 1 A flow chart of a link generation method according to an embodiment of the present specification is shown, which specifically includes the following steps.

[0046] Step 102: calling a third-party platform, and recording calling data generated by calling the third-party platform to a calling log.

[0047] In the present specification, the third-party platform can be understood as the external cooperation agency described above, such as a bank providing payment-related services such as deduction, loan, balance inquiry, and a shopping platform providing shopping, order inquiry, and the like. In actual application, the application platform of the link generation method is different, and the third-party platform is also different under the condition that the demand service of the application platform is different.

[0048] For example, if the link generation method of the present specification is applied to a certain payment platform, the payment platform needs to call the bank cooperating with it to provide services such as deduction, balance inquiry, etc., and the third-party platform can be understood as the bank cooperating with it. If the link generation method of the present specification is applied to a certain credit platform, the credit platform needs to call the shopping platform cooperating with it to provide the user's shopping record, and through the user's shopping record, the credit platform provides the user with credit scoring services, etc., and the third-party platform can be understood as the shopping platform cooperating with it.

[0049] Specifically, calling the third-party platform can be understood as calling a certain service provided by the third-party platform through the interaction interface of the third-party platform to process the project, such as the payment platform calling the deduction service of the bank through the calling interface provided by the bank to realize the payment operation of a certain item. And record the calling data generated by calling the third-party platform to the calling log.

[0050] In a specific implementation, the call data includes call time, success / failure, time consumption, third-party number, interface name, business feature, and / or return code, etc. The call time can be understood as the time of calling the third-party platform this time. The success / failure can be understood as successfully calling the service provided by the third-party platform or failing to call the service provided by the third-party platform. The time consumption can be understood as the duration of calling the service provided by the third-party platform this time. The third-party number can be understood as the unique identifier of the third-party platform, such as a string generated by letters, numbers, and / or special characters. The interface name can be understood as the interface name of the service provided by the third-party platform. The business feature can be understood as a dynamic value, which can be continuously refined according to different projects, such as loan, personal / enterprise loan, personal small loan / personal large loan, personal small scenario consumption loan / personal large loan, etc. The return code can be understood as the return value of calling the third-party platform this time. Each return value represents a meaning, such as return code 00 indicating successful call, return code 55 indicating password error, return code 99 indicating system exception, etc.

[0051] Step 104: preprocessing the call data of the call log.

[0052] Specifically, after recording the call data generated by calling the third-party platform to the call log, the method further includes:

[0053] uploading the call log to a log system;

[0054] Correspondingly, the preprocessing the call data of the call log includes:

[0055] streaming the call log in the log system back to a stream computing engine, and preprocessing the call data of the call log according to the stream computing engine.

[0056] If the link generation method of the embodiments of the present specification is applied to an application platform, the log system and the stream computing engine both belong to the processing system of the application platform.

[0057] In actual applications, after the calling data generated by calling the third-party platform is recorded in the calling log, in order to ensure the secure storage of the calling log, the calling log is uploaded to the log system for permanent storage in the log system. Moreover, if the third-party platform is frequently called, the calling data is frequently generated. If the calling data is preprocessed in real time, the data processing resources are greatly increased. When the calling amount of the third-party platform increases sharply, the calling data is gathered in the preprocessing stage, which may cause system crash. Therefore, the calling log recording the calling data is uploaded to the log system. In the subsequent preprocessing stage, the calling data in the calling log can be obtained from the log system according to a preset time interval for preprocessing, or the log system automatically issues the calling data in the calling log according to a preset time interval for preprocessing in the next stage. The preset time interval can be set according to actual applications, which is not limited herein. For example, the preset time interval can be set to 20 seconds or 1 minute.

[0058] In specific implementation, after the calling log is uploaded to the log system, the calling log in the log system is backflowed to the stream computing engine, and the calling data in the calling log is preprocessed according to the stream computing engine.

[0059] In the embodiments of the present specification, after the calling data generated by calling the third-party platform is recorded in the calling log, the calling log is uploaded to the log system to realize the secure and permanent storage of the calling data and avoid the loss of the calling data. Moreover, the calling log in the log system is backflowed to the stream computing engine, and the stream computing engine can well preprocess the large-scale calling data in real time and accurately.

[0060] In addition, the preprocessing of the calling data in the calling log according to the stream computing engine comprises:

[0061] According to the stream computing engine, the calling data in the calling log is summarized according to a preset time dimension, and the jitter data and noise data in the calling data in the calling log are deleted.

[0062] The preset time dimension can be set according to actual applications. For example, the preset time dimension is set to 1 minute, 2 minutes, etc.

