A service optimization method, device and equipment

Abstract

Embodiments of the present specification provide a service optimization method, device and equipment, wherein the method comprises: obtaining a target service node in which a service level is not at a safe water level in a recommendation system, performing parameter degradation processing on recall configuration parameters of a recall service node in the target service node based on a preset degradation gradient; performing parameter degradation processing on a truncation configuration parameter of an ordering service node in the target service node according to a current degradation round and a preset truncation step; performing object recall processing based on the parameter-degraded recall service node to obtain a first candidate object, and performing truncation screening processing on the first candidate object by using the parameter-degraded ordering service node to determine an object recommendation result according to a truncation screening processing result.

Description

Technical Field

[0001] This document relates to the field of computer technology, and in particular to a service optimization method, apparatus, and device. Background Technology

[0002] With the rapid development of computer technology and financial services, financial leasing services such as credit leasing have become an important part of the modern service industry. To improve the success rate of financial leasing services, services such as product recommendation and product search can be provided to users.

[0003] However, as the application of business scenarios such as product recommendation and product retrieval becomes more widespread, recommendation systems may face situations such as sudden traffic surges. When faced with sudden traffic surges (such as during major promotional events), recommendation systems often adopt a global on / off degradation strategy (such as shutting down all recall channels), which can lead to a precipitous drop in user service experience. Therefore, it is necessary to provide more reliable service optimization solutions to ensure the effectiveness of object recommendation under situations such as sudden traffic surges. Summary of the Invention

[0004] The purpose of the embodiments in this specification is to provide a more reliable service optimization scheme to ensure the effectiveness of object recommendation under conditions such as sudden traffic surges.

[0005] To achieve the above technical solution, the embodiments in this specification are implemented as follows: This specification provides a service optimization method, comprising: acquiring target service nodes in a recommendation system whose service level is not at a safe level; for recall service nodes among the target service nodes, performing parameter downgrading processing on the recall configuration parameters of the recall service nodes based on a preset downgrading gradient; for ranking service nodes among the target service nodes, performing parameter downgrading processing on the truncation configuration parameters of the ranking service nodes according to the current downgrading round and a preset truncation step size; performing object recall processing based on the parameter-downgraded recall service nodes to obtain a first candidate object, and using the parameter-downgraded ranking service nodes to perform truncation filtering processing on the first candidate object, so as to determine the object recommendation result based on the truncation filtering processing result.

[0006] This specification provides a service optimization device, comprising: a node acquisition module for acquiring target service nodes in a recommendation system whose service level is not at a safe level; a service degradation module for downgrading the recall configuration parameters of recall service nodes among the target service nodes based on a preset degradation gradient; a truncation update module for downgrading the truncation configuration parameters of ranking service nodes among the target service nodes based on the current degradation round and a preset truncation step size; and an object recommendation module for performing object recall processing based on the parameter-downgraded recall service nodes to obtain a first candidate object, and using the parameter-downgraded ranking service nodes to perform truncation filtering processing on the first candidate object, so as to determine the object recommendation result based on the truncation filtering processing result.

[0007] This specification provides a service optimization device, comprising: a processor; and a memory configured to store computer-executable instructions, wherein, when executed, the processor: acquires target service nodes in a recommendation system whose service level is not at a safe level; performs parameter downgrading processing on recall configuration parameters of recall service nodes among the target service nodes, based on a preset downgrading gradient; performs parameter downgrading processing on truncation configuration parameters of ranking service nodes among the target service nodes, based on the current downgrading round and a preset truncation step size; performs object recall processing based on the parameter-downgraded recall service nodes to obtain a first candidate object, and performs truncation filtering processing on the first candidate object using the parameter-downgraded ranking service nodes, so as to determine the object recommendation result based on the truncation filtering processing result.

[0008] This specification also provides a storage medium for storing computer-executable instructions. When executed by a processor, these instructions implement the following process: acquiring target service nodes in a recommendation system whose service level is not at a safe level; for recall service nodes among the target service nodes, performing parameter downgrading processing on the recall configuration parameters of the recall service nodes based on a preset downgrading gradient; for ranking service nodes among the target service nodes, performing parameter downgrading processing on the truncation configuration parameters of the ranking service nodes according to the current downgrading round and a preset truncation step size; performing object recall processing based on the parameter-downgraded recall service nodes to obtain a first candidate object, and using the parameter-downgraded ranking service nodes to perform truncation filtering processing on the first candidate object, thereby determining the object recommendation result based on the truncation filtering processing result.

[0009] This specification also provides a computer program product, including a computer program that, when executed by a processor, implements the following process: obtaining target service nodes in a recommendation system whose service level is not at a safe level; for recall service nodes among the target service nodes, performing parameter downgrading processing on the recall configuration parameters of the recall service nodes based on a preset downgrading gradient; for ranking service nodes among the target service nodes, performing parameter downgrading processing on the truncation configuration parameters of the ranking service nodes according to the current downgrading round and a preset truncation step size; performing object recall processing based on the parameter-downgraded recall service nodes to obtain a first candidate object, and using the parameter-downgraded ranking service nodes to perform truncation filtering processing on the first candidate object, so as to determine the object recommendation result based on the truncation filtering processing result. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Figure 1 This is a schematic diagram illustrating the implementation environment of one service optimization method described in this specification. Figure 2 This is a flowchart illustrating the processing procedure of a service optimization method described in this specification. Figure 3 This is a schematic diagram illustrating a process for obtaining water level information as described in this specification. Figure 4 This is a flowchart illustrating a parameter downgrade process for recalling configuration parameters as described in this specification. Figure 5 This is a flowchart illustrating a parameter downgrade process for truncating the object quantity parameter in this specification. Figure 6 This is a flowchart illustrating the process of determining the recommendation result for an object in this specification. Figure 7 This is a flowchart illustrating another parameter downgrade process for the truncated object quantity parameter in this specification. Figure 8 This is a schematic diagram illustrating the decision-making logic of a recommendation system as described in this specification. Figure 9 This is a flowchart illustrating a parameter recovery process described in this manual. Figure 10 This is a schematic diagram of a service optimization device described in this specification; Figure 11 This is a schematic diagram of a service optimization device described in this specification. Detailed Implementation

[0011] This specification provides a service optimization method, apparatus, and device through its embodiments.

[0012] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0013] This specification provides a more reliable service optimization scheme to ensure the effectiveness of object recommendation under conditions such as sudden traffic surges. With the rapid development of computer technology and financial services, credit leasing and other financial leasing businesses have become an important part of the modern service industry. As the application of business scenarios such as product recommendation and product retrieval becomes increasingly widespread, recommendation systems may face situations such as sudden traffic surges. For situations like sudden traffic surges, existing methods relying on manual configuration of rate limiting and degradation plans suffer from problems such as delayed response, incomplete coverage, and high maintenance costs, and cannot achieve second-level adaptive adjustment. When the request volume exceeds the system's capacity, without an effective protection mechanism, it can easily trigger a service avalanche, leading to a sharp drop in user experience or even service unavailability. Specifically, because traditional rate limiting and static degradation schemes cannot balance system stability and user experience, they often lead to a dilemma of "either all crash or tough it out." Furthermore, since most degradation methods rely on only a single indicator (such as queries per second (QPS, error rate, etc.) to determine whether the system is overloaded), they cannot accurately reflect the true resource bottleneck. For example, when the CPU does not reach the threshold but the response time (Response Time) is significantly reduced, the system may still experience a significant drop in user experience. When lock contention intensifies (Time to Release, RT), the system is practically unusable. However, solutions relying solely on a single metric for service optimization cannot trigger degradation due to the limited perspective. Furthermore, global on / off degradation (e.g., "disable all vector recall") lacks gradient control and priority ranking, leading to a precipitous drop in service experience. Moreover, it cannot automatically derive truncation parameters or degradation targets based on node type (e.g., recall service nodes, sorting service nodes), requiring manual maintenance of mapping relationships, which is error-prone and lacks scalability. Therefore, this specification provides a more reliable service optimization scheme to ensure object recommendation performance under conditions of sudden traffic surges. In this scheme, for recall service nodes not yet at a safe level, "gradual degradation" can be achieved through preset degradation gradients. To avoid a precipitous drop in user service experience caused by a global on / off degradation, and to ensure the core user experience, the system also addresses the issue of ranking service nodes not yet at a safe level. For these nodes, parameter degradation can be quickly and accurately applied based on the current degradation round and a preset truncation step size. This avoids the high error rate associated with manual parameter configuration, ensuring the maintainability of the recommendation system. Furthermore, by utilizing the recalled service nodes and the downgraded ranking service nodes, a closed-loop control mechanism of "accurate overload identification—intelligent decision-making degradation—smooth execution degradation" can be achieved. This solves the problem of uncontrollable degradation of recommendation services under high concurrency and ensures the effectiveness of object recommendation under sudden traffic surges. Specific details can be found in the following embodiments.

