Time sequence task processing method and device based on multi-neighborhood hybrid cross prediction model
By using a multi-neighborhood hybrid cross-prediction model, the problems of inaccurate prediction and lagging resource allocation in traditional resource allocation methods are solved, and efficient resource allocation and optimization for phased tasks are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CSC FINANCIAL CO LTD
- Filing Date
- 2025-12-19
- Publication Date
- 2026-04-28
AI Technical Summary
Traditional resource allocation methods struggle to accurately capture and utilize the relationships between complex time-series tasks, leading to inaccurate predictions, lagging or rigid resource allocation, and an inability to achieve forward-looking and refined dynamic resource allocation.
A multi-neighborhood hybrid cross-prediction model is adopted, which extracts task vectors through a temporal encoder and an inner product cross-neural network layer, and combines multi-neighborhood cross-attention processing to predict the task processing results of the next stage of the target task, and adjusts resource allocation based on the prediction results.
It enables accurate prediction of phased task results and optimization of resource allocation, thereby improving the efficiency and reliability of task processing.
Smart Images
Figure CN121935685A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the fields of artificial intelligence and computer application technology, specifically to a time-series task processing method, apparatus, device, and storage medium based on a multi-neighborhood hybrid cross-prediction model. Background Technology
[0002] In fields such as financial asset management and engineering, which involve complex, time-series, phased tasks and corresponding resource allocation, there is a common type of long-term, phased task. The core idea behind these tasks is "closed-loop feedback, dynamic adjustment, and resource optimization." After each phase is completed, resource allocation can be adjusted to optimize the performance of the next stage. Therefore, in the process of handling phased tasks, predicting the quality of the next stage's outcome allows for efficient resource allocation, scientific decision-making, and agile risk response. Traditional resource allocation methods, such as models based on fixed rules, simple thresholds, or historical statistics of a single task, struggle to accurately capture and utilize these complex, cross-task temporal relationships, leading to inaccurate predictions, delayed resource allocation, or rigidity.
[0003] For example, in cross-market ETF asset allocation, the price fluctuations of assets in different markets influence each other, and traditional models struggle to model this non-linear cross-market contagion effect. In enterprise network management, traffic in different subdomains is interconnected, and static bandwidth allocation cannot adapt to dynamically changing global traffic patterns. Therefore, there is an urgent need for an intelligent solution that can automatically learn the interactions between tasks from time-series data and achieve forward-looking, refined dynamic resource allocation. Summary of the Invention
[0004] In view of this, embodiments of this application provide a time-series task processing method and apparatus based on a multi-neighborhood hybrid cross-prediction model. This scheme can flexibly and efficiently make accurate predictions of the phased results of phased tasks.
[0005] The time-series task processing method and apparatus based on a multi-neighborhood hybrid cross-prediction model provided in this application are implemented as follows: One aspect of this application provides a time-series task processing method based on a multi-neighborhood hybrid cross-prediction model, the method comprising: Determine the target evaluation parameter sequence corresponding to the phased target task, and the relevant evaluation parameter sequence corresponding to at least one related neighboring task. The target evaluation parameter sequence includes the task evaluation parameters corresponding to the target task in at least one historical task phase, and the relevant evaluation parameter sequence includes the task evaluation parameters corresponding to the corresponding related neighboring task in at least one historical task phase. Feature extraction is performed on the target evaluation parameter sequence to obtain the target task vector; Feature extraction is performed on the relevant evaluation parameter sequences of each related neighborhood task to obtain the relevant task vector; The prediction and evaluation results corresponding to the target task are determined based on the target task vector and at least one related task vector. The prediction and evaluation results are used to predict the quality of the task processing results in the next task stage of the target task.
[0006] In one possible implementation, feature extraction is performed on the target evaluation parameter sequence to obtain the target task vector, including: The target evaluation parameter sequence is initially encoded by a time encoder to obtain the first candidate task vector. The first candidate task vector is input into the inner product cross neural network layer for internal feature cross learning to obtain the target task vector.
[0007] In one possible implementation, feature extraction is performed on the relevant evaluation parameter sequences of each related neighborhood task to obtain a relevant task vector, including: For each relevant neighborhood task, the relevant evaluation parameter sequence is initially encoded by a time encoder to obtain the second candidate task vector; After feature concatenation of each dimension in the second candidate task vector, adaptive weighting is performed to obtain the relevant task vector that highlights the time contribution.
