Resource early warning method and device based on computing node cluster and electronic equipment

Abstract

Embodiments of the present disclosure disclose a resource warning method and device based on a computing node cluster and an electronic device. A specific implementation of the method comprises: determining whether to call the computing node cluster for resource processing; for each interactive object, determining the corresponding model calculation amount in the information prediction process; selecting a first computing node from the computing node cluster to support the call of each interactive object; for each interactive object, generating first historical resource overdue summary information and first future resource overdue summary information; generating second historical resource overdue summary information and second future resource overdue summary information; generating target resource quality information; and sending corresponding encrypted information to a target monitoring terminal to perform target warning under the condition that the resource quality warning condition is met. The implementation accurately and efficiently generates target resource quality information on the basis of reducing waste of computing resources through reasonable allocation of hardware resources.

Description

Technical Field

[0001] The embodiments disclosed herein relate to the fields of big data and artificial intelligence, and specifically to resource early warning methods, apparatuses, electronic devices, and computer-readable media based on computing node clusters. Background Technology

[0002] Currently, with the continuous development of big data and artificial intelligence technologies, applying these technologies to various industries can significantly improve business efficiency. The typical approach to determining resource overdue information for a target object is as follows: First, obtain information on the target object's various resource overdue behaviors. Then, use relevant formulas or expert analysis to determine the resource overdue information under each behavior.

[0003] However, when using the above method, the following technical problems often arise:

[0004] With the advent of the big data era, target objects possess massive amounts of behavioral data, making formulas or expert analysis methods often inefficient. While configuring more hardware resources can help identify resource expiration information, this often fails to fully utilize hardware resources for massive datasets, resulting in wasted computing resources and inaccurate resource expiration information.

[0005] The information disclosed in this background section is only intended to enhance the understanding of the background of the inventive concept, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The summary portion of this disclosure is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description portion. This summary portion is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0007] Some embodiments of this disclosure propose a resource early warning method, apparatus, electronic device, and computer-readable medium based on computing node clusters to solve one or more of the technical problems mentioned in the background section above.

[0008] In a first aspect, some embodiments of this disclosure provide a resource early warning method based on a computing node cluster, comprising: in response to receiving a resource early warning instruction for a target object, determining whether to invoke the computing node cluster for resource processing based on the average storage space occupied by the data corresponding to each object stored on the storage end; in response to determining the invocation, for each interactive object corresponding to the target object, determining the corresponding model computation volume in the information prediction process based on the historical resource overdue information sequence and the historical resource overdue cause sequence corresponding to the interactive object; selecting a first computing node from the computing node cluster that supports the invocation of each interactive object based on the model computation volume set and the interactive object information set; and for each interactive object, utilizing the corresponding historical resource overdue information sequence and the historical resource overdue cause sequence. The first computing node deploys a prediction model and online prediction resources to generate first historical resource overdue summary information and first future resource overdue summary information. Based on the historical resource overdue information sequence and historical resource overdue reason sequence corresponding to the above-mentioned target object, the second computing node selected in the above-mentioned computing node cluster generates second historical resource overdue summary information and second future resource overdue summary information. Based on the first historical resource overdue summary information set, the first future resource overdue summary information set, the above-mentioned second historical resource overdue summary information, and the above-mentioned second future resource overdue summary information, target resource quality information of the above-mentioned target object in the target time period is generated. The encrypted information corresponding to the above-mentioned target resource quality information is sent to the target monitoring terminal to issue a target alarm when the resource quality alarm conditions are met.

[0009] Secondly, some embodiments of this disclosure provide a resource early warning device based on a computing node cluster, comprising: a first determining unit configured to, in response to receiving a resource early warning instruction for a target object, determine whether to invoke the computing node cluster for resource processing based on the average storage space occupied by the data corresponding to each object stored in the storage terminal; a second determining unit configured to, in response to determining the invocation, for each interactive object corresponding to the target object, determine the corresponding model computation volume in the information prediction process based on the historical resource overdue information sequence and the historical resource overdue cause sequence corresponding to the interactive object; a selecting unit configured to, based on the model computation volume set and the interactive object information set, select a first computing node from the computing node cluster that supports invocation by each interactive object; and a first generating unit configured to, for each interactive object, determine the corresponding historical resource overdue information sequence and the historical resource overdue cause sequence. The system comprises: a first generation unit, which generates first historical resource overdue summary information and first future resource overdue summary information using the prediction model and online prediction resources deployed on the corresponding first computing node; a second generation unit, which generates second historical resource overdue summary information and second future resource overdue summary information using the second computing node selected in the computing node cluster, based on the historical resource overdue information sequence and historical resource overdue reason sequence corresponding to the target object; a third generation unit, which generates target resource quality information of the target object in the target time period based on the first historical resource overdue summary information set, the first future resource overdue summary information set, the second historical resource overdue summary information, and the second future resource overdue summary information; and a sending unit, which sends the encrypted information corresponding to the target resource quality information to the target monitoring terminal to issue a target alarm under the condition of meeting the resource quality alarm requirements.

[0010] Thirdly, some embodiments of this disclosure provide an electronic device, including: one or more processors; and a storage device having one or more programs stored thereon, such that when the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any implementation of the first aspect.

[0011] Fourthly, some embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the method as described in any implementation of the first aspect.

[0012] The above embodiments of this disclosure have the following beneficial effects: Through the resource early warning method based on computing node clusters in some embodiments of this disclosure, target resource quality information is generated accurately and efficiently by rationally allocating hardware resources, while reducing the waste of computing resources. Specifically, the reason why the relevant target resource quality information is not accurate enough and a lot of computing resources are wasted is that: with the advent of the big data era, target objects have massive amounts of behavioral data, and related formulas or expert analysis methods are often inefficient. Of course, configuring more hardware resources can help determine resource overdue information, but for massive amounts of data, hardware resources often cannot be fully utilized, resulting not only in wasting a lot of computing resources but also leading to inaccurate resource overdue information. Based on this, the resource early warning method based on computing node clusters in some embodiments of this disclosure first, in response to receiving a resource early warning instruction for a target object, determines whether to call the computing node cluster for resource processing based on the average storage space occupied by the data corresponding to each object stored on the storage end. Here, when determining resource alerts for target objects, the average storage space occupied by the data corresponding to each object is determined to decide whether to call the computing node cluster to achieve reasonable subsequent resource allocation. This determines whether to expedite resource alerts and avoid further resource losses due to slow resource quality determination when the target object has abnormal resources. Then, for each interaction object corresponding to the target object, the corresponding model computational load during information prediction is determined based on the historical resource overdue information sequence and historical resource overdue cause sequence for the interaction object. This facilitates the subsequent selection of corresponding computing nodes for relevant resource prediction and calculation, ensuring reasonable allocation of computing resources and improving the efficiency of generating resource overdue summary information. Next, in response to the determined call, based on the model computational load set and the interaction object information set, the first computing node supporting the call for each interaction object is selected from the computing node cluster to provide hardware resources for the generation of resource overdue summary information for subsequent interaction objects. Next, for each interactive object, based on the corresponding historical resource overdue information sequence and historical resource overdue cause sequence, and utilizing the prediction model and online prediction resources deployed on the corresponding first computing node, the first historical resource overdue summary information and the first future resource overdue summary information can be accurately generated while making full use of computing resources. Furthermore, based on the historical resource overdue information sequence and historical resource overdue cause sequence corresponding to the target object, and using the second computing node selected from the aforementioned computing node cluster, the second historical resource overdue summary information and the second future resource overdue summary information are generated, thus providing sufficient hardware computing resources for the target object and achieving efficient and accurate generation of resource overdue summary information.Furthermore, based on the first historical resource overdue summary information set, the first future resource overdue summary information set, the aforementioned second historical resource overdue summary information, and the aforementioned second future resource overdue summary information, the resource quality status of the target object within the target time period can be fully explained from the perspective of the target object and each interactive object, so as to accurately generate the target resource quality information of the target object within the target time period. Finally, the encrypted information corresponding to the aforementioned target resource quality information is sent to the target monitoring terminal to trigger a target alarm under the condition of meeting the resource quality alarm. Here, encryption is used to ensure the security of the transmission of the target object's resource information and avoid the leakage of the object's personal information. In addition, through the alarm of the target monitoring terminal, the resources of the target object can be adjusted in a timely manner to improve resource quality. In summary, firstly, the resource overdue status of each interactive object corresponding to the target object indirectly reflects the implicit resource quality of the target object. Based on this, the accurate generation of resource quality can be achieved based on the resource overdue status of the target object itself. In the process of judging the resource overdue status, the generation efficiency of resource overdue status can be greatly improved by allocating computing nodes. Attached Figure Description