[0014] The service optimization methods described in one or more embodiments of this specification are applicable to the service optimization implementation environment, such as... Figure 1 As shown, the implementation environment includes at least: Client 100 and server 200. Furthermore, server 200 can be configured with various network models and algorithms, among which: Client 100 can run on terminal devices, which can be mobile phones, personal computers, tablets, e-book readers, wearable devices, devices that interact with information based on AR (Augmented Reality) and VR (Virtual Reality), and laptop computers, etc. Client 100 can be installed on terminal devices, and Client 100 can be an application, a browser, or a subroutine embedded in an application, etc.

[0015] Server 200 can run on a server, which can be one or more servers, a server cluster consisting of several servers, or a cloud server on a cloud computing platform. Server 200 can be installed on the server. Server 200 can be an application or a subroutine embedded in an application. Various network models and algorithms can be integrated into server 200, or server 200 can call one or more of various network models and algorithms to perform corresponding operations.

[0016] In addition, it may include a database 300, which may be located in the server on which the server 200 runs, or outside the server on which the server 200 runs. The database 300 may store graph structure data (i.e., graph data composed of recall service nodes, ranking service nodes, etc.) corresponding to the recommendation system, as well as the safety water level corresponding to each service node.

[0017] In this implementation environment, server 200 can obtain target service nodes in the recommendation system whose service level is not at a safe level. For recall service nodes among the target service nodes, based on a preset degradation gradient, the recall configuration parameters of the recall service nodes are downgraded. For ranking service nodes among the target service nodes, based on the current degradation round and a preset truncation step size, the truncation configuration parameters of the ranking service nodes are downgraded. Based on the parameter-downgraded recall service nodes, object recall processing is performed to obtain the first candidate object. Then, using the parameter-downgraded ranking service nodes, the first candidate object is truncated and filtered to determine the object recommendation result. Server 200 can return the object recommendation result to client 100.

[0018] like Figure 2As shown in the embodiments of this specification, a service optimization method is provided. The execution subject of this method can be a server, which can be a single independent server or a server cluster composed of multiple servers. The server can be a backend server for services such as credit leasing, commodity trading, and video viewing, or a backend server for a credit leasing application. Specifically, the method may include the following steps: In step S202, the target service nodes in the recommendation system whose service level is not at a safe level are obtained.

[0019] The recommendation system can be a system for recommending objects, such as a recommendation system for leasable goods, tradable goods, and watchable videos. The target service node can be a graph structure data where the service level is not at a safe level. The graph structure data can be graph data constructed based on recall service nodes, ranking service nodes, etc. For example, the graph structure data can be a directed acyclic graph (DAG), that is, graph structure data can be graph data composed of vertices and directed edges, and there are no directed cycles in the graph. The graph structure data can include multiple different types of nodes (such as recall service nodes, ranking service nodes, etc.), as well as node connection relationships determined based on the dependencies between services.

[0020] In implementation, the server can construct a graph-structured data based on the dependencies between services in the recommendation system. For example, the recommendation system may include a recall service, a ranking service, and a delivery service. The recall service can be used to initially filter out a set of items that may be relevant to the user from a massive candidate set. The ranking service can be used to refine the scoring and ranking of the recall results to improve the accuracy of the recommendations. This stage may include sub-stages such as coarse ranking, fine ranking, and re-ranking. Finally, the delivery service can generate object recommendation results from the objects filtered by the ranking service and return them to the user.

[0021] The server can construct nodes corresponding to each of the above services, and determine the node connection relationships between the nodes based on the dependencies between the multiple services. Based on the constructed nodes and the determined node connection relationships, the server can determine the graph structure data corresponding to the recommendation system.

[0022] Furthermore, the method for determining the nodes and node connection relationships in the above graph structure data is an optional and implementable method. In practical application scenarios, there can be a variety of different methods. For example, the recommendation system may also include a fusion truncation service between the recall service and the ranking service, a reordering service between the ranking service and the delivery service, etc. The server can select different methods according to the actual application requirements of the recommendation system. This specification does not specifically limit this method in the embodiments.

[0023] The server can perform water level detection processing on the graph structure data corresponding to the recommendation system to instantly filter out target service nodes whose service water level is not at a safe level.

[0024] The server can determine whether the service level of service nodes in the graph structure data is at a safe level based on preset detection indicators such as CPU utilization and response time (RT). For example, if the CPU utilization of the recalled service node is greater than 5% but not equal to 100% and the response time is greater than 5ms, it can be determined that the service level of the recalled service node is not at a safe level. In this way, by detecting both CPU utilization and response time at the same time, the misjudgment of a single indicator can be avoided, and the accuracy of overload identification can be improved through multi-dimensional water level fusion perception.

[0025] In addition, when collecting CPU utilization and response time, the server can start an independent thread for each service in the Distributed Resource Management (DRM) and periodically (e.g., every 3 seconds) call the MAAS function to obtain the service's CPU utilization, response time and other water level information from the cache. Based on the water level information of each service obtained, the server can filter out the target service nodes in the recommendation system whose service water level is not at a safe level.

[0026] Specifically, such as Figure 3 As shown, the server can asynchronously execute the water level information acquisition task using threads on a minute-by-minute basis. First, the server can obtain the identification information of the service nodes and save the service configuration information of each service node. Then, it can call the MAAS function to obtain the water level information. If the server determines that the service level of a service node is within a safe range based on the obtained water level information, it can update the cache (such as the Tbase cache, which does not expire) and record a detection log of the successful update of the water level information.

[0027] in, Figure 3The dynamic rule manager in the system can be used to configure, cache, and update rules related to flexible degradation. Flexible degradation refers to a technical mechanism that maintains core service capabilities and avoids cascading failures when the recommendation system faces high load or abnormal traffic by dynamically adjusting service strategies (such as parameter degradation processing for recall service nodes, parameter degradation processing for ranking service nodes, etc.). MAAS functions can be Model-as-a-Service functions, used to asynchronously obtain service runtime status data (such as CPU, memory, RT, etc.) from the recommendation system's configuration center as the water level information of that service node.

[0028] In this way, by combining local caching with asynchronous refresh using a small amount of traffic, both real-time performance and performance overhead can be balanced.

[0029] In step S204, for the recall service node in the target service node, the recall configuration parameters of the recall service node are downgraded based on the preset downgrade gradient.

[0030] The preset degradation gradient can be determined based on the experience requirements of the specific application scenario of the recommendation system. For example, for scenarios with high experience requirements, the corresponding preset degradation gradient can be smaller. That is, in object recommendation scenarios where user experience is the priority, user experience can be guaranteed by slow degradation. Conversely, for scenarios with low experience requirements, the corresponding preset degradation gradient can be larger. That is, in object recommendation scenarios where user experience is the priority, rapid degradation can be achieved to alleviate the pressure on computing resources. The recall configuration parameters can include parameters such as the number of recall tasks.

[0031] In implementation, the server can determine the node identifier of the recall service node in the graph structure data, and based on the node identifier, obtain the node's CPU utilization and response time, and then, based on the obtained information, determine whether the recall service node is the target service node, that is, determine whether the recall service node has exceeded its service limits.

[0032] If it is determined that the recall service node has exceeded its service limit, the server can generate a skip instruction based on a preset degradation gradient. The skip instruction can be used to downgrade the parameters of the recall service node, that is, the parameter value corresponding to the number of recall tasks can be updated through the skip instruction.

[0033] For example, if the preset degradation gradient is 20%, then if the recall service node currently has 100 pending recall tasks, it can skip 100 * 20% = 20 of these 100 recall tasks. This means the parameter value of the recall task quantity parameter can be downgraded from 100 to 80. In this way, the downgraded recall service node can process only 80 recall tasks, achieving parameter degradation by reducing the number of recall cycles. The server can select which recall tasks to skip based on factors such as the task volume and priority.