[0063] Taking 1 minute as an example of the preset time dimension, the calling data of the calling log is summarized according to the preset time dimension by the stream computing engine, and the jitter data and noise data in the calling data of the calling log are deleted; it can be understood that the calling data of the calling log is summarized according to the dimension of 1 minute by the stream computing engine, and the jitter data and noise data in the summarized calling data of the calling log are deleted. That is, the stream computing engine summarizes the calling data in the calling log according to the dimension of 1 minute, and then deletes the jitter data and noise data in each 1 minute of the summarized calling data, so as to ensure the accuracy of subsequent feature analysis based on the calling data.

[0064] In actual application, when the link of the third-party platform has communication exception, periodic maintenance and upgrading, data jitter is easy to occur, and the calling data generated at this time needs to be denoised and removed, otherwise it will affect normal analysis. In addition, unknown abnormal data can also be removed in advance, and then the removed abnormal calling data is analyzed separately to ensure the integrity of the data.

[0065] Step 106: performing feature labeling on the preprocessed calling data to obtain initial features of the calling data, and selecting target features from the initial features of the calling data based on a preset selection rule.

[0066] Specifically, as a cold start strategy, artificial data labeling or other labeling methods can be used to label the features of the calling data, for example, some obvious features are labeled for the data: large / small, personal / enterprise, direct / indirect, different payment institutions, etc. These features are most likely to correspond to different application architectures (calling links). If subsequent clustering analysis cannot be distinguished, it means that the feature labeling of the calling data is problematic or the external cooperation agency uses the same application architecture. The specific adjustment can be made according to the actual situation.

[0067] For example, if a certain piece of preprocessed calling data is: calling time: 10 o'clock in the morning, calling success, time consumption: 3 seconds, third-party number: 1***, interface name: ***, business feature: loan, return code: 11, then the feature labeling of the calling data can be: small, personal, direct, payment institution 1***, etc.

[0068] In actual application, the preprocessed each call data is marked with features to obtain initial features of each call data; then target features are selected from the initial features of the call data based on preset selection rules. The preset selection rules can be set according to actual application, for example, the preset selection rules are to select features of project burst peak / valley, project anomaly or special time period. For example, when the business peak, the personal loan project appears time consumption increase, success rate decline, but the enterprise loan project fluctuates little; or during 12:00 to 1:00, the system will probably perform batch processing, and the personal and enterprise projects appear different available rate fluctuations. These call data can all be used as important differentiated indicators for feature differentiation. That is, which of the above contents (such as time consumption, success rate, etc.) in the call data are selected as target features from the initial features.

[0069] Step 108: determining the call link of the third-party platform based on the initial features and the target features.

[0070] Specifically, the determination of the call link of the third-party platform based on the initial features and the target features comprises:

[0071] inputting the initial features and the target features into a link generation model to obtain the call link of the third-party platform; or

[0072] classifying the initial features and the target features by using a preset clustering algorithm, and generating the call link of the third-party platform based on the classification result.

[0073] The link generation model is a pre-trained machine learning model, and specifically, the link generation model is trained by the following steps:

[0074] calling a third-party sample platform, and recording call data generated by calling the third-party sample platform into a call log;

[0075] preprocessing the call data of the call log;

[0076] performing feature labeling on the preprocessed call data to obtain initial features of the call data, and selecting target features from the initial features of the call data based on preset selection rules;

[0077] classifying the initial features and the target features by using a preset clustering algorithm, and generating the call link of the third-party sample platform based on the classification result;

[0078] using the initial features and the target features as training samples, and using the call link corresponding to the initial features and the target features as training labels;

[0079] The link generation model is trained based on the training sample and the training label, and the link generation model is obtained.

[0080] In actual application, the call data generated by calling the third-party platform is preprocessed, feature extraction analysis and rule learning are performed, and corresponding call links are generated. Then, based on the features of the call data and the call links, a training sample is formed, and a link generation model is pre-trained. In subsequent use, the initial features and target features of the call data can be directly input into the link generation model to quickly and accurately obtain the call links of the third-party platform.

[0081] Before the link generation model is pre-trained, the initial features and target features can be classified by using a preset clustering algorithm. Based on the analysis of the classification results, the call links of the third-party platform are generated. The clustering algorithm includes but is not limited to K-means, MeanShift, DBSCAN, Gaussian Mixture Model (GMMs) and the like. By using any one of the above clustering algorithms to perform general classification on the initial features and target features, an application architecture link (call link) is determined for each classification in the subsequent. If the dispersion degree of a certain classification data is high, the next layer of clustering analysis can be continued to try to mine the child nodes. For example, the features such as loan, deduction, balance inquiry are classified as a class, the features such as personal / enterprise loan, personal / enterprise deduction are classified as a class, the features such as personal / enterprise loan, large / small loan are classified as a class, and the features such as personal / enterprise loan, large / small loan, real-time / asynchronous loan are classified as a class.