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

[0014] Figure 1 This is a flowchart of some embodiments of the resource early warning method based on computing node clusters according to the present disclosure;

[0015] Figure 2 This is a schematic diagram of the structure of some embodiments of the resource early warning device based on computing node clusters according to the present disclosure;

[0016] Figure 3 This is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. Detailed Implementation

[0017] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0018] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

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

[0020] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

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

[0022] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0023] refer to Figure 1 The diagram illustrates a flow 100 of some embodiments of a resource early warning method based on a compute node cluster according to the present disclosure. This resource early warning method based on a compute node cluster includes the following steps:

[0024] Step 101: In response to receiving a resource warning instruction for the target object, determine whether to call the computing node cluster for resource processing based on the average storage space occupied by the data corresponding to each object stored on the storage end.

[0025] In some embodiments, in response to receiving a resource warning instruction for a target object, the executing entity (e.g., a computing device) of the resource warning method based on a computing node cluster can determine whether to invoke the computing node cluster for resource processing based on the average storage space occupied by the data corresponding to each object stored on the storage terminal. The target object can be an object whose resource quality information within a target time period needs to be determined. For example, the target object can be an individual or a related organization (e.g., a company). The resource warning instruction can be an instruction to warn of the resource status corresponding to the target object. Here, the resource warning instruction can be sent from the target monitoring terminal to the computing device to determine the resource status corresponding to the target object. The storage terminal can be a data storage terminal that stores the data volume corresponding to each object. Each object can be an interactive object corresponding to the target object. An interactive object can be an object that has various interactive behaviors with the target object. Interactive behaviors can be behaviors of interaction between objects within the target domain. For example, interactive behaviors can be value-related interactive behaviors. For example, interactive behaviors can be equipment sales behaviors or investment behaviors. In practice, the interactive objects corresponding to the target object can be determined by the interactive behaviors stored on the storage terminal. The storage terminal stores the object data of each interactive object. The average amount of data stored for each interactive object can be determined as the average storage space occupied. This average storage space represents the average amount of data for each interactive object. In practice, the format and data type of the stored data are the same for all objects on the storage side.

[0026] As an example, if the average data storage space usage is determined to be within a first numerical range, then the compute node cluster will be invoked for resource processing. If the average data storage space usage is determined to be within a second numerical range, then the compute node cluster will not be invoked for resource processing. The value within the first numerical range is greater than the value within the second numerical range.

[0027] Step 102: In response to the determination call, for each interaction object corresponding to the target object, determine the corresponding model computation amount in the information prediction process based on the historical resource overdue information sequence and historical resource overdue reason sequence corresponding to the above interaction object.

[0028] In some embodiments, in response to a determined call, the aforementioned execution entity, for each interactive object corresponding to the target object, determines the corresponding model computational load during the information prediction process based on the historical resource overdue information sequence and the historical resource overdue reason sequence corresponding to the interactive object. The target time period can be a predetermined historical time period before the current time and a predetermined future time period after the current time. Resource quality information can be information about the resource quality of the target object corresponding to the target resource. In practice, resource quality information can be numerical information or text label information. For example, for numerical information, the higher the corresponding value, the higher the resource quality. For example, in a credit scoring scenario, the target object can be a target company, and the resource quality information can be the quality information of the company's assets. A higher value corresponding to the asset quality information indicates a better asset situation for the corresponding company. For example, company assets can include: company real estate, company equipment, company cash, and company intellectual property. The historical resource overdue information sequence can be a sequence of resource overdue information corresponding to interactive behaviors within the target historical time period. There is a one-to-one correspondence between the resource overdue information in the resource overdue information sequence and the historical time points in the target historical time period. Resource overdue information can be the overdue status of the interactive object in terms of resources. For example, in a credit reporting scenario, the interaction object can be a company engaging in value exchange. The corresponding resource overdue information can be information about the resource overdue status of the value exchange company. For instance, resource overdue information could be asset overdue information or credit overdue information of the interaction company. There is a one-to-one correspondence between historical resource overdue information in the historical resource overdue information sequence and historical resource overdue cause information in the historical resource overdue cause sequence. Historical resource overdue causes can be the reasons why historical resource overdue information is generated due to the interaction object experiencing historical resource overdue situations. In practice, if there are no historical resource overdue situations at the target historical time point, the corresponding historical resource overdue cause can be empty. For example, a historical resource overdue cause could be "a natural disaster caused resource overdue situations." The information prediction process can be the process of predicting the summary information of resource overdue situations over a certain period. The summary information of resource overdue situations can be the overall summary content of resource overdue situations. The model computational load can be the model computational load of the prediction model used in the information prediction process.

[0029] As an example, first, determine the data volume corresponding to the historical resource overdue information sequence and the historical resource overdue cause sequence. Then, determine the number of model parameters corresponding to the prediction model. Finally, based on the data volume and the number of model parameters, determine the model computational cost using the model computational cost formula.

[0030] Step 102: Based on the model computation set and the interaction object information set, select the first computing node from the computing node cluster that supports the calls of each interaction object.

[0031] In some embodiments, the execution entity can select a first computing node from the computing node cluster that supports calls from each interactive object, based on the model computational load set and the interactive object information set. The computing nodes in the computing node cluster are distributed. Each computing node in the cluster is equipped with a corresponding amount of computing hardware (e.g., graphics processing unit and central processing unit). The computing resources of each computing node can be different. The computing nodes support both online and offline resources. The interactive object information can be the object information corresponding to the interactive object. For example, the interactive object information may include: interactive object behavior information, interactive object asset information, and interactive object description information. Each interactive object has corresponding interactive object information.