[0034] In step S206, for the sorting service node in the target service node, the truncation configuration parameters of the sorting service node are downgraded according to the current downgrade round and the preset truncation step size.

[0035] In implementation, the server can determine the node identifier of the sorting service node in the graph structure data, and based on the node identifier, obtain the node's CPU utilization and response time, etc., and then, based on the obtained information, determine whether the sorting service node is the target service node, that is, determine whether the sorting service node has exceeded its service limits. The truncation configuration parameters can include parameters such as the number of truncation objects.

[0036] If it is determined that the sorting service node has exceeded its service limits, the server can perform parameter degradation processing on the truncation configuration parameters of the sorting service node according to the current degradation round and the preset truncation step size. Specifically, for scenarios with high user experience requirements, the corresponding preset truncation step size can be smaller. That is, in object recommendation scenarios where user experience is prioritized, more candidate objects can be retained during the sorting stage to ensure user experience. Conversely, for scenarios with low user experience requirements, the corresponding preset truncation step size can be larger. That is, in object recommendation scenarios where user experience is prioritized, fewer candidate objects can be retained during the sorting stage to alleviate the pressure on computing resources.

[0037] For example, the server can use pre-trained parameter values ​​to determine the model, based on the current degradation round, the preset truncation step size, and the current water level information of the sorting service node, to determine the target parameter value, and then use the target parameter value as the parameter value of the truncation configuration parameter of the sorting node after parameter degradation. The parameter value determination model can be a model built based on a preset machine learning algorithm.

[0038] In step S208, object recall processing is performed based on the recall service node after parameter downgrading to obtain the first candidate object. Then, the sorting service node after parameter downgrading is used to perform truncation and filtering processing on the first candidate object to determine the object recommendation result based on the truncation and filtering processing result.

[0039] In practice, taking the information flow recommendation scenario of rentable goods as an example, the server can perform object recall processing through the recall service node after parameter downgrading based on user information to recall the first candidate object that matches the user. Then, the server can use the sorting service node after parameter downgrading to perform relevance scoring processing on the first candidate object and the user, and perform truncation and filtering processing on the first candidate object based on the relevance scoring results. Finally, the server can determine the object recommendation result for the user based on the truncation and filtering processing result, so as to recommend the filtered result to the user in the rental platform.

[0040] Among them, the information flow recommendation scenario refers to the intelligent recommendation system that pushes rentable goods to users in the form of an information flow within the rental platform. It is a key path to facilitate the conversion of rental orders. The rental platform refers to the core business entry platform for users in the circular economy rental industry, which may include core functional modules such as product display, search, recommendation flow, and order placement.

[0041] This specification provides a service optimization method that can acquire target service nodes in a recommendation system whose service level is not at a safe level. For recall service nodes among the target service nodes, the recall configuration parameters of the recall service nodes are downgraded based on a preset downgrade gradient. For ranking service nodes among the target service nodes, the truncation configuration parameters of the ranking service nodes are downgraded based on the current downgrade round and a preset truncation step size. Based on the parameter-downgraded recall service nodes, object recall processing is performed to obtain a first candidate object. Then, using the parameter-downgraded ranking service nodes, the first candidate object is truncated and filtered to determine the object recommendation result based on the truncation and filtering results. In this way, on the one hand, for recall service nodes that are not at a safe level, "gradual degradation" can be achieved through preset degradation gradients, avoiding a precipitous drop in user service experience caused by a global on / off degradation, thus ensuring the core user experience. On the other hand, for ranking service nodes that are not at a safe level, parameter degradation can be quickly and accurately performed on the truncation configuration parameters based on the current degradation round and preset truncation step size, avoiding the high error rate problem of manually configuring truncation parameters, thus ensuring the maintainability of the recommendation system. Furthermore, by using the recall service nodes and ranking service nodes after parameter degradation, a closed-loop control of "accurate overload identification - intelligent decision degradation - smooth execution degradation" can be achieved, solving the problem of uncontrollable degradation of recommendation services under high concurrency and ensuring the object recommendation effect under sudden traffic.

[0042] In practical applications, the specific processing method for downgrading the recall configuration parameters of the recall service node in step S204 of the above-mentioned target service node based on the preset degradation gradient can be varied. The recall service node can contain multiple recall child nodes. Accordingly, the following provides an optional processing method, such as... Figure 4 As shown, the specific process may include the following steps S2042 to S2044.

[0043] In step S2042, the target recall sub-node is determined among multiple recall sub-nodes based on the recall strategy corresponding to each recall sub-node and the recall priority corresponding to the recall strategy.

[0044] The recall sub-nodes can be used to recall objects based on a preset recall strategy. The recall strategy can be a vector recall strategy, a U2I recall strategy, or a constant recall strategy. The priority of the node corresponding to the recall sub-node can be determined based on the historical number of objects recalled by the recall strategy corresponding to the recall sub-node, the recall requirements or experience requirements of the specific application scenario of the recommendation system, etc.

[0045] In implementation, the server can build different recall sub-nodes based on different recall strategies, and when the recall service node exceeds its limits, it can select the target recall sub-node that needs to be downgraded from the multiple recall sub-nodes contained in the recall service node.

[0046] For example, the server can determine the node priority of each recall sub-node based on the recall strategy corresponding to the recall sub-node and the recall priority corresponding to the recall strategy. Then, based on the node priority, the server can select the target recall sub-node. Specifically, assuming that the recall sub-nodes include recall sub-node 1 corresponding to the vector recall strategy, recall sub-node 2 corresponding to the U2I recall strategy, and recall sub-node 3 corresponding to the constant recall strategy, if the priority of the vector recall strategy is higher than that of the U2I recall strategy and the constant recall strategy, then the node priority of recall sub-node 1 is higher than that of recall sub-node 2 and recall sub-node 3. That is, the server can determine recall sub-node 1 as the target recall sub-node.

[0047] In step S2044, the recall configuration parameters of the target recall sub-node are downgraded based on the preset downgrade gradient.

[0048] In implementation, the server can downgrade the recall configuration parameters of the target recall sub-nodes based on a preset degradation gradient. Furthermore, after downgrading the recall configuration parameters of the target recall sub-nodes, the server can still determine whether the service level of the recall service node is at a safe level.

[0049] If the recalled service node after parameter downgrading is still not at a safe level, the server can continue to determine target recalled sub-nodes based on the node priority of the recalled sub-nodes, and perform parameter downgrading on the currently determined target recalled sub-nodes. For example, taking the above three recalled sub-nodes as an example, if the service level of the recalled service node after parameter downgrading for recalled sub-node 1 is still not at a safe level, then if the node priority of recalled sub-node 2 is higher than that of recalled sub-node 3, then the server can continue to perform parameter downgrading on recalled sub-node 2 based on the preset downgrading gradient.

[0050] Furthermore, since the number of candidate objects recalled by different recall strategies may vary significantly, different preset degradation gradients can be configured for different recall sub-nodes to achieve a more even distribution of recalled objects. For example, if the number of historical recalled objects for recall sub-node 1 is greater than the number of historical recalled objects for recall sub-node 2, then the preset degradation gradient for recall sub-node 1 can be greater than the preset degradation gradient for recall sub-node 2.

[0051] In practical applications, the specific processing method for downgrading the truncation configuration parameters of the sorting service node based on the current downgrading round and the preset truncation step size in step S206 above can vary. The truncation configuration parameters may include the number of truncation objects. Accordingly, the following provides an optional processing method, such as... Figure 5 As shown, the specific process may include the following steps S2062 to S2064.

[0052] In step S2062, the target update quantity is determined based on the current downgrade round and the preset truncation step size.

[0053] In practice, after the sorting service node is downgraded in terms of parameters, the service level of the sorting service node after parameter downgrading can be monitored. If the service level of the sorting service node after parameter downgrading is still not at a safe level, parameter downgrading can continue. Therefore, there may be multiple rounds of parameter economy, that is, the server can determine the target update quantity based on the current downgrading round and the preset truncation step size.

[0054] For example, the server can determine the target update quantity by multiplying the current downgrade round and the preset truncation step size.