[0082] In specific implementation, the data is labeled (i.e., the data is feature-labeled), which is divided into hierarchical labels and non-hierarchical labels. The non-hierarchical labels are, for example, transaction type, direct connection, time period [hour / minute / second], time consumption interval [100 / 200 / X milliseconds]; and the hierarchical labels are, for example, 1st level-loan, 2nd level-small loan, 3rd level-personal small loan, and 4th level-personal small consumption scenario loan.

[0083] In specific implementation, the data is labeled (i.e., the data is feature-labeled), which is divided into hierarchical labels and non-hierarchical labels. The non-hierarchical labels are, for example, transaction type, direct connection, time period [hour / minute / second], time consumption interval [100 / 200 / X milliseconds]; and the hierarchical labels are, for example, 1st level-loan, 2nd level-small loan, 3rd level-personal small loan, and 4th level-personal small consumption scenario loan.

[0083] In specific implementation, the data is labeled (i.e., the data is feature-labeled), which is divided into hierarchical labels and non-hierarchical labels. The non-hierarchical labels are, for example, transaction type, direct connection, time period [hour / minute / second], time consumption interval [100 / 200 / X milliseconds]; and the hierarchical labels are, for example, 1st level-loan, 2nd level-small loan, 3rd level-personal small loan, and 4th level-personal small consumption scenario loan.

[0083] In specific implementation, the data is labeled (i.e., the data is feature-labeled), which is divided into hierarchical labels and non-hierarchical labels. The non-hierarchical labels are, for example, transaction type, direct connection, time period [hour / minute / second], time consumption interval [100 / 200 / X milliseconds]; and the hierarchical labels are, for example, 1st level-loan, 2nd level-small loan, 3rd level-personal small loan, and 4th level-personal small consumption scenario loan.

[0083] In specific implementation, the data is labeled (i.e., the data is feature-labeled), which is divided into hierarchical labels and non-hierarchical labels. The non-hierarchical labels are, for example, transaction type, direct connection, time period [hour / minute / second], time consumption interval [100 / 200 / X milliseconds]; and the hierarchical labels are, for example, 1st level-loan, 2nd level-small loan, 3rd level-personal small loan, and 4th level-personal small consumption scenario loan.

[0084] Specifically, the generating the calling link of the third-party platform based on the classification result comprises:

[0085] determining initial features and target features of each category based on the classification result, and generating the calling link of the third-party platform according to the initial features and the target features of each category.

[0086] In actual application, after the initial features and the target features are classified, the initial features and the target features of each category are determined, and a corresponding calling link of the third-party platform is generated for each category of initial features and target features.

[0087] In the embodiments of the present specification, a corresponding calling link of the third-party platform is generated for each category of initial features and target features after classification, so as to generate a complete calling link of the third-party platform based on the above method.

[0088] In addition, after the generating the calling link of the third-party platform according to the initial features and the target features of each category, the method further comprises:

[0089] taking the initial features and the target features of each category as training samples, and taking the corresponding calling link generated according to the initial features and the target features of each category as a training label;

[0090] training the link generation model based on the training samples and the training label, and updating the link generation model.

[0091] Specifically, after the generating the calling link of the third-party platform according to the initial features and the target features of each category, the link generation model can be updated based on the initial features and the target features of each category and the corresponding calling link, so that the link generation model can learn in real time based on its new data, greatly improving the accuracy of subsequent link generation model applications.

[0092] In actual application, after the updating the link generation model, the method further comprises:

[0093] calling a target third-party platform, obtaining calling data of the target third-party platform, and determining data features of the calling data of the target third-party platform;

[0094] inputting the data features into the link generation model to obtain a calling link of the target third-party platform.

[0095] The target third-party platform can be understood as an external cooperative organization different from the third-party platform.

[0096] In actual implementation, after the link generation model is updated, when other third-party platforms are called again, based on the data characteristics of the calling data generated when other third-party platforms are called, the calling link of other third-party platforms can be obtained more quickly and accurately through the updated link generation model.

[0097] In another embodiment of the present specification, the calling link of the third-party platform is generated according to each type of initial feature and target feature, including:

[0098] The calling link of the third-party platform is generated according to each type of initial feature and target feature according to a preset feature level.