[0032] As an example, the aforementioned execution entity can randomly select the first computing node from the computing node cluster that supports the calls of each interactive object based on the model computational load set and the interactive object information set, so as to ensure that the computing resources that the first computing node can provide are higher than the model computational load.

[0033] In some optional implementations of certain embodiments [A1], the execution entity may select a first computing node from the computing node cluster that supports calls from each interactive object based on the model computation set and the interaction object information set, including the following steps:

[0034] The first step is to determine the object importance level corresponding to each interactive object in the aforementioned interactive object information set, thus obtaining an object importance level set. Object importance level can characterize the importance of an interactive object to a target object. For example, object importance level can characterize the degree to which an interactive object provides value to a target object. For instance, if the target object and interactive object are companies, the corresponding object importance level could be the importance of the company corresponding to the interactive object to the customer of the company corresponding to the target object. In practice, object importance level can be numerical information or tag-type information.

[0035] As an example, the aforementioned executing entity can determine the importance of each interaction object by querying an interaction object information database. This database can include interaction object information of varying importance to the target object.

[0036] The second step is to determine the precise information prediction information corresponding to each object importance level in the aforementioned object importance set, thus obtaining a precise information prediction information set. This precise information prediction information can be the required accuracy of the first historical resource overdue summary information and the first future resource overdue summary information for each interactive object, depending on its object importance level. In other words, interactive objects with different object importance levels have different precise information prediction information. In practice, interactive objects with higher object importance levels have higher precise information prediction information.

[0037] Here, by setting different information prediction accuracy based on the importance of objects, we can reduce the situation where each computing node takes a long time to predict the prediction information corresponding to the subsequent interactive objects in order to pursue accuracy, resulting in a lot of wasted computing resources.

[0038] As an example, the aforementioned executing entity can determine the accurate information prediction information corresponding to each object's importance based on the correlation table between the object's importance and the accuracy of the information prediction, thereby obtaining the accurate information prediction information set.

[0039] The third step is to determine the node status information of each computing node in the aforementioned computing node cluster at the current time, thus obtaining a node status information set. This node status information can be the running status of the computing node at the current time. In practice, node running information may include: CPU utilization and memory usage.

[0040] As an example, the aforementioned execution entity can obtain node status information from the master node to monitor each computing node in real time, thus obtaining a set of node status information.

[0041] The fourth step involves determining the first computational node supported by each interactive object through various methods, based on the aforementioned set of object importance, the aforementioned set of information prediction accuracy, and the aforementioned set of node status information.

[0042] Optionally [A2], the node status information includes: the model accuracy corresponding to the prediction model, the model prediction duration, and the node's online prediction resources. The model accuracy can be the prediction accuracy of the prediction model deployed on the computing node. In practice, prediction models with similar network structures (but with differences in simplicity and complexity) can be deployed on different computing nodes. However, the model structures of the prediction models corresponding to each computing node can be different, and their corresponding model accuracies can also be different. This design allows different computing nodes to handle computational tasks of varying complexity. In practice, the accuracy of each model corresponding to each computing node is greater than the target accuracy, ensuring the basic accuracy of the models corresponding to each computing node. In practice, the accuracy of each model can be in an arithmetic progression form. The model prediction duration can be the estimated time consumed by the model in making predictions. The node's online prediction resources can be the estimated amount of resources consumed by the node in making online predictions. In practice, the node's online prediction resources can be determined based on the deployed model structure and the size of the input data.

[0043] Optionally, the aforementioned execution entity can determine the first computing node supported by each interactive object based on the aforementioned set of object importance, the aforementioned set of information prediction accuracy, and the aforementioned set of node status information, including the following steps:

[0044] The first step is to divide the set of interaction objects into groups based on the aforementioned set of object importance, resulting in a sequence of interaction object groups. Within each interaction object group, the object importance level is the same. The importance level of the objects within the interaction object groups increases sequentially.

[0045] The second step involves determining the corresponding computational node groups for each interactive object group in the sequence, based on the corresponding model accuracy set, model prediction duration set, and online prediction resource set, in descending order of object importance. This process utilizes various methods to identify the first computational node supported by each interactive object. Each computational node group must contain at least one supporting computational node. The computational node group can consist of individual computational nodes that perform information prediction for each interactive object within the group. Here, the computational nodes within the computational node group are treated as a whole to complete the overall information prediction for the interactive object group.

[0046] In practice, after determining the matching computing node group corresponding to the interaction object group, at least one interaction object in the aforementioned interaction object group can be determined based on the master node. Here, the master node can perform a second matching based on the data storage occupancy of the interaction object and the node status of the corresponding computing node, so that each interaction object in the interaction object group can achieve effective prediction of the corresponding information. Optionally [A3], the aforementioned execution entity can determine the matching computing node group corresponding to the interaction object group in the aforementioned interaction object group sequence in descending order of object importance based on the corresponding model accuracy set, model prediction duration set, node online prediction resource quantity set, and the aforementioned information prediction accuracy information set, to obtain the first computing node supported by each interaction object, including the following steps:

[0047] The first step is to perform the generation step for the target interactive object group in the sequence of interactive object groups:

[0048] Sub-step 1: In response to determining that the target interaction object group is not the object group with the highest object importance level, at least one interaction object group preceding the target interaction object group is determined. Here, at least one interaction object group can be any interaction object group whose corresponding object importance level is higher than the corresponding object importance level of the target interaction object group and which has already undergone computation node allocation.

[0049] Sub-step 2 involves determining at least one first computing node group corresponding to the aforementioned at least one interactive object group, wherein each computing node in the computing node group supports corresponding to at least two interactive objects. There is a one-to-one correspondence between the interactive object groups in the at least one interactive object group and the first computing node groups in the at least one first computing node group.

[0050] Sub-step 3: Remove at least one of the first computing node groups from the above computing node cluster to obtain the computing node cluster after removal.

[0051] Sub-step 4: Based on the aforementioned information prediction accuracy set, select at least one computing node group from the removed computing node clusters whose model accuracy matches the target interaction object group. The model accuracy achievable by the corresponding model in each of the at least one computing node group is higher than or equal to the information prediction accuracy requirement for the target interaction object group.

[0052] As an example, firstly, the aforementioned execution entity can filter out the information prediction accuracy group corresponding to the target interaction object group from the information prediction accuracy information set. Then, it filters out the information prediction accuracy information with the highest accuracy value from the aforementioned information prediction accuracy group as the target information prediction accuracy information. Next, it filters out the set of computing node groups that support combination from the aforementioned removed computing node cluster. Then, it determines the minimum model accuracy corresponding to each computing node group in the computing node group set. Finally, it filters out at least one computing node group from the computing node group set whose minimum model accuracy is higher than or equal to the aforementioned target information prediction accuracy information.