[0055] In step S2064, based on the first parameter value corresponding to the truncated object quantity parameter of the current sorting service node and the target update quantity, the truncated object quantity parameter of the sorting service node is downgraded.

[0056] In implementation, the server can downgrade the truncation object quantity parameter of the sorting service node based on the difference between the first parameter value and the target update quantity. For example, assuming the current truncation object quantity parameter of the sorting service node is 100 (i.e., the first parameter value is 100), if the difference between the first parameter value and the target update quantity is 50, then the truncation object quantity parameter of the sorting service node after parameter downgrading will be 50.

[0057] That is, Y = trunc - n * b, where Y is the parameter value of the number of truncated objects in the sorting service node after parameter downgrading, trunc is the first parameter value, n is the current downgrading round, and b is the preset truncation step size (e.g., 50, etc.).

[0058] In practical applications, the specific processing method for downgrading the parameter of the truncated object quantity parameter of the sorting service node based on the first parameter value corresponding to the truncated object quantity parameter of the current sorting service node and the target update quantity in step S2064 can be varied. The following provides an optional processing method, which may specifically include the processing of steps A1 to A2.

[0059] In step A1, based on the target update quantity and the first parameter value, the second parameter value is determined, and it is determined whether the second parameter value is less than the preset parameter threshold.

[0060] In step A2, if the value of the second parameter is not less than the preset parameter threshold, then the value of the second parameter is determined as the parameter value corresponding to the truncated object number parameter of the sorting service node after parameter downgrading; if the value of the second parameter is less than the preset parameter threshold, then the preset parameter threshold is determined as the parameter value corresponding to the truncated object number parameter of the sorting service node after parameter downgrading.

[0061] In implementation, the server can determine the second parameter value as the parameter value corresponding to the truncated object number parameter of the sorting service node after parameter downgrading, provided that the second parameter value is not less than the preset parameter threshold. Conversely, if the second parameter value is less than the preset parameter threshold, the preset parameter threshold can be determined as the parameter value corresponding to the truncated object number parameter of the sorting service node after parameter downgrading. This ensures that the number of candidate objects output by the sorting service node is not less than the preset parameter threshold, thereby guaranteeing the user's service experience.

[0062] In practical applications, the sorting service node with downgraded parameters in step S208 above performs truncation and filtering on the first candidate object. The specific processing method for determining the object recommendation result based on the truncation and filtering results can vary. The following provides one optional processing method, such as... Figure 6 As shown, the specific process may include the following steps S2082 to S2084.

[0063] In step S2082, the sorting service node after parameter downgrading is used to perform truncation and filtering on the first candidate object to obtain the truncation and filtering result.

[0064] In implementation, the server can use the sorting service node with downgraded parameters to score the relevance between the first candidate object and the user, and based on the relevance score, perform truncation and filtering on the first candidate object to obtain the truncation and filtering result.

[0065] In practical applications, the sorting service node after parameter downgrading in step S2082 above is used to perform truncation and filtering on the first candidate object. The specific processing method for obtaining the truncation and filtering result can be varied. The sorting service node can include the first sorting node and the second sorting node. Accordingly, the following provides an optional processing method, which can specifically include the processing of steps B1 to B2.

[0066] In step B1, using the first sorting node after parameter downgrading, based on the pre-trained first model, the first matching degree between the first candidate object and the user is determined, and based on the first matching degree and the first truncation number corresponding to the first sorting node, the first candidate object is truncated and filtered to obtain the second candidate object.

[0067] The first model can be a model built based on a preset machine learning algorithm to determine the matching degree between candidate objects and users.

[0068] In step B2, using the second sorting node after parameter downgrading, based on the pre-trained second model, the second matching degree between the second candidate object and the user is determined. Based on the second matching degree and the second truncation number corresponding to the second sorting node, the second candidate object is truncated and filtered to obtain the truncation and filtering result.

[0069] The number of model parameters in the first model can be smaller than the number of model parameters in the second model.

[0070] In implementation, the server can use the first sorting node after parameter downgrading to perform coarse sorting of the first candidate objects based on the first model with a smaller parameter magnitude. Then, using the second sorting node after parameter downgrading to perform further truncation and filtering of the coarse sorting results based on the second model with a larger parameter magnitude, the filtered first candidate objects can be obtained.

[0071] The server can determine the truncation configuration parameters to be downgraded based on the type of the sorting node (i.e., coarse-sorting type and fine-sorting type). For example, if the upstream node of the sorting node is the merge node RECALL_MERGE, then the sorting node is the first sorting node. In this case, merge.config.finalCutSize or global recall.base.trunc.config can be obtained as the truncation configuration parameters to be downgraded for the current first sorting node. Based on the current downgrade round and the preset truncation step size, the parameter value corresponding to the truncation configuration parameter is downgraded.

[0072] Similarly, if the upstream node of the sorting node is the truncated node ITEM_TRUNC, then the sorting node can be the second sorting node. In this case, the server can obtain configAbKey as the truncated configuration parameter to be downgraded corresponding to the current second sorting node, and perform parameter downgrading processing on the parameter value corresponding to the truncated configuration parameter based on the current downgrade round and the preset truncated step size.

[0073] In addition, if the upstream node of the sorted node is not the aforementioned truncated node or merged node, the graph structure data can be recursively searched upwards, and the update process can be skipped if no truncated node or merged node is found.

[0074] In step S2084, a preset fallback candidate object corresponding to the location information of the recommendation system is obtained, and the object recommendation result is determined based on the preset fallback candidate object and the truncation and filtering results.

[0075] Among them, the preset backup candidate can be a candidate selected based on the popularity information of the object determined by historical exposure number, historical transaction data, etc.

[0076] In implementation, for example, the server can obtain the historical exposure count and historical transaction data of candidate objects based on the location information of the recommendation system, and determine the object popularity information of each candidate object based on the obtained information, so as to filter out the preset fallback candidate objects based on the object popularity information, and store the preset fallback candidate objects in the local cache, so as to supplement the object recommendation results as a fallback result when service degradation is required.

[0077] To avoid a shortage of candidate objects in the object recommendation results due to parameter downgrading, the server can determine the object recommendation results based on preset fallback candidate objects and truncation filtering results. For example, if the number of objects in the truncation filtering results is less than a preset threshold, the server can filter the preset fallback candidate objects to determine the objects in the filtered preset fallback candidate objects and the objects in the truncation filtering results as the object recommendation results.

[0078] In this way, when the recall component in the recommendation system receives the skip instruction, on the one hand, it can skip the corresponding recall loop (i.e., recall child nodes) according to the node priority, and on the other hand, it can inject the city popular cache (i.e., cache data containing the fallback candidate object) as a fallback recall during the recall stage.

[0079] In the ranking component of the recommendation system, when consuming the first candidate after degradation, the popular city fallback results (i.e., the pre-recalled fallback candidates) can be added during the fine ranking stage. In addition, the recommendation system can be configured with a reasonable failure tolerance to avoid the overall ranking failure due to the degradation of individual nodes. If there is a manual degradation backdoor instruction, the manually configured skip instruction can be executed first.

[0080] In practical applications, even if the recall service node after parameter downgrading is still not at a safe level, parameter downgrading can continue. There are various ways to continue parameter downgrading; the following is one optional method: Figure 7 As shown, the specific process may include the following steps S702 to S704.

[0081] In step S702, it is determined whether the service level corresponding to the recall service node after parameter downgrading is at a safe level within the first detection period.

[0082] The first detection period can be any period, such as the last 1 minute, the last 5 minutes, etc. The first detection period can be determined based on the current degradation gradient and the corresponding experience requirements of the recommendation system. For example, if the current degradation gradient is large, the corresponding first detection period can be long, and if the current degradation gradient is small, the corresponding first detection period can be short.

[0083] In implementation, the decision-making logic of a recommendation system can be as follows: Figure 8 As shown, firstly, the server can read the identification information of the service nodes related to the recommendation system in the graph structure data, and based on the obtained identification information, read the water level information of each service node.

[0084] In this process, once the water level information of the service node is read, it can be determined whether the service water level of the service node is at a safe level based on the current water level information. If the service water level of the recalled service node is not at a safe level, parameter downgrading can be performed on the recalled service node, and the service water level of the recalled service node after parameter downgrading can be determined whether it is at a safe level within the first detection period.