[0008] In one possible implementation, the prediction and evaluation result corresponding to the target task is determined based on the target task vector and at least one related task vector, including: The target task encoding vector is obtained by performing multi-neighbor cross-attention processing on the target task vector and at least one related task vector. The task encoding vector is input into the trained classification network to obtain the prediction and evaluation results corresponding to the target task.
[0009] In one possible implementation, multi-neighborhood cross-attention processing is performed based on the target task vector and at least one related task vector to obtain the task encoding vector of the target task, including: By concatenating the target task vector with at least one related task vector, a global task vector is obtained. The global task vector is processed by multi-neighbor cross-attention and multi-neighbor cross-product calculation to obtain the task encoding vector of the target task.
[0010] In one possible implementation, the method also includes: Determine the corresponding task resource adjustment plan based on the forecast and assessment results; Adjust the current task resource configuration corresponding to the target task according to the task resource adjustment plan.
[0011] In one possible implementation, the task resource configuration includes at least one task resource parameter corresponding to a task node, and the task resource adjustment scheme includes at least one adjustment weight corresponding to a task resource parameter. Adjust the current task resource configuration corresponding to the target task according to the task resource adjustment plan, including: Adjust the current task resource parameters according to the adjustment weight corresponding to each task node.
[0012] Another aspect of this application embodiment provides a time-series task processing apparatus based on a multi-neighborhood hybrid cross-prediction model, the apparatus comprising: The sequence determination module is used to determine the target evaluation parameter sequence corresponding to the phased target task and the related evaluation parameter sequence corresponding to at least one related neighboring task. The target evaluation parameter sequence includes the task evaluation parameters corresponding to the target task in at least one historical task phase, and the related evaluation parameter sequence includes the task evaluation parameters corresponding to the corresponding related neighboring task in at least one historical task phase. The first encoding module is used to extract features from the target evaluation parameter sequence to obtain the target task vector; The second encoding module is used to extract features from the relevant evaluation parameter sequences of each relevant neighborhood task to obtain the relevant task vector. The results evaluation module is used to determine the predicted evaluation result corresponding to the target task based on the target task vector and at least one related task vector. The predicted evaluation result is used to predict the quality of the task processing result in the next task stage of the target task.
[0013] In one possible implementation, the first encoding module is further used for: The target evaluation parameter sequence is initially encoded by a time encoder to obtain the first candidate task vector. The first candidate task vector is input into the inner product cross neural network layer for internal feature cross learning to obtain the target task vector.
[0014] In one possible implementation, the second encoding module is further used for: For each relevant neighborhood task, the relevant evaluation parameter sequence is initially encoded by a time encoder to obtain the second candidate task vector; After feature concatenation of each dimension in the second candidate task vector, adaptive weighting is performed to obtain the relevant task vector that highlights the time contribution.
[0015] In one possible implementation, the result evaluation module is further used for: The target task encoding vector is obtained by performing multi-neighbor cross-attention processing on the target task vector and at least one related task vector. The task encoding vector is input into the trained classification network to obtain the prediction and evaluation results corresponding to the target task.
[0016] In one possible implementation, the result evaluation module is further used for: By concatenating the target task vector with at least one related task vector, a global task vector is obtained. The global task vector is processed by multi-neighbor cross-attention and multi-neighbor cross-product calculation to obtain the task encoding vector of the target task.
[0017] In one possible implementation, the device further includes: The scheme determination module is used to determine the corresponding task resource adjustment scheme based on the prediction and evaluation results. The resource configuration module is used to adjust the current task resource configuration corresponding to the target task according to the task resource adjustment plan.
[0018] In one possible implementation, the task resource configuration includes at least one task resource parameter corresponding to a task node, and the task resource adjustment scheme includes at least one adjustment weight corresponding to a task resource parameter. The resource configuration module is further used for: Adjust the current task resource parameters according to the adjustment weight corresponding to each task node.
[0019] The electronic device provided in this application includes a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the program, it implements the method described in this application.
[0020] The computer-readable storage medium provided in this application embodiment stores a computer program thereon, which, when executed by a processor, implements the method provided in this application embodiment.