[0053] Sub-step 5 involves removing computing node groups from at least one of the aforementioned computing node groups whose sum of online prediction resources is less than a multiple of the computing resources required for the prediction process, thus obtaining a set of computing node groups. The computing resources required for the prediction process can be the total computing resources needed to predict the information of each interactive object in the interactive object group. The sum of online prediction resources can be the sum of the online prediction resources of each computing node in the computing node group. The multiple corresponding to the sum of online prediction resources can be the sum of online prediction resources multiplied by a multiple. For example, the multiple can be 1.5. The multiple can be a value set based on historical experience. By setting a multiple, it is possible to avoid resource shortages when computing nodes in the computing node group, due to the large amount of data involved in their tasks and the potentially long execution time, suddenly receive a task of a higher priority from the master node, forcing the current task to be paused and affecting its execution.

[0054] Sub-step 6: Determine the total duration corresponding to each computing node group in the above computing node group set.

[0055] As an example, the aforementioned execution entity can determine the model prediction time with the highest model prediction time among the model prediction time groups corresponding to each computing node in the computing node group, and use this as the total time. Here, since each computing node performs information prediction in parallel, the highest model prediction time is the total prediction time of the computing node group.

[0056] The second step is to select the computing node group with the smallest total duration from the above set of computing node groups, and use it as the first computing node group corresponding to the target interaction object group.

[0057] Step 103: For each interactive object, based on the corresponding historical resource overdue information sequence and historical resource overdue reason sequence, use the prediction model deployed on the corresponding first computing node and online prediction resources to generate the first historical resource overdue summary information and the first future resource overdue summary information.

[0058] In some embodiments, the aforementioned execution entity can, for each interactive object, generate first historical resource overdue summary information and first future resource overdue summary information based on the corresponding historical resource overdue information sequence and historical resource overdue cause sequence, using the prediction model deployed on the corresponding first computing node and online prediction resources. The prediction model can be a model that predicts resource overdue summary information at a future time or a historical time. In practice, the prediction model can be a neural network model used for information prediction. For example, the prediction model can be two temporal neural network models. The first temporal neural network model is used to predict resource overdue summary information at a historical time. The second temporal neural network model is used to predict resource overdue summary information at a future time. Online prediction resources can be resource information that the first computing node can provide at the current time to support the model for online prediction. The first historical resource overdue summary information can be a summary of resource overdue information for interactive objects within a target historical time period. The first future resource overdue summary information can be a summary of resource overdue information for interactive objects within a target future time period. For example, the first historical resource overdue summary information can include: the resource overdue value of the object at a historical time period. For example, in a credit reporting scenario, the resource overdue value can characterize the severity of asset overdue. In practice, resource delinquency values ​​can range from 0 to 100. Higher values ​​indicate greater severity of delinquency and lower resource quality. In addition, the first historical resource delinquency summary information may also include: estimated delinquency time. This estimated delinquency time can be the delinquent maintenance period of the object.

[0059] As an example, the aforementioned implementing entity can input the historical resource overdue information sequence and the historical resource overdue reason sequence into the prediction model to generate the first historical resource overdue summary information and the first future resource overdue summary information using online prediction resources.

[0060] In some optional implementations of some embodiments [A4], the aforementioned execution entity can generate first historical resource overdue summary information and first future resource overdue summary information based on the corresponding historical resource overdue information sequence and historical resource overdue reason sequence, using the prediction model deployed on the corresponding first computing node and online prediction resources, including the following steps:

[0061] The first step is to combine each historical resource overdue information in the above historical resource overdue information sequence with the corresponding historical resource overdue reason to obtain a combined information sequence.

[0062] As an example, the aforementioned implementing entity can concatenate each historical resource overdue information with the corresponding historical resource overdue reason to obtain concatenated information, which can then be used as combined information.

[0063] The second step utilizes online prediction resources to input the combined information sequence into the historical prediction sub-model of the prediction model deployed on the corresponding first computing node, thereby generating the first historical resource overdue summary information. The historical prediction sub-model can be a neural network model that predicts overdue situations at historical time points. It can be a temporal neural network model, for example, an LSTM-based prediction model. In practice, the prediction model can include an encoding layer, a historical prediction sub-model, and a future prediction sub-model. The model structures of the historical and future prediction sub-models can be identical. However, during training, the encoding layer + historical prediction sub-model and the encoding layer + future prediction sub-model are trained separately. That is, the encoding layer + historical prediction sub-model and the encoding layer itself are trained based on the training dataset. Then, based on the training dataset, the model parameters corresponding to the encoding layer are fixed, and the encoding layer + historical prediction sub-model is trained. Here, the encoding layer can be a network layer that encodes resource information into vector form. In practice, the encoding layer can be a BERT pre-trained model. Alternatively, the encoding layer can be pre-trained. The historical prediction sub-model can be trained independently based on the training dataset. Similarly, the future prediction sub-model can be trained independently based on the training dataset. Here, the historical training data can include: sequences of resource delinquency information and reasons for delinquency within the current time period, and sequences of resource delinquency information within historical time periods. The training dataset can be a summary of historical resource delinquency information and corresponding historical reasons for delinquency over a given period.

[0064] Here, online prediction resources are scheduled to drive the historical prediction sub-model to generate historical resource delinquency information.

[0065] The third step involves utilizing online prediction resources to input the combined information sequence into the future prediction sub-model, which is part of the prediction model deployed on the corresponding first computing node, to generate the first summary information on future resource delinquencies. The aforementioned historical prediction sub-model and future prediction sub-model are trained on the same training set. The future prediction sub-model can be a model that predicts resource delinquency within a future time period. Further details are omitted.

[0066] In some optional implementations of certain embodiments [A5], the aforementioned historical prediction sub-model includes: a cause feature extraction layer, a resource overdue information feature extraction layer, a cause importance generation layer, a feature fusion layer, and an encoding and decoding layer. The cause feature extraction layer can be a network layer that extracts the semantic content of the cause corresponding to the overdue cause. In practice, the cause feature extraction layer can be an RNN-based network layer. Here, by inputting the resource overdue cause into the encoding layer, cause feature information is obtained. Then, the cause feature information is input into the cause feature extraction layer to obtain the cause semantic content. The resource overdue information feature extraction layer can be a network layer that extracts the semantic content of the resource overdue information. Here, the resource overdue information feature extraction layer can be an RNN-based network layer. The cause importance generation layer can be a network layer that generates the cause importance. Here, the cause importance generation layer can be a network layer based on an attention mechanism. Here, the attention mechanism can be the attention mechanism layer in the Transformer model. Here, the cause importance generation layer can determine the degree of influence of the cause feature information corresponding to the cause feature semantic content based on the resource overdue information feature information. The feature fusion layer can be a network layer that fuses the output features of the cause feature extraction layer and the resource overdue information feature extraction layer. In practice, the feature fusion layer can be a feature concatenation layer. The encoding and decoding layer can include a feature encoding layer and a feature decoding layer. Here, the encoding and decoding layer can be a seq2seq model. The encoding and decoding layer is used to predict historical resource overdue situations.

[0067] Optionally, the aforementioned executing entity may input the combined information sequence into the historical prediction sub-model included in the prediction model deployed on the corresponding first computing node to generate the first historical resource overdue summary information, including the following steps:

[0068] The first step is to perform processing steps for each combination of information in the above combined information sequence:

[0069] Sub-step 1 involves inputting the historical resource overdue information, including the combined information, into the aforementioned resource overdue information feature extraction layer to generate resource overdue information feature information. This resource overdue information feature information can characterize the semantic content of resource overdue features corresponding to the historical resource overdue information. The resource overdue information feature information can be in vector form.