[0085] Additionally, if the water level information of the service node is not read, such as Figure 8As shown, it can be determined whether there is downgraded snapshot information in the local cache. If there is, the current downgrade parameters can be consumed. For example, the recall configuration parameters of the recall service node can be downgraded based on the current preset downgrade gradient. If there is no downgrade, no downgrade processing can be performed.

[0086] In step S704, if the service level corresponding to the recall service node after parameter downgrading is not at a safe level within the first detection period, the preset downgrading gradient is updated, and based on the updated downgrading gradient, the recall configuration parameters of the recall service node are further downgraded.

[0087] In implementation, if the service level of the recall service node after parameter downgrading is not at a safe level within the first detection period, then... Figure 8 As shown, the preset degradation gradient can be updated, and based on the updated degradation gradient, the parameter degradation processing can continue to be performed on the recall service node after parameter degradation, such as updating 50% of the preset degradation gradient to 100% of the preset degradation gradient.

[0088] The server can update the preset degradation gradient based on the preset degradation range, which can be determined according to the experience requirements of the recommendation system.

[0089] Or, such as Figure 8 As shown, if the service level of the recall service node after parameter downgrading is not at a safe level, it is also possible to determine whether the current downgrading time has exceeded the first detection period. If it has exceeded the first detection period, the preset downgrading gradient can be updated, and the parameter downgrading process can continue to be performed on the recall service node after parameter downgrading based on the updated downgrading gradient. If it has not exceeded the first detection period, the current downgraded parameters can continue to be consumed, that is, the recall service can continue to be performed on the recall service node after parameter downgrading based on the current parameters, and parameter downgrading will not be performed again.

[0090] In addition, such as Figure 8 As shown, the recommendation system can also record execution operation logs, including service status information and degradation information for each service node.

[0091] In practical applications, parameter recovery can also be performed when the recall service node after parameter degradation is at a safe level. There are various methods for parameter recovery; the following provides one optional method: Figure 9 As shown, the specific process may include the following steps S902 to S906.

[0092] In step S902, if the service level of the recall service node after parameter downgrading is at a safe level within the first detection period, then the recall configuration parameters of the recall service node are restored based on the preset recovery gradient.

[0093] In implementation, such as Figure 8 As shown, if the service level of the recall service node after parameter downgrading is at a safe level during the first detection period, parameter restoration processing can be performed on the recall service node to ensure the object recommendation effect.

[0094] In step S904, it is determined whether the service level corresponding to the recall service node after parameter recovery is at a safe level within the second detection period.

[0095] The second detection cycle is longer than the first detection cycle. For example, the first detection cycle can be 1 minute and the second detection cycle can be 2 minutes.

[0096] In step S906, if the service level corresponding to the recalled service node after parameter recovery is at a safe level within the second detection period, the preset recovery gradient is updated, and based on the updated recovery gradient, the recall configuration parameters of the recalled service node are further processed for parameter recovery.

[0097] The preset recovery gradient can be multiple, such as 10%, 30%, 50%, 60%, 70%, 80%, 90%, and 100%, which can be determined according to the current recovery status and the experience requirements of the recommendation system.

[0098] In implementation, the server can update the preset recovery gradient based on the preset recovery step size. For example, the preset recovery step size can be 20%. That is, if the current preset recovery gradient is 10%, then the updated recovery gradient can be 10% + 20% = 30%.

[0099] In addition, such as Figure 8 As shown, during parameter recovery processing, it can also be determined whether the parameter recovery time exceeds the second detection period. If it exceeds the second detection period, the preset recovery gradient can be updated, and the parameter recovery processing can continue for the recalled service node after parameter recovery processing based on the updated recovery gradient. If it does not exceed the second detection period, the current recovery parameters (i.e., the preset recovery gradient) can be consumed, that is, the recalled service can be recalled based on the recalled service node after the current parameter recovery processing, and the service recovery processing can not continue.

[0100] In this way, by deeply integrating multi-dimensional system level perception (such as CPU utilization combined with response time RT), graph structure data parsing, and gradient-based priority degradation strategies, a closed-loop control of "accurate overload identification—intelligent decision-making degradation—smooth execution degradation" can be achieved. This fundamentally solves the problem of uncontrollable degradation of recommendation services under high concurrency. Specifically, simultaneously detecting CPU utilization and response time RT can avoid misjudgment based on a single indicator, improve the accuracy of overload identification, and achieve multi-dimensional level fusion perception. By traversing graph structure data and automatically extracting the truncated configuration parameters and service dependencies corresponding to the sorting service nodes, manual configuration errors can be eliminated. Through graph-driven rule self-parsing, the maintainability of the system is improved, and multiple preset degradation gradients can be supported. Parameter degradation can be performed according to the node priority of the recalled child nodes. Through the gradient-based degradation mechanism combined with priority, "gradual degradation" can be achieved, ensuring the core experience. In addition, local rule caching, combined with asynchronous refresh of a small amount of traffic, can balance real-time performance and performance overhead, enabling hot rule updates without affecting the performance of the main link. In scenarios with sudden traffic surges, the P99 latency fluctuation of the recommendation system can be reduced by more than 60%, and the service availability can be maintained at more than 99.9%. At the same time, the decline in recommendation conversion rate can also be controlled within an acceptable range (e.g., no more than 15%).

[0101] This specification provides a service optimization method that can acquire target service nodes in a recommendation system whose service level is not at a safe level. For recall service nodes among the target service nodes, the recall configuration parameters of the recall service nodes are downgraded based on a preset downgrade gradient. For ranking service nodes among the target service nodes, the truncation configuration parameters of the ranking service nodes are downgraded based on the current downgrade round and a preset truncation step size. Based on the parameter-downgraded recall service nodes, object recall processing is performed to obtain a first candidate object. Then, using the parameter-downgraded ranking service nodes, the first candidate object is truncated and filtered to determine the object recommendation result based on the truncation and filtering results. In this way, on the one hand, for recall service nodes that are not at a safe level, "gradual degradation" can be achieved through preset degradation gradients, avoiding a precipitous drop in user service experience caused by a global on / off degradation, thus ensuring the core user experience. On the other hand, for ranking service nodes that are not at a safe level, parameter degradation can be quickly and accurately performed on the truncation configuration parameters based on the current degradation round and preset truncation step size, avoiding the high error rate problem of manually configuring truncation parameters, thus ensuring the maintainability of the recommendation system. Furthermore, by using the recall service nodes and ranking service nodes after parameter degradation, a closed-loop control of "accurate overload identification - intelligent decision degradation - smooth execution degradation" can be achieved, solving the problem of uncontrollable degradation of recommendation services under high concurrency and ensuring the object recommendation effect under sudden traffic.

[0102] The above describes the service optimization method provided in the embodiments of this specification. Based on the same idea, the embodiments of this specification also provide a service optimization device, such as... Figure 10 As shown.

[0103] The service optimization device includes: a node acquisition module 1001, a service degradation module 1002, a truncation update module 1003, and an object determination module 1004, wherein: Node acquisition module 1001 acquires target service nodes in the recommendation system whose service level is not at a safe level. The service degradation module 1002 performs parameter degradation processing on the recall configuration parameters of the recall service node in the target service node based on a preset degradation gradient. The truncation update module 1003 performs parameter downgrading processing on the truncation configuration parameters of the sorting service node in the target service node according to the current downgrading round and the preset truncation step size. The object recommendation module 1004 performs object recall processing based on the recall service node after parameter downgrading to obtain the first candidate object, and uses the sorting service node after parameter downgrading to perform truncation and filtering processing on the first candidate object, so as to determine the object recommendation result based on the truncation and filtering processing result.

[0104] In the embodiments described in this specification, the recall service node includes multiple recall sub-nodes, and the service degradation module 1002 is used for: Based on the recall strategy corresponding to each of the recall sub-nodes and the recall priority corresponding to the recall strategy, the target recall sub-node among the plurality of recall sub-nodes is determined; Based on the preset degradation gradient, the recall configuration parameters of the target recall sub-node are downgraded.

[0105] In the embodiments of this specification, the truncation configuration parameters include a parameter for the number of truncation objects, and the truncation update module 1003 is used for: Based on the current downgrade round and the preset cutoff step size, determine the target update quantity; Based on the first parameter value corresponding to the truncated object quantity parameter of the current sorting service node, and the target update quantity, the truncated object quantity parameter of the sorting service node is downgraded.