[0099] The preset feature level can be set according to actual needs, and the present specification does not make any limitation on this. For example, if the preset feature level is 1 layer, the calling link of the third-party platform generated according to each type of initial feature and target feature according to the preset feature level can be: calling party-third-party platform; if the preset feature level is 2 layers, the calling link of the third-party platform can be: calling party-third-party platform-system 1, and if the preset feature level is 3 layers, the calling link of the third-party platform can be: calling party-third-party platform-system 1-system 11, and so on.

[0100] In the above example, if the classified initial features and target features are: loan, deduction, balance inquiry; personal / enterprise loan, personal / enterprise deduction; personal / enterprise loan, large / small loan; personal / enterprise loan, large / small loan, real-time / asynchronous loan, the first layer is: loan, deduction, balance inquiry; the second layer is: personal / enterprise loan, personal / enterprise deduction; the third layer is: personal / enterprise loan, large / small loan; and the fourth layer is: personal / enterprise loan, large / small loan, real-time / asynchronous loan; then the calling link of the third-party platform generated according to each type of initial feature and target feature according to the four feature levels can be:

[0101] Calling party (for example, ** technology);

[0102] First layer: third-party platform (for example, institution 1) that can realize loan, deduction, and balance inquiry;

[0103] Second layer: system in the third-party platform that can realize personal / enterprise loan and personal / enterprise deduction (for example, system 1 in institution 1 that can provide a certain service);

[0104] Third layer: system in the third-party platform that can realize personal / enterprise loan and large / small loan (for example, system 11 in institution 1 that can provide a certain service);

[0105] The fourth layer: third-party platforms that can provide loans to individuals / enterprises, large / small amounts, and real-time / asynchronous loans (for example, system 111 in institution 1 that can provide a certain service).

[0106] In the embodiments of this specification, when generating the call chain of a third-party platform based on each type of initial feature and target feature, a personalized call chain can be generated based on a preset feature hierarchy to meet user experience requirements.

[0107] In another embodiment of this specification, after determining the call chain of the third-party platform, the method further includes:

[0108] A visual link view of a preset shape is generated based on the call link and the relationships between nodes in the call link.

[0109] See Figure 2 , Figure 2 A visual link view of a link generation method provided in one embodiment of this specification is shown.

[0110] Figure 2 The image displayed is a visual link view generated based on the call chain and the relationships between nodes in the call chain. In this view, **Technology, Institutional Interface Name A, System A, System B, System C, System A1, System A2, System C1, System C2, and System C11 are all nodes in the call chain. The connection between each node and other nodes represents the relationship between the two nodes.

[0111] Specifically, **Technology** comprises multiple systems, such as System A, System ..., etc. In practical applications, when **Technology** needs to complete a project, System A within **Technology** will call internal systems ..., which in turn will call services provided by their internal systems through interface A of an external partner organization to complete the project. Systems A, B, C, A1, A2, C1, C2, and C11 can all provide different services; therefore, using a specific service requires navigating different call chains.

[0112] Figure 2In the specific implementation, the **technology-agency interface name A is one calling link, the **technology-agency interface name A-system A is one calling link, the **technology-agency interface name A-system A-system A1 is one calling link, the **technology-agency interface name A-system A-system A2 is one calling link, the **technology-agency interface name A-system B is one calling link, the **technology-agency interface name A-system C is one calling link, the **technology-agency interface name A-system C-system C1 is one calling link, the **technology-agency interface name A-system C-system C2 is one calling link, and the **technology-agency interface name A-system C-system C1-system C11 is one calling link.

[0113] In the embodiments of the present specification, a visual link view can be generated based on the calling link, and the user can intuitively understand the service calling process based on the visual link view, which provides better help for production emergency, link stress testing, capacity assessment, and other operation and maintenance adjustments. In addition, in addition to displaying the visual view of the calling link, the calling link importance, transaction volume, stability, new / old, and the like can be labeled to provide more data display and assist in making better decisions.

[0114] In another embodiment of the present specification, after the calling link of the target third-party platform is obtained, the method further includes:

[0115] In the case that the calling of the target third-party platform fails, the calling link of the target third-party platform is determined;

[0116] The calling link of the target third-party platform is fault detected, a fault calling link is obtained, and fault processing is performed on the fault calling link.

[0117] Specifically, in the case that the calling of the target third-party platform fails, the calling link of the target third-party platform can be determined, and then each branch calling link in the calling link of the target third-party platform is detected to determine a fault calling link, and fault processing is performed on the fault calling link.