[0021] In this embodiment, a target evaluation parameter sequence corresponding to a phased target task and a related evaluation parameter sequence corresponding to at least one related neighboring task are determined. The evaluation parameter sequence of each task includes its task evaluation parameters from at least one historical task phase. Feature extraction is performed on the target evaluation parameter sequence to obtain a target task vector. Feature extraction is performed on the related evaluation parameter sequences of each related neighboring task to obtain a related task vector. Based on the target task vector and the related task vectors, a predicted evaluation result corresponding to the target task is determined. This predicted evaluation result is used to predict the quality of the task processing result in the next task phase. This embodiment can accurately evaluate the task processing result in the next phase based on the evaluation parameters of the current task and the historical phases of related neighboring tasks. Attached Figure Description
[0022] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0023] Figure 1 A flowchart illustrating a time-series task processing method based on a multi-neighborhood hybrid cross-prediction model according to an embodiment of this application is shown. Figure 2 A schematic diagram of a time-series task processing algorithm based on a multi-neighborhood hybrid cross-prediction model according to an embodiment of this application is shown. Figure 3 A flowchart illustrating another time-series task processing method based on a multi-neighborhood hybrid cross-prediction model according to an embodiment of this application is shown. Figure 4 A schematic diagram of a time-series task processing apparatus based on a multi-neighborhood hybrid cross-prediction model according to an embodiment of this application is shown. Figure 5 A schematic diagram of an electronic device according to an embodiment of this application is shown. Detailed Implementation
[0024] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the specific technical solutions of this application will be further described in detail below with reference to the accompanying drawings of the embodiments of this application. The following embodiments are used to illustrate this application, but are not intended to limit the scope of this application.
[0025] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.
[0026] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.
[0027] It should be noted that the terms "first, second, third" used in the embodiments of this application are used to distinguish similar or different objects and do not represent a specific order of objects. It can be understood that "first, second, third" can be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.
[0028] The time-series task processing method based on a multi-neighborhood hybrid cross-prediction model in this application can be executed by any electronic device, including but not limited to mobile phones, wearable devices (such as smartwatches, smart bracelets, smart glasses, etc.), tablet computers, laptops, in-vehicle terminals, PCs (Personal Computers), etc. The functions implemented by this method can be achieved by a processor in the electronic device calling program code. Of course, the program code can be stored in a computer storage medium. Therefore, the electronic device includes at least a processor and a storage medium.
[0029] The time-series task processing method based on a multi-neighborhood hybrid cross-prediction model in this application can be used in scenarios where the phased results of any phased task are automatically evaluated based on task evaluation parameters. For example, this application can be applied to historical scientific research evaluation parameters to evaluate the quality of each research group's next-stage research results in scientific research and drug development scenarios. It can also be applied to historical network evaluation parameters to evaluate the efficiency of at least one department in the next stage of work in enterprise network management scenarios. Alternatively, it can be applied to historical stock evaluation parameters to evaluate the profit and loss performance of each fund on the next trading day in portfolio investment scenarios such as funds.
[0030] The following describes in detail the time-series task processing scheme based on the multi-neighborhood hybrid cross-prediction model of this application embodiment with reference to the accompanying drawings.
[0031] Figure 1 A flowchart illustrating a time-series task processing method based on a multi-neighborhood hybrid cross-prediction model according to an embodiment of this application is shown. Figure 1 As shown, the time-series task processing method based on the multi-neighborhood hybrid cross-prediction model in this application embodiment may include the following steps S10-S13.
[0032] For ease of description, the time-series task processing method based on a multi-neighborhood hybrid cross-prediction model in this application embodiment is described using an electronic device as the execution subject. It should be understood that the execution subject in this application embodiment can also be a processor or chip in an electronic device, and this application embodiment does not impose any limitations.
[0033] Step S10: Determine the target evaluation parameter sequence corresponding to the phased target task, and the relevant evaluation parameter sequence corresponding to at least one related neighboring task.
[0034] In one possible implementation, the electronic device can determine a phased target task and at least one related neighboring task. This related neighboring task is also a phased task that allows for the phased acquisition of task results and adjustment of task evaluation parameters. The target task and the related neighboring task can be tasks with the same corresponding task resource type and the same task result type.
[0035] Optionally, the relevant neighboring tasks corresponding to the target task in different application scenarios can be determined in different ways. For example, if the target task is the enterprise network management task of the technology department, the relevant neighboring tasks may include the enterprise network management tasks of the human resources department, the marketing department, and the finance department, etc. Alternatively, if the target task is the trading task of a single exchange-traded open-ended index fund in the A-share capital market, the relevant neighboring tasks may be the trading tasks of single exchange-traded open-ended index funds in the US stock market or Hong Kong stock market that have an impact on the target task in the A-share capital market.