[0106] In the embodiments described in this specification, the truncation update module 1003 is used for: Based on the target update quantity and the first parameter value, determine the second parameter value, and determine whether the second parameter value is less than a preset parameter threshold; If the second parameter value is not less than the preset parameter threshold, then the second parameter value is determined as the parameter value corresponding to the truncated object number parameter of the sorting service node after parameter downgrading. If the second parameter value is less than the preset parameter threshold, then the preset parameter threshold is determined as the parameter value corresponding to the truncated object number parameter of the sorting service node after parameter downgrading.

[0107] In this embodiment of the specification, the object recommendation module 1004 is used for: Using the sorting service node after parameter downgrading, the first candidate object is truncated and filtered to obtain the truncated filtering result; Obtain a preset fallback candidate object corresponding to the location information of the recommendation system, and determine the object recommendation result based on the preset fallback candidate object and the truncation and filtering processing result.

[0108] In this embodiment of the specification, the sorting service node includes a first sorting node and a second sorting node, and the object recommendation module 1004 is used for: Using the first sorting node after parameter downgrading, based on the pre-trained first model, the first matching degree between the first candidate object and the user is determined, and based on the first matching degree and the first truncation number corresponding to the first sorting node, the first candidate object is truncated and filtered to obtain the second candidate object. Using the second sorting node after parameter downgrading, and based on the pre-trained second model, the second matching degree between the second candidate object and the user is determined. Based on the second matching degree and the second truncation number corresponding to the second sorting node, the second candidate object is truncated and filtered to obtain the truncation and filtering result. The number of model parameters in the first model is smaller than the number of model parameters in the second model.

[0109] In the embodiments described in this specification, the device further includes: The first judgment module determines whether the service level of the recall service node after the parameter downgrade is at a safe level within the first detection period. If the service level of the recall service node after parameter downgrading is not at a safe level within the first detection period, the downgrading module updates the preset downgrading gradient and continues to downgrade the recall configuration parameters of the recall service node based on the updated downgrading gradient.

[0110] In the embodiments described in this specification, the device further includes: The first recovery module, if the service level corresponding to the recall service node after parameter downgrading is at a safe level within the first detection period, performs parameter recovery processing on the recall configuration parameters of the recall service node based on a preset recovery gradient. The second judgment module determines whether the service level of the recall service node after parameter recovery is at a safe level within the second detection period, wherein the second detection period is longer than the first detection period. If the service level corresponding to the recalled service node after parameter recovery is at a safe level within the second detection period, the second recovery module updates the preset recovery gradient and continues to perform parameter recovery processing on the recall configuration parameters of the recalled service node based on the updated recovery gradient.

[0111] This specification provides a service optimization device that can acquire target service nodes in a recommendation system whose service level is not at a safe level. For recall service nodes among the target service nodes, the device performs parameter downgrading processing on the recall configuration parameters of the recall service nodes based on a preset downgrading gradient. For ranking service nodes among the target service nodes, the device performs parameter downgrading processing on the truncation configuration parameters of the ranking service nodes based on the current downgrading round and a preset truncation step size. Based on the parameter-downgraded recall service nodes, the device performs object recall processing to obtain a first candidate object. Then, using the parameter-downgraded ranking service nodes, the device performs truncation and filtering processing on the first candidate object to determine the object recommendation result based on the truncation and filtering processing result. In this way, on the one hand, for recall service nodes that are not at a safe level, "gradual degradation" can be achieved through preset degradation gradients, avoiding a precipitous drop in user service experience caused by a global on / off degradation, thus ensuring the core user experience. On the other hand, for ranking service nodes that are not at a safe level, parameter degradation can be quickly and accurately performed on the truncation configuration parameters based on the current degradation round and preset truncation step size, avoiding the high error rate problem of manually configuring truncation parameters, thus ensuring the maintainability of the recommendation system. Furthermore, by using the recall service nodes and ranking service nodes after parameter degradation, a closed-loop control of "accurate overload identification - intelligent decision degradation - smooth execution degradation" can be achieved, solving the problem of uncontrollable degradation of recommendation services under high concurrency and ensuring the object recommendation effect under sudden traffic.

[0112] The above are service optimization devices provided in the embodiments of this specification. Based on the same idea, the embodiments of this specification also provide a service optimization device, such as... Figure 11 As shown.

[0113] The service optimization device can provide terminal devices or servers, etc., for the above embodiments.

[0114] like Figure 11 As shown, device 1100 mainly consists of a communication interface 1102, a user interface 1104, a processor 1106, and a data storage 1108. These components are interconnected and communicate with each other via a system bus, network, or other connection mechanism 1110. The communication interface 1102 enables device 1100 to communicate with other devices, access networks, and transmission networks via analog or digital modulation. For example, the communication interface 1102 may include a chipset and antenna for wireless communication with a radio access network or access point. Furthermore, the communication interface 1102 can be a wired interface such as Ethernet, Token Ring, or a USB port, or a wireless interface such as Wi-Fi, Bluetooth, Global Positioning System (GPS), or a wide-area wireless interface (e.g., WiMAX or LTE). Of course, the communication interface 1102 can also support other forms of physical layer interfaces and standard or proprietary communication protocols. The communication interface 1102 may also include multiple physical communication interfaces, such as Wi-Fi, Bluetooth, and wide-area wireless interfaces.

[0115] User interface 1104 includes receiving user input and providing output to the user. Therefore, user interface 1104 may include input components such as a keypad, keyboard, touch-sensitive or presence-sensitive panel, computer mouse, trackball, joystick, microphone, still camera, and video camera, and output components such as a display screen (which may be combined with a touch-sensitive panel), CRT, LCD, LED, display using DLP technology, printer, and other similar devices known or developed in the future. User interface 1104 may also generate auditory output via speakers, speaker jacks, audio output ports, audio output devices, headphones, and other similar devices known or developed in the future. In some embodiments, user interface 1104 may include software, circuitry, or other forms of logic capable of transmitting and receiving data from external user input / output devices. Additionally or alternatively, device 1100 may support remote access from other devices via communication interface 1102 or another physical interface (not shown). User interface 1104 may be configured to receive user input, the position and movement of which may be indicated by indicators or cursors described herein. User interface 1104 can also be configured as a display device for rendering or displaying text fragments.

[0116] Processor 1106 may contain one or more general-purpose processors and / or special-purpose processors.

[0117] Data storage 1108 may include one or more volatile storage components and may be integrated wholly or partially with processor 1106. Data storage 1108 may include removable and non-removable components.

[0118] Processor 1106 is capable of executing program instructions 1118 (e.g., compiled or uncompiled program logic and / or machine code) stored in data storage 1108 to perform the various functions described herein. Data storage 1108 may contain a non-transitory computer-readable medium on which program instructions are stored, which, when executed by device 1100, enable device 1100 to perform any methods, processes, or functions disclosed in this specification and / or the accompanying drawings. Execution of program instructions 1118 by processor 1106 may result in processor 1106 using data 1112.

[0119] For example, program instructions 1118 may include an operating system 1122 (e.g., an operating system kernel, device drivers, and / or other modules) installed on device 1100 and one or more applications 1120 (e.g., a browser, social application, or game application). Similarly, data 1112 may include operating system data 1116 and application data 1114. Operating system data 1116 is primarily accessible to the operating system 1122, while application data 1114 is primarily accessible to one or more applications 1120. Application data 1114 may reside in a file system visible or hidden from the user of device 1100.

[0120] Application 1120 can communicate with operating system 1112 through one or more application programming interfaces (APIs). These APIs help application 1120 read and / or write application data 1114, transmit or receive information via communication interface 1102, receive or display information on user interface 1104, etc.

[0121] In some terminology, application 1120 may be simply referred to as "app". Furthermore, application 1120 can be downloaded to device 1100 through one or more online app stores or app markets. However, applications can also be installed on device 1100 in other ways, such as through a web browser or a physical interface on device 1100 (e.g., a USB port).