[0118] Referring to Figure 3 , Figure 3 A fault processing schematic diagram of a visual link view in a link generation method provided by one embodiment of the present specification is shown.

[0119] For example Figure 3 In the specific implementation, the **technology-agency interface name A is one calling link, the **technology-agency interface name A-system A is one calling link, the **technology-agency interface name A-system A-system A1 is one calling link, the **technology-agency interface name A-system A-system A2 is one calling link, the **technology-agency interface name A-system B is one calling link, the **technology-agency interface name A-system C is one calling link, the **technology-agency interface name A-system C-system C1 is one calling link, the **technology-agency interface name A-system C-system C2 is one calling link, and the **technology-agency interface name A-system C-system C1-system C11 is one calling link. Figure 3If the call of **Technology-Organization Interface Name A-System C-System C1-System C11 in the system fails, the system **Technology-Organization Interface Name A-System C-System C1-System C11 link can be shut down, and the service of the system C11 is started again after the link of **Technology-Organization Interface Name A-System C-System C1-System C11 is gradually recovered.

[0120] In specific implementation, when the call link of the target third-party platform fails, all links in the call link are detected for failure, and then the failure is disposed of through a failure disposal strategy (shutting down the link or reducing the traffic), and the service of the failed link is started again after the link is gradually recovered.

[0121] In the embodiments of the present specification, through such a fine dynamic processing strategy, the failure of a link corresponding to a certain project can be found and accurately processed, avoiding shutting down the entire service of the interface of the third-party platform. Moreover, the link generation method in the embodiments of the present specification improves the speed of the entire feature calculation, machine learning, and link view construction through real-time calculation. For most stable projects, TH or T1 offline data can be used, which consumes less resources.

[0122] The link generation method provided by the embodiments of the present specification generates the call link associated with the third-party platform through preprocessing, feature analysis, link rule analysis, and learning of the call data generated when the third-party platform is called. Subsequently, when the third-party platform fails, the call link of the third-party platform can be analyzed to determine a certain failure link, and only the service of the failure link is shut down or throttled, without affecting other call links of the third-party platform to process the project, thereby achieving fine control of the call link of the third-party platform and high flexibility. Moreover, through real-time calculation, the timeliness and universality of the entire feature calculation, machine learning, and link view construction are improved, and real-time disposal of link failure is also possible, thereby improving the timeliness of the entire emergency response.

[0123] In addition, the non-parametric clustering technology of machine learning does not require prior knowledge of the number of clusters and does not limit the shape of the cluster, avoiding the communication and maintenance costs of adjustment after the change of the application architecture of the other party, and regardless of whether the other party cooperates or the information is incorrect, the correct identification can be achieved, with high reliability and availability. Furthermore, the non-parametric clustering technology of machine learning avoids the need for prior knowledge of data and special attention to feature changes through statistical analysis, and adopts a universal solution for different organizations, with good flexibility and scalability.

[0124] Referring to Figure 4 , Figure 4A process flow diagram of a link generation method provided by one embodiment of the present specification is shown, and specifically includes the following steps.

[0125] Step 402: External link call starts.

[0126] Specifically, the external link call starts, which can be understood as calling the service link of the external cooperation agency through the interface of the external cooperation agency.

[0127] Step 404: Record the call log and report to the log system.

[0128] Among them, the call log includes call data, call time, success / failure, time consumption, agency number, API name, business characteristics 1 / 2 / 3, return code, etc., wherein the business characteristics 1...n are dynamic values that can be continuously refined according to different businesses, such as loans, personal / enterprise loans, personal small loans / personal large loans, personal small loan / loan for personal large loan, etc.

[0129] Step 406: Call log backflow to stream computing engine.

[0130] Specifically, after the stream computing engine aggregates the call data according to the minute dimension, it starts to preprocess the data, eliminating jitter or noise data, etc. Then the preprocessed call data is stored to the online analysis system, providing real-time data support for subsequent feature analysis and learning.

[0131] Step 408: Perform feature analysis.

[0132] Specifically, the online analysis system performs feature labeling on the call data, and selects important features from the initial features after feature labeling. Among them, the important features, when the system is running normally, the feature indicators converge, and the feature performance in the business peak / valley, business anomaly or special period is observed, such as when the business peak, the personal loan business appears time consumption increase, success rate decline, but the enterprise loan business fluctuation is very small; Or during 12:00 to 1:00, the system will probably perform batch processing, and the personal and enterprise businesses appear different availability rate fluctuations. These data can all be used as important differentiated indicators to realize important feature differentiation.