[0036] In some embodiments, after determining a target task and at least one related neighboring task corresponding to the target task, the electronic device can determine a target evaluation parameter sequence corresponding to the target task and a related evaluation parameter sequence corresponding to at least one related neighboring task. The target evaluation parameter sequence includes task evaluation parameters corresponding to the target task in at least one historical task stage. The related evaluation parameter sequence includes task evaluation parameters corresponding to the relevant neighboring task in at least one historical task stage.
[0037] Optionally, the content of the task evaluation parameters in the target evaluation parameter sequence and the related neighboring task evaluation parameter sequence may differ under different application scenarios, and can be determined according to the actual application scenario. Specifically, the task evaluation parameters may include task evaluation parameters for at least one task node, and these parameters may include at least one parameter used to evaluate the historical phase-specific task processing results. For example, if the target task is an enterprise network management task for the technology department, at least one task node may be an enterprise user terminal corresponding to the technology department, and the task evaluation parameters for each enterprise user terminal may include usage traffic, latency data, and packet loss rate. If the target task is a trading task for a single exchange-traded fund (ETF) in the A-share capital market, at least one task node may be at least one stock included in the ETF, and the task evaluation parameters for each stock may include the corresponding daily return, daily volatility, and average trading volume within a preset time period.
[0038] Step S11: Extract features from the target evaluation parameter sequence to obtain the target task vector.
[0039] In one possible implementation, after determining the target evaluation parameter sequence for the target task, the electronic device can encode the target evaluation parameter sequence to achieve feature extraction, thereby obtaining the target task vector corresponding to the target task. Since the target evaluation parameter sequence includes task evaluation parameters from multiple historical task stages, the extracted target task vector can capture the longitudinal temporal features of the target task.
[0040] Optionally, in the embodiments of this application, when extracting the target evaluation parameter sequence features, temporal features can be captured through a temporal encoder and an inner product cross-calculation module. That is, the electronic device can input the target evaluation parameter sequence into the temporal encoder to perform preliminary encoding of the target evaluation parameter sequence, obtaining a first candidate task vector. Then, the first candidate task vector is input into the inner product cross-calculation module to perform inner product cross-calculation on the first candidate task vector, obtaining the target task vector. The temporal encoder can use a bidirectional LSTM structure for feature encoding, that is, feature encoding is performed separately through forward LSTM and backward LSTM network structures, and the encoding results of the two network structures are concatenated to obtain the first candidate task vector. After determining the first candidate task vector, the electronic device can perform inner product cross-calculation on the first candidate task vector and itself through the inner product cross-calculation module to extract the relationships and interactions between different task nodes, obtaining the target task vector.
[0041] Specifically, the inner product cross calculation process inputs the first candidate task vector into the inner product cross neural network layer for internal feature cross learning to obtain the target task vector. p =(l p 1 ,l p 2 ... l p n Each element in the target task vector is represented by the formula... The calculation yields the result, where n is the number of neurons in the inner product cross neural network layer, N is the number of target evaluation parameters in the target evaluation parameter sequence, i is the index of the target evaluation parameter, and f is the number of neurons in the inner product cross neural network layer. i Let θ be the dense vector obtained after embedding the i-th feature domain through the embedding layer. i n These are learnable scalar weights used to measure the contribution of the i-th objective evaluation parameter to the n-th cross neuron.
[0042] Step S12: Extract features from the relevant evaluation parameter sequences of each of the relevant neighborhood tasks to obtain the relevant task vectors.
[0043] In one possible implementation, after determining the relevant evaluation parameter sequences for each related neighboring task corresponding to the target task, the electronic device can encode these sequences to extract features, thereby obtaining the relevant task vectors for each related neighboring task. Since the relevant evaluation parameter sequences include task evaluation parameters from multiple historical task stages, the extracted relevant task vectors can capture the longitudinal temporal features of the related neighboring tasks.
[0044] Optionally, when extracting the relevant evaluation parameter sequence features in this embodiment, temporal features can be captured through a temporal encoder, a vector concatenation module, and a temporal attention module. Specifically, for each relevant neighborhood task, the electronic device can input the relevant evaluation parameters into the temporal encoder, which performs preliminary encoding on the relevant evaluation parameter sequence to obtain a second candidate task vector. The second candidate task vector is then input into the vector concatenation module to perform feature concatenation on each dimension of the second candidate task vector. The concatenated features are then input into the temporal attention module for adaptive weighting to obtain a relevant task vector that highlights the temporal contribution. The temporal attention module is used to identify and focus on the most important part of the input relevant evaluation parameters, that is, the part most relevant to the stage-specific task results of the relevant neighborhood task.