[0122] Specifically, in this embodiment, the service optimization device includes a memory and one or more programs, wherein one or more programs are stored in the memory, and one or more programs may include one or more modules, and each module may include a series of computer-executable instructions for the service optimization device, and is configured to be executed by one or more processors. The one or more programs include computer-executable instructions for performing the following: Identify target service nodes in the recommendation system whose service level is not at a safe level. For the recall service node in the target service node, the recall configuration parameters of the recall service node are downgraded based on the preset downgrade gradient. For the sorting service node in the target service node, the truncation configuration parameters of the sorting service node are downgraded according to the current downgrade round and the preset truncation step size. The object recall process is performed based on the recall service node after parameter downgrading to obtain the first candidate object. Then, the sorting service node after parameter downgrading is used to perform truncation and filtering on the first candidate object to determine the object recommendation result based on the truncation and filtering result.

[0123] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the service optimization device embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0124] This specification provides a service optimization device that can acquire target service nodes in a recommendation system whose service level is not at a safe level. For recall service nodes among the target service nodes, the device performs parameter downgrading processing on the recall configuration parameters of the recall service nodes based on a preset downgrading gradient. For ranking service nodes among the target service nodes, the device performs parameter downgrading processing on the truncation configuration parameters of the ranking service nodes based on the current downgrading round and a preset truncation step size. Based on the parameter-downgraded recall service nodes, the device performs object recall processing to obtain a first candidate object. Then, using the parameter-downgraded ranking service nodes, the device performs truncation and filtering processing on the first candidate object to determine the object recommendation result based on the truncation and filtering processing result. In this way, on the one hand, for recall service nodes that are not at a safe level, "gradual degradation" can be achieved through preset degradation gradients, avoiding a precipitous drop in user service experience caused by a global on / off degradation, thus ensuring the core user experience. On the other hand, for ranking service nodes that are not at a safe level, parameter degradation can be quickly and accurately performed on the truncation configuration parameters based on the current degradation round and preset truncation step size, avoiding the high error rate problem of manually configuring truncation parameters, thus ensuring the maintainability of the recommendation system. Furthermore, by using the recall service nodes and ranking service nodes after parameter degradation, a closed-loop control of "accurate overload identification - intelligent decision degradation - smooth execution degradation" can be achieved, solving the problem of uncontrollable degradation of recommendation services under high concurrency and ensuring the object recommendation effect under sudden traffic.

[0125] Furthermore, based on the above Figures 1 to 9This specification also provides a storage medium for storing computer-executable instruction information in one or more embodiments. In one specific embodiment, the storage medium may be a USB flash drive, optical disc, hard disk, etc. When the computer-executable instruction information stored in the storage medium is executed by a processor, it can realize the following process: Identify target service nodes in the recommendation system whose service level is not at a safe level. For the recall service node in the target service node, the recall configuration parameters of the recall service node are downgraded based on the preset downgrade gradient. For the sorting service node in the target service node, the truncation configuration parameters of the sorting service node are downgraded according to the current downgrade round and the preset truncation step size. The object recall process is performed based on the recall service node after parameter downgrading to obtain the first candidate object. Then, the sorting service node after parameter downgrading is used to perform truncation and filtering on the first candidate object to determine the object recommendation result based on the truncation and filtering result.

[0126] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the above-described storage medium embodiment is basically similar to the method embodiment, so the description is relatively simple; relevant parts can be referred to the description of the method embodiment.

[0127] This specification provides a storage medium that can acquire target service nodes in a recommendation system whose service level is not at a safe level. For recall service nodes among the target service nodes, based on a preset degradation gradient, the recall configuration parameters of the recall service nodes are downgraded. For ranking service nodes among the target service nodes, based on the current degradation round and a preset truncation step size, the truncation configuration parameters of the ranking service nodes are downgraded. Based on the parameter-downgraded recall service nodes, object recall processing is performed to obtain a first candidate object. Then, using the parameter-downgraded ranking service nodes, the first candidate object is truncated and filtered to determine the object recommendation result based on the truncation and filtering results. In this way, on the one hand, for recall service nodes that are not at a safe level, "gradual degradation" can be achieved through preset degradation gradients, avoiding a precipitous drop in user service experience caused by a global on / off degradation, thus ensuring the core user experience. On the other hand, for ranking service nodes that are not at a safe level, parameter degradation can be quickly and accurately performed on the truncation configuration parameters based on the current degradation round and preset truncation step size, avoiding the high error rate problem of manually configuring truncation parameters, thus ensuring the maintainability of the recommendation system. Furthermore, by using the recall service nodes and ranking service nodes after parameter degradation, a closed-loop control of "accurate overload identification - intelligent decision degradation - smooth execution degradation" can be achieved, solving the problem of uncontrollable degradation of recommendation services under high concurrency and ensuring the object recommendation effect under sudden traffic.

[0128] Furthermore, based on the above Figures 1 to 9 This specification also provides one or more embodiments of a computer program product, including a computer program, which, when executed by a processor, can perform the following processes: Identify target service nodes in the recommendation system whose service level is not at a safe level. For the recall service node in the target service node, the recall configuration parameters of the recall service node are downgraded based on the preset downgrade gradient. For the sorting service node in the target service node, the truncation configuration parameters of the sorting service node are downgraded according to the current downgrade round and the preset truncation step size. The object recall process is performed based on the recall service node after parameter downgrading to obtain the first candidate object. Then, the sorting service node after parameter downgrading is used to perform truncation and filtering on the first candidate object to determine the object recommendation result based on the truncation and filtering result.

[0129] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the above-described embodiment of a computer program product is relatively simple in description because it is fundamentally similar to the method embodiment; relevant parts can be referred to the description of the method embodiment.

[0130] This specification provides a computer program product that can acquire target service nodes in a recommendation system whose service level is not at a safe level. For recall service nodes among the target service nodes, based on a preset degradation gradient, the recall configuration parameters of the recall service nodes are downgraded. For ranking service nodes among the target service nodes, based on the current degradation round and a preset truncation step size, the truncation configuration parameters of the ranking service nodes are downgraded. Based on the parameter-downgraded recall service nodes, object recall processing is performed to obtain a first candidate object. Then, using the parameter-downgraded ranking service nodes, the first candidate object is truncated and filtered to determine the object recommendation result based on the truncation and filtering results. In this way, on the one hand, for recall service nodes that are not at a safe level, "gradual degradation" can be achieved through preset degradation gradients, avoiding a precipitous drop in user service experience caused by a global on / off degradation, thus ensuring the core user experience. On the other hand, for ranking service nodes that are not at a safe level, parameter degradation can be quickly and accurately performed on the truncation configuration parameters based on the current degradation round and preset truncation step size, avoiding the high error rate problem of manually configuring truncation parameters, thus ensuring the maintainability of the recommendation system. Furthermore, by using the recall service nodes and ranking service nodes after parameter degradation, a closed-loop control of "accurate overload identification - intelligent decision degradation - smooth execution degradation" can be achieved, solving the problem of uncontrollable degradation of recommendation services under high concurrency and ensuring the object recommendation effect under sudden traffic.

[0131] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0132] In the 1990s, improvements to a technology could be clearly distinguished as either hardware improvements (e.g., improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the methodology). However, with technological advancements, many methodological improvements today can be considered direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved methodology into the hardware circuit. Therefore, it cannot be said that a methodological improvement cannot be implemented using a hardware physical module. For example, a Programmable Logic Device (PLD) (such as a Field Programmable Gate Array (FPGA)) is such an integrated circuit whose logic function is determined by the user programming the device. Designers can program and "integrate" a digital system onto a PLD themselves, without needing chip manufacturers to design and manufacture dedicated integrated circuit chips. Furthermore, nowadays, instead of manually manufacturing integrated circuit chips, this programming is mostly implemented using "logic compiler" software. Similar to the software compiler used in program development, the original code before compilation must also be written in a specific programming language, called a Hardware Description Language (HDL). There are many HDLs, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, and RHDL (Ruby Hardware Description Language). Currently, the most commonly used are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also understand that by simply performing some logic programming on the method flow using one of these hardware description languages ​​and programming it into an integrated circuit, the hardware circuit implementing the logical method flow can be easily obtained.

[0133] The controller can be implemented in any suitable manner. For example, it can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320. A memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also recognize that, in addition to implementing the controller in purely computer-readable program code form, the same functionality can be achieved by logically programming the method steps to make the controller take the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the means included therein for implementing various functions can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0134] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0135] For ease of description, the above apparatus is described by dividing it into various functional units. Of course, when implementing one or more embodiments of this specification, the functions of each unit can be implemented in one or more software and / or hardware.