[0133] Step 410: Perform link rule analysis and learning.

[0134] Specifically, the online analysis system performs generalization classification on the features analyzed in step 408 by clustering analysis, such as using K-means, MeanShift, DBSCAN, Gaussian mixture model (GMMs) and other clustering algorithms, and then generates a new application architecture link (i.e. call link) for each classification.

[0135] Step 412: the result analysis is stored in the online analysis system.

[0136] Specifically, the analysis result model is trained based on the application architecture link and the analyzed features, and the trained analysis result model is stored in the online analysis system.

[0137] Step 414: a visual view is constructed.

[0138] Specifically, the visual view is constructed based on the application architecture link.

[0139] In actual application, when the external link is called again, the visual link view can be enriched or modified based on the calling data in the new calling log, and the above method is repeated to construct a more accurate and complete visual link view.

[0140] Step 416: dynamic disposal.

[0141] Specifically, in the case of failure to call the external link, the failure in the application architecture link is detected, and the failure link is closed or the traffic is reduced, and after the link is gradually restored, the service is started.

[0142] In actual application, the above scheme is completed through four modules, specifically including: a log record backflow calculation engine module (module one), a feature analysis and link rule learning module (module two), a visual view construction module (module three), and a dynamic disposal module (module four).

[0143] The log record backflow calculation engine module executes steps 402-406, mainly solves the link calling log structure, real-time reporting to the log system, and backflow to the stream calculation engine, and provides real-time data support for subsequent feature analysis and learning. Among them, the calling log structure can be understood as a general log printing format "2020-**-26 14:42:51aabbccddee calls a mechanism b service, result SUCCESS, time consumption 300ms, transaction scenario is personal small loan, code = ABCDE", this format is a natural language, which needs to be cut, enumerated and translated, etc. converted into K / V structure, for example: [instId = a, api = b, result = S, rt = 300ms, transCode = ABCDE], only such structured data can be processed by machine.

[0144] The feature analysis and link rule learning module performs steps 408-412, mainly solving the problems of link call log structured data, feature extraction analysis and rule learning; necessary labeling and enumeration are performed on the features according to the institutional dimension, and the extracted features are classified by a clustering algorithm, and a new application architecture link is generated for each classification. The visualization view module performs step 414, responsible for the construction of the visualization link view, the data comes from the link feature model after clustering analysis of the feature analysis and link rule learning module, and the core parameters include: institution number, business name, clustering result, and finally the visualization link view is constructed. The dynamic processing module performs step 416, responsible for the dynamic processing of the link after the problem occurs, such as when a fault is found in only one link, then try to close or reduce the traffic of the link business, and after the link is gradually restored, the business is started again.

[0145] The technical innovations of the link generation method provided by the specification are four points, the first is structured log record backflow calculation, the second is feature analysis and link rule learning, the third is visualization view construction, and the fourth is dynamic processing of fault links. Specifically, the structured log record backflow calculation: the link call log is structured, reported to the log system in real time, and backflowed to the stream computing engine, providing real-time data support for subsequent feature analysis and learning. Feature analysis and link rule learning: solve the problems of feature extraction analysis and rule learning of structured call log data, denoise the data, label and enumerate the features according to the institutional dimension, and classify the labeled features by a clustering algorithm, and generate a new application architecture link for each classification. Visualization view construction: the data comes from the link feature model after clustering analysis, and the core parameters include: institution number, business name, clustering result. Dynamic processing of fault links: dynamic processing of links after problems occur, such as when a fault is found in only one link, then try to close or reduce the traffic of the link business, and after the link is gradually restored, the business is started again, realizing fine control.

[0146] Corresponding to the method embodiments described above, the specification also provides link generation device embodiments, Figure 5 A structural schematic diagram of a link generation device provided by an embodiment of the specification is shown. As shown in the figure, Figure 5 The device includes:

[0147] The data processing module 502 is configured to call a third-party platform and record the call data generated by calling the third-party platform to a call log, and pre-process the call data of the call log;

[0148] The feature analysis module 504 is configured to perform feature labeling on the preprocessed call data to obtain initial features of the call data, select target features from the initial features of the call data based on a preset selection rule, and determine the call link of the third-party platform based on the initial features and the target features.

[0149] Optionally, the feature analysis module 504 is further configured to:

[0150] input the initial features and the target features into a link generation model to obtain the call link of the third-party platform; or

[0151] classify the initial features and the target features by using a preset clustering algorithm, and generate the call link of the third-party platform based on a classification result.

[0152] Optionally, the feature analysis module 504 is further configured to:

[0153] determine each type of initial features and target features based on the classification result, and generate the call link of the third-party platform according to each type of initial features and target features.