[0070] Sub-step 2 involves inputting the combined information, including the historical reasons for resource delinquency, into the aforementioned reason feature extraction layer to generate reason feature information. This reason feature information characterizes the semantic content of the reasons for historical resource delinquency. The reason feature information can be in vector form.

[0071] Sub-step 3 involves inputting the aforementioned resource overdue information feature information and cause feature information into the cause importance generation layer to generate cause importance information. Cause importance information can be a value between 0 and 1; the higher the value, the more severe the impact of the cause on resource overdueness.

[0072] Sub-step 4: Input the above-mentioned resource overdue information feature information and the above-mentioned reason importance information into the above-mentioned feature fusion layer to generate feature fusion information.

[0073] The second step is to input the obtained feature fusion information sequence into the above encoding and decoding layer to generate the above first historical resource overdue summary information.

[0074] Step 104: Based on the historical resource overdue information sequence and historical resource overdue reason sequence corresponding to the above target object, use the second computing node selected in the above computing node cluster to generate the second historical resource overdue summary information and the second future resource overdue summary information.

[0075] In some embodiments, the execution entity may generate second historical resource overdue summary information and second future resource overdue summary information based on the historical resource overdue information sequence and historical resource overdue reason sequence corresponding to the target object, using the second computing node selected in the computing node cluster.

[0076] Step 105: Based on the first historical resource overdue summary information set, the first future resource overdue summary information set, the aforementioned second historical resource overdue summary information, and the aforementioned second future resource overdue summary information, generate the target resource quality information of the aforementioned target object within the target time period.

[0077] In some embodiments, the executing entity may generate target resource quality information of the target object within a target time period based on a first historical resource overdue summary information set, a first future resource overdue summary information set, the second historical resource overdue summary information set, and the second future resource overdue summary information set. The target time period may be a predetermined time period. For example, the target time period may be within a predetermined future time period.

[0078] In some optional implementations of certain embodiments [A6], the target time period includes: a target historical time period and a target future time period. The target historical time period can be a time period prior to the current time. The target future time period can be a future time period after the current time.

[0079] Optionally, the aforementioned executing entity may generate target resource quality information of the target object within the target time period based on the first historical resource overdue summary information set, the first future resource overdue summary information set, the aforementioned second historical resource overdue summary information, and the aforementioned second future resource overdue summary information, including the following steps:

[0080] The first step is to generate a third set of historical resource overdue information for the target object within the target historical time period, based on the aforementioned first set of historical resource overdue information. This third set of historical resource overdue information can include the target object's historical resource overdue status within the target historical time period. The third set of historical resource overdue information may include: the overdue summary time and the severity of the resource overdue.

[0081] As an example, firstly, for each first historical resource overdue summary, the overdue severity of the corresponding interactive object is determined based on this summary. Then, based on the closeness of the relationship between the interactive object and the target object, the overdue severity levels in the resulting set are weighted to obtain a weighted overdue severity. Finally, the historical resource overdue summary corresponding to the weighted overdue severity is retrieved from the resource overdue association table and used as the third historical resource overdue summary. The resource overdue association table serves as a mapping relationship between overdue severity and resource overdue summary information.

[0082] The second step is to generate historical predicted resource overdue summary information for the target object within the target historical time period based on the above-mentioned second historical resource overdue summary information and the above-mentioned third historical resource overdue summary information.

[0083] As an example, the aforementioned implementing entity can weight the weighted overdue severity corresponding to the second historical resource overdue summary information and the weighted overdue severity corresponding to the third historical resource overdue summary information to obtain the historical predicted resource overdue severity. Then, it can query the historical resource overdue summary information corresponding to the historical predicted resource overdue severity from the resource overdue association table, and use this as the historical predicted resource overdue summary information.

[0084] The third step involves generating, based on the aforementioned first set of future resource overdue information, third set of future resource overdue information for the target object within the aforementioned target future time period. This third set of future resource overdue information can represent the target object's future resource overdue status within the target future time period. The third set of future resource overdue information may include: the overdue summary time and the severity of the resource overdue.

[0085] As an example, firstly, for each first set of future resource overdue summary information, the overdue severity of the corresponding interactive object is determined based on this information. Then, based on the closeness of the association between the interactive object and the target object, the overdue severity levels in the resulting set are weighted to obtain a weighted overdue severity level. Finally, the future resource overdue summary information corresponding to the weighted overdue severity level is retrieved from the resource overdue association table and used as the third set of future resource overdue summary information. The resource overdue association table serves as a mapping relationship between overdue severity levels and future resource overdue summary information.

[0086] The fourth step is to generate the future predicted resource overdue summary information for the target object within the target future time period based on the above-mentioned second future resource overdue summary information and the above-mentioned third future resource overdue summary information.

[0087] As an example, the aforementioned executing entity can weight the weighted overdue severity corresponding to the second future resource overdue summary information and the weighted overdue severity corresponding to the third future resource overdue summary information to obtain the future predicted resource overdue degree. Then, it can query the future resource overdue summary information corresponding to the future predicted resource overdue degree from the resource overdue association table, and use it as the future predicted resource overdue summary information.

[0088] The fifth step is to determine the above-mentioned historical forecast resource overdue summary information and the above-mentioned future forecast resource overdue summary information as the above-mentioned target resource quality information.

[0089] Step 106: Send the encrypted information corresponding to the above target resource quality information to the target monitoring terminal so as to issue a target alarm under the condition of meeting the resource quality alarm.

[0090] In some embodiments, the aforementioned executing entity can send encrypted information corresponding to the target resource quality information to the target monitoring terminal to trigger a target alarm when the resource quality alarm condition is met. The encrypted information can be the result of encrypting the target resource quality information. In practice, the target resource quality information can be encrypted using a pre-determined encryption algorithm (e.g., AES (Advanced Encryption Standard), DES (Data Encryption Standard), 3DES (Triple DES)). The target monitoring terminal can be a terminal that monitors the resource quality corresponding to the target object. The resource quality alarm condition can be that the resource quality information meets a target condition to trigger an alarm operation. For example, when the resource quality information is lower than a target value, an alarm operation will be automatically triggered. The corresponding alarm operation can be an email alarm.

[0091] In some optional implementations of certain embodiments [A7], the encrypted information includes: quality encrypted information corresponding to the resource quality information of the first encryption method, resource overdue impact encrypted information of each interaction object on the target object of the first encryption method, and resource overdue encrypted information corresponding to the target interaction object of the first encryption method. The first encryption method can be an algorithm for encrypting information related to resource quality warnings. The quality encrypted information can be the result of encrypting the resource quality information using the first encryption method. There is a one-to-one correspondence between the interaction objects in each interaction object and the resource overdue impact encrypted information in each resource overdue impact encrypted information. The resource overdue impact encrypted information can be the result of encrypting the resource overdue impact information using the first encryption method. The resource overdue impact information can be the impact information of the resource overdue situation corresponding to the interaction object on the target object. In practice, the resource overdue impact information can be a weighted output impact value based on the first historical resource overdue summary information, the first future resource overdue summary information, and the importance of the object. Here, the higher the impact value corresponding to the resource overdue impact information, the more the overdue situation of the corresponding interaction object will affect the determination of the resource quality of the target object. The encrypted information on overdue resources can be the result of encrypting overdue resource information (including the above-mentioned second historical overdue resource summary information and the above-mentioned second future overdue resource summary information).