[0136] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, one or more embodiments of this specification may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, one or more embodiments of this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0137] The embodiments described herein are illustrated with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0138] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0139] These computer program instructions may also be loaded onto a computer or other programmable device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable device for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0140] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0141] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0142] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0143] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Without further limitations, the presence of additional identical elements in the process, method, article, or apparatus that includes said elements is not excluded.

[0144] It should also be noted that the terms "one," "an," and "the" do not specifically refer to the singular; they can also include the plural. Ordinal numbers such as "first," "second," etc., do not necessarily indicate order; often they are used to distinguish between objects. For example, "first server" and "second server" usually refer to two servers. To differentiate between these two servers, they are described as "first server" and "second server." Of course, sometimes these two servers may be the same server.

[0145] In this specification, unless explicitly stated otherwise, "receiving and sending data" does not necessarily mean direct receiving and sending; it can also mean indirect receiving and sending. For example, A receiving data sent by B can be understood as A directly receiving data sent by B, or it can be understood as A indirectly receiving data sent by B through other entities such as C. Similarly, B sending data to A can be understood as B sending data directly to A, or it can be understood as B indirectly sending data to A through other entities such as C. Here, C can be a single subject, or two or more subjects.

[0146] This specification uses specific terms to describe embodiments thereof. Terms such as "an embodiment," "one embodiment," and / or "some embodiments" refer to a particular feature, structure, or characteristic associated with at least one embodiment of this specification. Therefore, it should be emphasized and noted that references to "an embodiment," "one embodiment," or "an alternative embodiment" in different locations within this specification do not necessarily refer to the same embodiment. Furthermore, those skilled in the art can combine and integrate the different embodiments or examples described herein, as well as the features of those different embodiments or examples, without contradiction.

[0147] Although one or more embodiments of this specification provide method steps as described in the embodiments or flowcharts, it is understood that the order of steps listed in the embodiments or flowcharts is only one of many possible execution orders and does not represent the only execution order. Therefore, when the claims involve method steps, any changes or adjustments to the order of such steps, or the parallelism between steps, are also within the scope of protection of the claims.

[0148] It should be noted that the user data obtained in this manual is authorized by the user and does not involve user privacy.

[0149] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, one or more embodiments of this specification may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, one or more embodiments of this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0150] One or more embodiments of this specification can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. One or more embodiments of this specification can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0151] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0152] The above description is merely an embodiment of this specification and is not intended to limit this document. Various modifications and variations can be made to this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this specification should be included within the scope of the claims of this specification.

Claims

1. A service optimization method, comprising: Identify target service nodes in the recommendation system whose service level is not at a safe level. For the recall service node in the target service node, the recall configuration parameters of the recall service node are downgraded based on the preset downgrade gradient. For the sorting service node in the target service node, the truncation configuration parameters of the sorting service node are downgraded according to the current downgrade round and the preset truncation step size. The object recall process is performed based on the recall service node after parameter downgrading to obtain the first candidate object. Then, the sorting service node after parameter downgrading is used to perform truncation and filtering on the first candidate object to determine the object recommendation result based on the truncation and filtering result.

2. The method according to claim 1, wherein the recall service node comprises multiple recall sub-nodes, and the step of performing parameter downgrading processing on the recall configuration parameters of the recall service node based on a preset downgrading gradient includes: Based on the recall strategy corresponding to each of the recall sub-nodes and the recall priority corresponding to the recall strategy, the target recall sub-node among the plurality of recall sub-nodes is determined; Based on the preset degradation gradient, the recall configuration parameters of the target recall sub-node are downgraded.

3. The method according to claim 1, wherein the truncation configuration parameters include a truncation object quantity parameter, and the step of performing parameter downgrading processing on the truncation configuration parameters of the sorting service node according to the current downgrading round and a preset truncation step size includes: Based on the current downgrade round and the preset cutoff step size, determine the target update quantity; Based on the first parameter value corresponding to the truncated object quantity parameter of the current sorting service node, and the target update quantity, the truncated object quantity parameter of the sorting service node is downgraded.

4. The method according to claim 3, wherein the parameter downgrading processing of the truncated object quantity parameter of the sorting service node based on the first parameter value corresponding to the truncated object quantity parameter of the current sorting service node and the target update quantity includes: Based on the target update quantity and the first parameter value, determine the second parameter value, and determine whether the second parameter value is less than a preset parameter threshold; If the second parameter value is not less than the preset parameter threshold, then the second parameter value is determined as the parameter value corresponding to the truncated object number parameter of the sorting service node after parameter downgrading. If the second parameter value is less than the preset parameter threshold, then the preset parameter threshold is determined as the parameter value corresponding to the truncated object number parameter of the sorting service node after parameter downgrading.

5. The method according to claim 1, wherein the step of using the sorting service node after parameter downgrading to perform truncation and filtering processing on the first candidate object, so as to determine the object recommendation result based on the truncation and filtering processing result, includes: Using the sorting service node after parameter downgrading, the first candidate object is truncated and filtered to obtain the truncated filtering result; Obtain a preset fallback candidate object corresponding to the location information of the recommendation system, and determine the object recommendation result based on the preset fallback candidate object and the truncation and filtering processing result.

6. The method according to claim 5, wherein the sorting service node includes a first sorting node and a second sorting node. The ranking service node, after being downgraded using the parameters, performs truncation and filtering on the first candidate object to obtain the truncation and filtering result, including: Using the first sorting node after parameter downgrading, based on the pre-trained first model, the first matching degree between the first candidate object and the user is determined, and based on the first matching degree and the first truncation number corresponding to the first sorting node, the first candidate object is truncated and filtered to obtain the second candidate object. Using the second sorting node after parameter downgrading, and based on the pre-trained second model, the second matching degree between the second candidate object and the user is determined. Based on the second matching degree and the second truncation number corresponding to the second sorting node, the second candidate object is truncated and filtered to obtain the truncation and filtering result. The number of model parameters in the first model is smaller than the number of model parameters in the second model.

7. The method according to claim 1, further comprising: Determine whether the service level of the recall service node after the parameter downgrade is at a safe level within the first detection period. If the service level of the recall service node after parameter downgrading is not at a safe level within the first detection period, the preset downgrading gradient is updated, and based on the updated downgrading gradient, the recall configuration parameters of the recall service node are further downgraded.

8. The method according to claim 7, further comprising: If the service level of the recall service node after parameter downgrading is at a safe level during the first detection period, then the recall configuration parameters of the recall service node are restored based on the preset recovery gradient. Determine whether the service level of the recall service node after parameter recovery is at a safe level within the second detection period, wherein the second detection period is longer than the first detection period; If the service level of the recall service node after parameter recovery is at a safe level within the second detection period, the preset recovery gradient is updated, and based on the updated recovery gradient, the recall configuration parameters of the recall service node are further restored.

9. A service optimization apparatus, comprising: The node acquisition module acquires target service nodes in the recommendation system whose service level is not at a safe level. The service degradation module performs parameter degradation processing on the recall configuration parameters of the recall service node in the target service node based on a preset degradation gradient. The truncation update module performs parameter downgrading processing on the truncation configuration parameters of the sorting service node in the target service node, based on the current downgrading round and the preset truncation step size. The object recommendation module performs object retrieval processing based on the retrieval service node after parameter downgrading to obtain the first candidate object, and then uses the sorting service node after parameter downgrading to perform truncation and filtering processing on the first candidate object, so as to determine the object recommendation result based on the truncation and filtering processing result.

10. A service optimization device, wherein the service optimization includes: processor; as well as A memory configured to store computer-executable instructions, which, when executed, cause the processor to: Identify target service nodes in the recommendation system whose service level is not at a safe level. For the recall service node in the target service node, the recall configuration parameters of the recall service node are downgraded based on the preset downgrade gradient. For the sorting service node in the target service node, the truncation configuration parameters of the sorting service node are downgraded according to the current downgrade round and the preset truncation step size. The object recall process is performed based on the recall service node after parameter downgrading to obtain the first candidate object. Then, the sorting service node after parameter downgrading is used to perform truncation and filtering on the first candidate object to determine the object recommendation result based on the truncation and filtering result.

Images