[0154] Optionally, the apparatus further comprises:

[0155] a model updating module configured to:

[0156] use each type of initial features and target features as training samples, and use a corresponding call link generated according to each type of initial features and target features as a training label;

[0157] train the link generation model based on the training samples and the training label to update the link generation model.

[0158] Optionally, the apparatus further comprises:

[0159] a link obtaining module configured to:

[0160] call a target third-party platform, obtain call data of the target third-party platform, and determine data features of the call data of the target third-party platform;

[0161] input the data features into the link generation model to obtain a call link of the target third-party platform.

[0162] Optionally, the apparatus further comprises:

[0163] a view generation module configured to:

[0164] generate a visual link view of a preset shape based on the call link and an association relationship between nodes in the call link.

[0165] Optionally, the feature analysis module 504 is further configured to:

[0166] generate the call link of the third-party platform according to the preset feature hierarchy for each type of initial feature and target feature.

[0167] Optionally, the apparatus further comprises:

[0168] a log uploading module configured to:

[0169] upload the call log to a log system;

[0170] Correspondingly, the data processing module 502 is further configured to:

[0171] backflow the call log in the log system to a stream computing engine, and pre-process the call data of the call log according to the stream computing engine.

[0172] Optionally, the data processing module 502 is further configured to:

[0173] aggregate the call data of the call log according to a preset time dimension according to the stream computing engine, and delete jitter data and noise data in the call data of the call log.

[0174] Optionally, the apparatus further comprises:

[0175] a fault processing module configured to:

[0176] determine the call link of the target third-party platform in the case of failure or exception in calling the target third-party platform;

[0177] detect faults of the call link of the target third-party platform, obtain a fault call link, and perform fault processing on the fault call link.

[0178] Optionally, the call data comprises call time, success / failure, time consumption, third-party number, interface name, business feature, and / or return code.

[0179] The link generation apparatus provided by the embodiments of the present specification generates the call link associated with the third-party platform through pre-processing, feature analysis, link rule analysis, and learning of the call data generated when calling the third-party platform. Subsequently, when a fault occurs in calling the third-party platform, the call link of the third-party platform can be analyzed to determine a certain fault link, and only the service of the fault link is stopped or throttled, etc., without affecting other call links of the third-party platform to perform project processing, thereby realizing fine control of the call link of the third-party platform and having strong flexibility.

[0180] The above is a schematic scheme of the link generation apparatus of the embodiment. It should be noted that the technical scheme of the link generation apparatus and the technical scheme of the link generation method described above belong to the same concept, and the details of the technical scheme of the link generation apparatus which are not described in detail can be referred to the description of the technical scheme of the link generation method.

[0181] Figure 6 A structural block diagram of a computing device 600 according to one embodiment of the present specification is shown. The components of the computing device 600 include, but are not limited to, a memory 610 and a processor 620. The processor 620 is connected to the memory 610 through a bus 630, and a database 650 is used to save data.

[0182] The computing device 600 also includes an access device 640, which enables the computing device 600 to communicate via one or more networks 660. Examples of these networks include a public switched telephone network (PSTN), a local area network (LAN), a wide area network (WAN), a personal area network (PAN), or a combination of communication networks such as the Internet. The access device 640 can include one or more of any type of network interface (e.g., network interface card (NIC)), wired or wireless, such as an IEEE 802.11 wireless local area network (WLAN) wireless interface, a Worldwide Interoperability for Microwave Access (Wi-MAX) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, a near field communication (NFC) interface, and the like.

[0183] In one embodiment of the present specification, the above-mentioned components of the computing device 600 and other components not shown in the Figure 6 may be connected to each other, for example, through a bus. It should be understood that Figure 6 The structural block diagram of the computing device shown is only for the purpose of example, and is not a limitation on the scope of the present specification. Other components can be added or replaced as needed by those skilled in the art.

[0184] The computing device 600 can be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., a tablet computer, a personal digital assistant, a laptop computer, a notebook computer, a netbook, etc.), a mobile phone (e.g., a smartphone), a wearable computing device (e.g., a smartwatch, smart glasses, etc.), or other types of mobile devices, or a stationary computing device such as a desktop computer or a PC. The computing device 600 can also be a mobile or stationary server.

[0185] The processor 620 is configured to execute computer-executable instructions, which, when executed by the processor, implement the steps of the link generation method described above.