[0045] For example, the adaptive weighting process is achieved through the formula The calculation yielded that, , , Let Wa, Ua, and va be the last dimension in the relevant task vector, and let Wa, Ua, and va be learnable parameters. To obtain the attention weights for time steps τ across different dimensions.
[0046] Step S13: Determine the prediction and evaluation result corresponding to the target task based on the target task vector and at least one of the related task vectors.
[0047] In one possible implementation, after determining the target task vector corresponding to the target task and the related task vectors of at least one related neighboring task, the electronic device can determine the prediction and evaluation result corresponding to the target task based on the target task vector and the at least one related task vector. Here, since the target task vector includes the longitudinal temporal features of the target task, and the related neighboring task features of the related neighboring tasks corresponding to the target task can be used to characterize the lateral features of similar tasks to the target task, the electronic device in this embodiment can determine the prediction and evaluation result corresponding to the target task based on both lateral and longitudinal features, thus obtaining a prediction and evaluation result with high accuracy.
[0048] Optionally, the prediction and evaluation results in this application embodiment are used to predict the quality of the task processing results in the next task stage of the target task. These prediction and evaluation results differ in different application scenarios. For example, if the target task is an enterprise network management task for the technology department, the prediction and evaluation results may include the terminal status of each employee's user terminal in the technology department, which may include normal, warning, emergency, etc. If the target task is a trading task for a single exchange-traded open-ended index fund in the A-share capital market, the prediction and evaluation results may include the trading status of each stock within it, which may include significant positive, positive, neutral, negative, and significant negative, etc.
[0049] In some embodiments, the electronic device determines the predicted evaluation result corresponding to the target task based on the target task vector and at least one related task vector by first performing temporal attention processing on the target task vector and at least one related task vector to obtain the task encoding vector of the target task. Then, the task encoding vector is input into a trained classification network to obtain the predicted evaluation result corresponding to the target task. The classification network can be a classifier such as a multilayer perceptron or a support vector machine.
[0050] Optionally, the temporal attention processing can be implemented through a vector concatenation module, a cross-attention module, and an outer product calculation module. Specifically, the electronic device can input the target task vector and at least one related task vector into the vector concatenation module for concatenation to obtain a global task vector. Further, the global task vector is input into the cross-attention module for multi-neighborhood cross-attention processing, and then the cross-attention-processed global task vector is input into the outer product calculation module for multi-neighborhood outer product cross-calculation to obtain the task encoding vector of the target task. The cross-attention processing is used to calculate the mutual influence weights between different tasks, capturing the linkage between different types of tasks. For example, in the case where the target task is the trading of a single exchange-traded open-ended index fund in the A-share capital market, the cross-attention module is used to calculate the mutual influence weights between single exchange-traded open-ended index funds in different markets, capturing the contagion effect of investment sentiment and the linkage between capital flows in different markets.
[0051] For example, the above multi-domain outer product cross calculation is performed using the formula The implementation is defined as follows: where ht is the input target task vector, d is the number of feature dimensions of the target task vector, M is the number of relevant neighboring tasks of the target task, hi is the relevant task vector of the i-th relevant neighboring task, zt is the calculated task encoding vector, and Wq and Wk are pre-defined learnable parameter matrices. This is the outer product operator. The outer product operation yields a outer product matrix Pi, where the elements of the outer product matrix are Pi[m,n]=ht[m]. hi[n]; the outer product matrix explicitly captures the pairwise second-order interactions between the target task and the i-th related neighboring task across all feature dimensions, which is key to capturing complex nonlinear relationships; Flatten() is a tensor vectorization operation used to convert the outer product matrix of size d×d into a vector of length d2, which is used to concatenate it with the target task vector ht.
[0052] Figure 2 This diagram illustrates a time-series task processing algorithm based on a multi-neighborhood hybrid cross-prediction model according to an embodiment of this application. Figure 2 As shown in the embodiments of this application, the above modules can be integrated into a multi-neighborhood hybrid cross-prediction model. This model includes a backbone network for determining the prediction and evaluation results of the target task, and branch networks for extracting features from relevant neighboring tasks. The electronic device can input the target evaluation parameter sequence and relevant evaluation parameter sequences into the backbone network and branch networks respectively for feature extraction. The branch networks then input the feature extraction results back into the backbone network, which, together with the feature extraction results of the target evaluation parameter sequence, determines the prediction and evaluation results for the next stage of the target task.