[0092] Optionally, the encrypted information corresponding to the target resource quality information is sent to the target monitoring terminal to issue a target alarm when the resource quality alarm condition is met, including:

[0093] The encrypted information and the decryption encryption method for the second encryption method are sent to the target monitoring terminal. The second encryption method can be different from the first encryption method. The encryption algorithms corresponding to the first and second encryption methods are different and can be dynamically set on the resource early warning terminal. The decryption encryption method can be the encrypted result after encrypting the decryption method.

[0094] Optionally, you can send a resource quality complaint by following these steps:

[0095] Step 1. Obtain the encrypted information and the encryption method sent by the target monitoring terminal.

[0096] The second step is to use the second decryption method to decrypt the above-mentioned encryption method to obtain the decryption method.

[0097] The third step is to use the above parsing method to parse the above encrypted information to obtain the above quality encrypted information, the encrypted information on the impact of each resource's expiration, and the above resource expiration encrypted information.

[0098] The fourth step involves using the first decryption method to parse the aforementioned quality encryption information, the aforementioned overdue impact encryption information for each resource, and the aforementioned overdue encryption information for each resource, respectively, to obtain the target resource quality information, the overdue impact information for each resource, and the overdue information for each resource.

[0099] The fifth step involves displaying the target resource quality information, the impact of each resource's expiration, and the resource expiration information on the resource quality expiration processing page. This page displays information related to the resource quality expiration status of the target object.

[0100] Step 6: In response to determining that quality feedback information is entered in the edit box within the page area corresponding to the aforementioned target resource quality information, at least one resource overdue impact feedback message is entered in the edit box within the at least one page area corresponding to each of the aforementioned resource overdue impact information, and resource overdue feedback information is entered in the edit box within the page area corresponding to the aforementioned resource overdue information. At least one page area can be at least one page area selected by the target object from among the page areas corresponding to each resource overdue impact information.

[0101] Step 7: Using the first encryption method and the aforementioned decryption encryption method, encrypt and combine the aforementioned quality feedback information, the aforementioned feedback information on at least one resource overdue impact, and the aforementioned resource overdue feedback information to obtain the third encrypted information, wherein the aforementioned first encryption method and the aforementioned first decryption method correspond to each other. The first decryption method can be a decryption method designed specifically for the first encryption method.

[0102] The eighth step is to send the aforementioned third encrypted information to the resource early warning terminal (i.e., the corresponding execution entity).

[0103] Optionally, the above-mentioned resource early warning terminal performs the following steps:

[0104] The first step is to obtain the third encrypted information sent by the terminal corresponding to the target object.

[0105] The second step is to decrypt the third encrypted information to obtain the quality feedback information, the feedback information on the impact of at least one overdue resource, and the feedback information on the overdue resource.

[0106] The third step involves calling the master node in the aforementioned computing node cluster to determine whether to correct the aforementioned target resource quality information and generate a correction reason based on the aforementioned quality feedback information, the aforementioned feedback information on the impact of at least one resource overdue, the aforementioned resource overdue feedback information, and the resource quality generation record corresponding to the aforementioned target object.

[0107] Here, the large language model deployed corresponding to the master node is used to determine whether to correct the above target resource quality information and generate the correction reason based on the above quality feedback information, the above at least one resource overdue impact feedback information, the above resource overdue feedback information, and the resource quality generation record corresponding to the above target object.

[0108] The fourth step, in response to confirmation, is to send the corrected resource quality information and the reasons for the correction to the aforementioned target terminal.

[0109] Here, the aforementioned "optional" aspect serves as another inventive point of this disclosure, addressing another requirement: "In resource confidentiality scenarios, how to ensure secure communication between the target and the resource early warning terminal, and how to ensure the target can effectively appeal resource quality." Based on this, this application, by setting a first encryption method and a second encryption method, can effectively ensure secure communication between the target and the resource early warning terminal. By setting corresponding page interactions, the target can appeal for each resource quality characteristic, thereby achieving an effective appeal.

[0110] The above embodiments of this disclosure have the following beneficial effects: Through the resource early warning method based on computing node clusters in some embodiments of this disclosure, target resource quality information is generated accurately and efficiently by rationally allocating hardware resources, while reducing the waste of computing resources. Specifically, the reason why the relevant target resource quality information is not accurate enough and a lot of computing resources are wasted is that: with the advent of the big data era, target objects have massive amounts of behavioral data, and related formulas or expert analysis methods are often inefficient. Of course, configuring more hardware resources can help determine resource overdue information, but for massive amounts of data, hardware resources often cannot be fully utilized, resulting not only in wasting a lot of computing resources but also leading to inaccurate resource overdue information. Based on this, the resource early warning method based on computing node clusters in some embodiments of this disclosure first, in response to receiving a resource early warning instruction for a target object, determines whether to call the computing node cluster for resource processing based on the average storage space occupied by the data corresponding to each object stored on the storage end. Here, when determining resource alerts for target objects, the average storage space occupied by the data corresponding to each object is determined to decide whether to call the computing node cluster to achieve reasonable subsequent resource allocation. This determines whether to expedite resource alerts and avoid further resource losses due to slow resource quality determination when the target object has abnormal resources. Then, for each interaction object corresponding to the target object, the corresponding model computational load during information prediction is determined based on the historical resource overdue information sequence and historical resource overdue cause sequence for the interaction object. This facilitates the subsequent selection of corresponding computing nodes for relevant resource prediction and calculation, ensuring reasonable allocation of computing resources and improving the efficiency of generating resource overdue summary information. Next, in response to the determined call, based on the model computational load set and the interaction object information set, the first computing node supporting the call for each interaction object is selected from the computing node cluster to provide hardware resources for the generation of resource overdue summary information for subsequent interaction objects. Next, for each interactive object, based on the corresponding historical resource overdue information sequence and historical resource overdue cause sequence, and utilizing the prediction model and online prediction resources deployed on the corresponding first computing node, the first historical resource overdue summary information and the first future resource overdue summary information can be accurately generated while making full use of computing resources. Furthermore, based on the historical resource overdue information sequence and historical resource overdue cause sequence corresponding to the target object, and using the second computing node selected from the aforementioned computing node cluster, the second historical resource overdue summary information and the second future resource overdue summary information are generated, thus providing sufficient hardware computing resources for the target object and achieving efficient and accurate generation of resource overdue summary information.Furthermore, based on the first historical resource overdue summary information set, the first future resource overdue summary information set, the aforementioned second historical resource overdue summary information, and the aforementioned second future resource overdue summary information, the resource quality status of the target object within the target time period can be fully explained from the perspective of the target object and each interactive object, so as to accurately generate the target resource quality information of the target object within the target time period. Finally, the encrypted information corresponding to the aforementioned target resource quality information is sent to the target monitoring terminal to trigger a target alarm under the condition of meeting the resource quality alarm. Here, encryption is used to ensure the security of the transmission of the target object's resource information and avoid the leakage of the object's personal information. In addition, through the alarm of the target monitoring terminal, the resources of the target object can be adjusted in a timely manner to improve resource quality. In summary, firstly, the resource overdue status of each interactive object corresponding to the target object indirectly reflects the implicit resource quality of the target object. Based on this, the accurate generation of resource quality can be achieved based on the resource overdue status of the target object itself. In the process of judging the resource overdue status, the generation efficiency of resource overdue status can be greatly improved by allocating computing nodes.