[0186] The above is a schematic scheme of the computing device of the embodiment. It should be noted that the technical scheme of the computing device and the technical scheme of the link generation method described above belong to the same concept, and the details of the technical scheme of the computing device that are not described in detail can be referred to the description of the technical scheme of the link generation method.

[0187] An embodiment of the present specification also provides a computer readable storage medium storing computer executable instructions, the computer executable instructions being executed by a processor to implement the steps of the link generation method described above.

[0188] The above is a schematic scheme of the computer readable storage medium of the embodiment. It should be noted that the technical scheme of the storage medium and the technical scheme of the link generation method described above belong to the same concept, and the details of the technical scheme of the storage medium that are not described in detail can be referred to the description of the technical scheme of the link generation method.

[0189] The above describes specific embodiments of the present specification. Other embodiments are within the scope of the appended claims. In some cases, the acts or steps recited in the claims can be performed in a different order than the order described in the embodiments and still achieve desirable results. In addition, the processes depicted in the figures do not necessarily require the particular order shown or sequential order in order to achieve the desired results. In some implementations, multitasking and parallel processing can be advantageous.

[0190] The computer instructions include computer program code, which can be in the form of source code, object code, executable code, or some intermediate form. The computer readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction, for example, in some jurisdictions, according to legislation and patent practice, the computer readable medium does not include electrical carrier signals and telecommunication signals.

[0191] It should be noted that, for the aforementioned method embodiments, the sequences of the described actions are not necessarily required to implement the present application, and certain actions can be performed in other sequences, or even at the same time, in accordance with the present application. Furthermore, certain actions can not be required to implement the present application. Additionally, the described embodiments are not necessarily the only possible implementation of the present application.

[0192] In the above embodiments, the description of each embodiment focuses on different aspects, and the parts not described in detail in a certain embodiment can be referred to the relevant description of other embodiments.

[0193] The above disclosed preferred embodiments of the present application are only used to help explain the present application. Alternative embodiments do not describe all the details and limit the present application to the specific embodiments described. Obviously, according to the content of the embodiments of the present application, many modifications and changes can be made. The present application selects and specifically describes these embodiments in order to better explain the principles and practical applications of the embodiments of the present application, so that those skilled in the art can well understand and use the present application. The present application is limited by the claims and their full scope and equivalents.

Claims

1. A link generation method, comprising: calling a third-party platform, and recording calling data generated by calling the third-party platform to a calling log; preprocessing calling data of the calling log; performing feature labeling on the preprocessed calling data to obtain initial features of the calling data, and selecting target features from the initial features of the calling data based on a preset selection rule; classifying the initial features and the target features using a preset clustering algorithm, and generating a calling link of the third-party platform based on a classification result, including: determining each class of initial features and target features based on the classification result, and generating the calling link of the third-party platform according to a preset feature hierarchy for each class of initial features and target features. 2.The link generation method of claim 1, after the generating the calling link of the third-party platform based on the classification result, further comprising: generating a visual link view of a preset shape based on the calling link and an association relationship between nodes in the calling link. 3.The link generation method of claim 1, after the recording the calling data generated by calling the third-party platform to the calling log, further comprising: uploading the calling log to a log system; correspondingly, the preprocessing the calling data of the calling log, comprising: streaming the calling log in the log system back to a stream computing engine, and preprocessing the calling data of the calling log according to the stream computing engine. 4.The link generation method of claim 3, the preprocessing the calling data of the calling log according to the stream computing engine, comprising: summarizing the calling data of the calling log according to the stream computing engine in a preset time dimension, and deleting jitter data and noise data in the calling data of the calling log. 5.The link generation method of any one of claims 1-4, wherein the calling data comprises calling time, success / failure, time consumption, third-party number, interface name, business feature, and / or return code. 6.A link generation apparatus, comprising: a data processing module configured to call a third-party platform, and record calling data generated by calling the third-party platform to a calling log, and preprocess calling data of the calling log; a feature analysis module configured to perform feature labeling on the preprocessed calling data to obtain initial features of the calling data, select target features from the initial features of the calling data based on a preset selection rule, and determine a calling link of the third-party platform based on the initial features and the target features; the feature analysis module is further configured to determine each class of initial features and target features based on a classification result, and generate the calling link of the third-party platform according to a preset feature hierarchy for each class of initial features and target features. 7.A computing device, comprising: a memory and a processor; the memory is configured to store computer executable instructions, and the processor is configured to execute the computer executable instructions, which, when executed by the processor, implement the steps of the link generation method of any one of claims 1-5.

8. A computer readable storage medium storing computer executable instructions which, when executed by a processor, implement the steps of the link generation method of any one of claims 1 to 5.

Images