[0111] Further reference Figure 2 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of a resource early warning device based on a computing node cluster. These device embodiments are similar to... Figure 1 Corresponding to the method embodiments shown, this resource early warning device based on computing node clusters can be specifically applied to various electronic devices.

[0112] like Figure 2As shown, a resource early warning device 200 based on a computing node cluster includes: a first determining unit 201, a second determining unit 202, a selecting unit 203, a first generating unit 204, a second generating unit 205, a third generating unit 206, and a sending unit 207. The first determining unit 201 is configured to, in response to receiving a resource early warning instruction for a target object, determine whether to invoke the computing node cluster for resource processing based on the average storage space occupied by the data corresponding to each object stored in the storage terminal; the second determining unit 202 is configured to, in response to determining the invocation, for each interactive object corresponding to the target object, determine the corresponding model computation amount in the information prediction process based on the historical resource overdue information sequence and the historical resource overdue cause sequence corresponding to the interactive object; the selecting unit 203 is configured to, based on the model computation amount set and the interactive object information set, select a first computing node from the computing node cluster that supports invocation for each interactive object; the first generating unit 204 is configured to, for each interactive object, utilize the corresponding first computing node's pre-processing mechanism based on the corresponding historical resource overdue information sequence and the historical resource overdue cause sequence. The system uses a testing model and online resource prediction to generate first historical resource overdue summary information and first future resource overdue summary information; a second generation unit 205 is configured to generate second historical resource overdue summary information and second future resource overdue summary information based on the historical resource overdue information sequence and historical resource overdue reason sequence corresponding to the target object, using a second computing node selected from the computing node cluster; a third generation unit 206 is configured to generate target resource quality information of the target object in the target time period based on the first historical resource overdue summary information set, the first future resource overdue summary information set, the second historical resource overdue summary information, and the second future resource overdue summary information; and a sending unit 207 is configured to send the encrypted information corresponding to the target resource quality information to the target monitoring terminal to issue a target alarm under the condition of meeting the resource quality alarm.

[0113] It is understandable that the units and references described in the resource early warning device 200 based on computing node clusters are... Figure 1 The steps in the described method correspond to each other. Therefore, the operations, features, and beneficial effects described above for the method also apply to the resource early warning device 200 based on a computing node cluster and the units contained therein, and will not be repeated here.

[0114] The following is for reference. Figure 3 It shows a schematic diagram of the structure of an electronic device (e.g., an electronic device) 300 suitable for implementing some embodiments of the present disclosure. Figure 3 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0115] like Figure 3 As shown, the electronic device 300 may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. The RAM 303 also stores various programs and data required for the operation of the electronic device 300. The processing unit 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0116] Typically, the following devices can be connected to I / O interface 305: input devices 306 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 307 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 308 including, for example, magnetic tapes, hard disks, etc.; and communication devices 309. Communication device 309 allows electronic device 300 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 3 An electronic device 300 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 3 Each box shown can represent a device or multiple devices as needed.

[0117] In particular, according to some embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication device 309, or installed from storage device 308, or installed from ROM 302. When the computer program is executed by processing device 301, it performs the functions defined in the methods of some embodiments of this disclosure.

[0118] It should be noted that, in some embodiments of this disclosure, the computer-readable medium described above may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0119] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0120] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to: upon receiving a resource warning instruction for a target object, determine whether to invoke a computing node cluster for resource processing based on the average storage space occupied by the data corresponding to each object stored in the storage terminal; upon determining the invocation, for each interactive object corresponding to the target object, determine the corresponding model computational load during the information prediction process based on the historical resource overdue information sequence and historical resource overdue cause sequence corresponding to the interactive object; select a first computing node from the computing node cluster that supports the invocation of each interactive object based on the model computational load set and the interactive object information set; and for each interactive object, determine the corresponding model computational load during the information prediction process based on the corresponding historical resource overdue information sequence and historical resource overdue cause sequence. The system generates first historical resource overdue summary information and first future resource overdue summary information using the prediction model and online prediction resources deployed on the corresponding first computing node. Based on the historical resource overdue information sequence and historical resource overdue cause sequence corresponding to the target object, the system generates second historical resource overdue summary information and second future resource overdue summary information using the second computing node selected in the computing node cluster. Based on the first historical resource overdue summary information set, the first future resource overdue summary information set, the second historical resource overdue summary information, and the second future resource overdue summary information, the system generates target resource quality information for the target object in the target time period. The system sends the encrypted information corresponding to the target resource quality information to the target monitoring terminal to issue a target alarm when the resource quality alarm conditions are met.

[0121] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0122] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0123] The units described in some embodiments of this disclosure can be implemented in software or hardware. The described units can also be housed in a processor; for example, a processor can be described as including a first determining unit, a second determining unit, a selecting unit, a first generating unit, a second generating unit, a third generating unit, and a sending unit. The names of these units do not necessarily limit the specific unit itself; for example, the sending unit can also be described as "a unit that sends encrypted information corresponding to the aforementioned target resource quality information to a target monitoring terminal to issue a target alarm when the resource quality alarm condition is met."

[0124] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0125] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A resource early warning method based on a computing node cluster, comprising: In response to receiving a resource warning instruction for the target object, determine whether to call the computing node cluster for resource processing based on the average storage space occupied by the data corresponding to each object stored on the storage end; In response to the determined call, for each interaction object corresponding to the target object, the corresponding model computation amount in the information prediction process is determined based on the historical resource overdue information sequence and the historical resource overdue reason sequence corresponding to the interaction object; Based on the model computational load set and the interaction object information set, a first computing node supporting the calls of each interaction object is selected from the computing node cluster. This selection includes: determining the object importance corresponding to each interaction object information in the interaction object information set, obtaining an object importance set; determining the information prediction accuracy information corresponding to each object importance in the object importance set, obtaining an information prediction accuracy information set; determining the node state information of each computing node in the computing node cluster at the current time, obtaining a node state information set; and selecting the first computing node supporting the calls of each interaction object from the computing node cluster based on the object importance set. The method further includes, after determining the first computing node supported by each interactive object based on the set of model computational load and the set of interactive object information, the method also includes, for each interactive object, performing the following communication steps: generating node usage event information for performing resource prediction for the interactive object; communicating the node usage event information to the corresponding first computing node so that the first computing node can arrange task timing and resource scheduling, wherein the node status information includes: the model accuracy corresponding to the prediction model, the model prediction duration, and the amount of online prediction resources of the node; For each interactive object, based on the corresponding historical resource overdue information sequence and historical resource overdue reason sequence, the first historical resource overdue summary information and the first future resource overdue summary information are generated using the prediction model deployed on the corresponding first computing node and the online prediction resources. Based on the historical resource overdue information sequence and historical resource overdue reason sequence corresponding to the target object, the second computing node selected in the computing node cluster is used to generate the second historical resource overdue summary information and the second future resource overdue summary information. Based on the first historical overdue resource summary information set, the first future overdue resource summary information set, the second historical overdue resource summary information, and the second future overdue resource summary information, the target resource quality information of the target object in the target time period is generated; The encrypted information corresponding to the target resource quality information is sent to the target monitoring terminal so as to trigger a target alarm when the resource quality alarm conditions are met.

2. The method according to claim 1, wherein, The step of determining the first computing node that each interactive object can call based on the object importance set, the information prediction accuracy set, and the node status information set includes: Based on the set of object importance, the set of interactive objects is divided into objects to obtain a sequence of interactive object groups. Each interactive object in an interactive object group has the same object importance level, and the object importance levels of the interactive object groups in the sequence of interactive object groups increase sequentially. Based on the corresponding model accuracy set, model prediction duration set, node online prediction resource set, and the information prediction accuracy set, the corresponding computing node groups in the interaction object group sequence are determined in descending order of object importance level, thereby obtaining the first computing node supported by each interaction object.

3. The method according to claim 2, wherein, Based on the corresponding model accuracy set, model prediction duration set, node online prediction resource set, and the information prediction accuracy set, and in descending order of object importance, the corresponding matching computation node groups in the interaction object group sequence are determined sequentially to obtain the first computation node supported by each interaction object, including: For the target interactive object group in the sequence of interactive object groups, perform the generation steps: In response to determining that the target interactive object group is not the object group with the highest object importance level, at least one interactive object group preceding the target interactive object group is determined. Determine at least one first computing node group corresponding to the at least one interactive object group, wherein the computing nodes in the computing node group support corresponding to at least two interactive objects; Remove at least one first computing node group from the computing node cluster to obtain a computing node cluster after removal; Based on the information, predict the accurate information set, and select at least one computing node group from the removed computing node clusters whose model accuracy matches the target interaction object group. Remove computing node groups whose sum of online prediction resources for corresponding nodes is less than the number of computing resources required for the prediction process from the at least one computing node group to obtain a computing node group set. Determine the total duration corresponding to each computing node group in the computing node group set; The computing node group with the smallest total duration is selected from the set of computing node groups and used as the first computing node group corresponding to the target interaction object group.

4. The method according to claim 1, wherein, The step of generating first historical resource overdue information summary information and first future resource overdue summary information based on the corresponding historical resource overdue information sequence and historical resource overdue reason sequence, using the prediction model deployed on the corresponding first computing node and online prediction resources, includes: Each historical resource overdue information in the historical resource overdue information sequence is combined with its corresponding historical resource overdue reason to obtain a combined information sequence; Using online prediction resources, the combined information sequence is input into the historical prediction sub-models included in the prediction model deployed on the corresponding first computing node to generate the first historical resource overdue summary information; Using online prediction resources, the combined information sequence is input into the future prediction sub-model of the prediction model deployed on the corresponding first computing node to generate the first future resource overdue summary information, wherein the historical prediction sub-model and the future prediction sub-model are trained on the same training set.

5. The method according to claim 1, wherein, The target time period includes: a target historical time period and a target future time period; and The step of generating target resource quality information for the target object within a target time period based on the first historical resource overdue summary information set, the first future resource overdue summary information set, the second historical resource overdue summary information, and the second future resource overdue summary information includes: Based on the first set of historical overdue resource summary information, generate the third set of historical overdue resource summary information for the target object within the target historical time period; Based on the second historical resource overdue summary information and the third historical resource overdue summary information, generate the historical predicted resource overdue summary information of the target object within the target historical time period; Based on the first set of future resource overdue summary information, generate the third set of future resource overdue summary information for the target object within the target future time period; Based on the second future resource overdue summary information and the third future resource overdue summary information, generate the future predicted resource overdue summary information of the target object within the target future time period; The historical predicted resource overdue summary information and the future predicted resource overdue summary information are determined as the target resource quality information.

6. The method according to claim 4, wherein, The historical prediction sub-model includes: a cause feature extraction layer, a resource overdue information feature extraction layer, a cause importance generation layer, a feature fusion layer, and an encoding and decoding layer; and The step of inputting the combined information sequence into the prediction model deployed on the corresponding first computing node includes a historical prediction sub-model to generate first historical resource overdue summary information, including: For each combination of information in the combined information sequence, perform the following processing steps: The historical overdue information included in the combined information is input into the resource overdue information feature extraction layer to generate resource overdue information feature information; The combined information, including the reasons for the historical resource overdue dates, is input into the reason feature extraction layer to generate reason feature information; The resource overdue information feature information and the cause feature information are input into the cause importance generation layer to generate cause importance information; The resource overdue information feature information and the cause importance information are input into the feature fusion layer to generate feature fusion information; The obtained feature fusion information sequence is input into the encoding and decoding layer to generate the first historical resource overdue summary information.

7. A resource early warning device based on a computing node cluster, used to execute the resource early warning method based on a computing node cluster as described in claim 1, comprising: The first determining unit is configured to, in response to receiving a resource warning instruction for a target object, determine whether to call the computing node cluster for resource processing based on the average storage space occupied by the data corresponding to each object stored on the storage end. The second determining unit is configured to, in response to a determining call, determine the corresponding model computation amount in the information prediction process for each interactive object corresponding to the target object, based on the historical resource overdue information sequence and the historical resource overdue reason sequence corresponding to the interactive object. The selection unit is configured to select a first computing node from the computing node cluster that supports calls from each interactive object, based on the model computation set and the interactive object information set, wherein the selection is based on the model computation set and the interactive object information set. The first generation unit is configured to generate, for each interactive object, first historical resource overdue information summary information and first future resource overdue information summary information based on the corresponding historical resource overdue information sequence and historical resource overdue reason sequence, using the prediction model deployed on the corresponding first computing node and online prediction resources. The second generation unit is configured to generate second historical resource overdue information and second future resource overdue information based on the historical resource overdue information sequence and historical resource overdue reason sequence corresponding to the target object, using the second computing node selected in the computing node cluster. The third generation unit is configured to generate target resource quality information of the target object in the target time period based on the first historical resource overdue summary information set, the first future resource overdue summary information set, the second historical resource overdue summary information and the second future resource overdue summary information. The sending unit is configured to send encrypted information corresponding to the target resource quality information to the target monitoring terminal so as to issue a target alarm when the resource quality alarm condition is met.

8. An electronic device, comprising: One or more processors; Storage device, on which one or more programs are stored, When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-6.

9. A computer-readable medium having a computer program stored thereon, wherein, When the program is executed by the processor, it implements the method as described in any one of claims 1-6.

Images