[0053] In other embodiments, the electronic device can further adjust the task resource configuration of the target task based on the prediction and evaluation results to optimize the target task processing results in the next stage. That is, the electronic device can determine the corresponding task resource adjustment scheme based on the prediction and evaluation results, and then adjust the current task resource configuration corresponding to the target task according to the task resource adjustment scheme. Optionally, the task resource configuration includes task resource parameters corresponding to at least one task node, and the task resource adjustment scheme includes adjustment weights corresponding to at least one task resource parameter. Adjusting the current task resource configuration corresponding to the target task according to the task resource adjustment scheme can be achieved by adjusting the current task resource parameters according to the adjustment weights corresponding to each task node.
[0054] For example, when the target task is an enterprise network management task for the technology department, at least one task node can be a user terminal of a technology department employee, and the corresponding task resource parameter can be the bandwidth allocated to the technology department employee user terminal. Relatively speaking, the task resource adjustment scheme can include an adjustment weight for the bandwidth of each employee user terminal, and the electronic device can reallocate the bandwidth of the employee user terminal by calculating the product of this adjustment weight and the current bandwidth of the employee user terminal. When the target task is a trading task for a single exchange-traded fund (ETF) in the A-share capital market, at least one task node can be one of the included stocks, and the corresponding task resource parameter can be the proportion of that stock in the single ETF. Relatively speaking, the task resource adjustment scheme can include an adjustment weight for the proportion of each stock, and the electronic device can reallocate the proportion of the stock in the single ETF by calculating the product of this adjustment weight and the proportion of each stock.
[0055] Figure 3 A flowchart illustrating another time-series task processing method based on a multi-neighborhood hybrid cross-prediction model according to an embodiment of this application is shown. Figure 3 As shown, in this embodiment of the application, when the electronic device processes a phased task, it can first execute step S30 to obtain task processing data. This task processing data may include a sequence of target evaluation parameters corresponding to the target task, and a sequence of relevant evaluation parameters corresponding to at least one related neighboring task associated with the target task. Then, step S31 is executed to determine a prediction evaluation result based on the obtained task processing data using a multi-domain hybrid cross-prediction model. This prediction evaluation result is used to predict the quality of the task processing result in the next phase of the target task. Further, the electronic device executes step S32 to determine a task resource adjustment scheme corresponding to the prediction evaluation result. Then, step S33 is executed to adjust the current task resource configuration corresponding to the target task according to the task resource adjustment scheme, thereby optimizing the task processing result in the next phase.
[0056] Based on the aforementioned technical features, this application embodiment can obtain horizontally related and vertically related feature information to the next stage task processing result based on the evaluation parameters of the current task and related neighboring tasks in historical stages, thus accurately evaluating the next stage task processing result. Furthermore, it can adjust task resource allocation based on the predicted quality of the next stage task processing result, optimizing the next stage task processing result. Therefore, this application embodiment can automatically and efficiently optimize phased tasks based on prediction results, improving the efficiency and reliability of the entire target task processing process.
[0057] It should be understood that although the steps in the above flowcharts are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the above flowcharts may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.
[0058] Based on the foregoing embodiments, this application provides a time-series task processing device based on a multi-neighborhood hybrid cross-prediction model. The device includes various modules and units included in each module, which can be implemented by a processor; of course, it can also be implemented by specific logic circuits. In the implementation process, the processor can be a central processing unit (CPU), microprocessor (MPU), digital signal processor (DSP), or field-programmable gate array (FPGA), etc.
[0059] Figure 4 This diagram illustrates a time-series task processing apparatus based on a multi-neighborhood hybrid cross-prediction model according to an embodiment of this application. Figure 4 As shown, the time-series task processing apparatus based on a multi-neighborhood hybrid cross-prediction model in this application includes: The sequence determination module 40 is used to determine the target evaluation parameter sequence corresponding to the phased target task and the related evaluation parameter sequence corresponding to at least one related neighboring task. The target evaluation parameter sequence includes the task evaluation parameters corresponding to the target task in at least one historical task phase, and the related evaluation parameter sequence includes the task evaluation parameters corresponding to the corresponding related neighboring task in at least one historical task phase. The first encoding module 41 is used to extract features from the target evaluation parameter sequence to obtain the target task vector; The second encoding module 42 is used to extract features from the relevant evaluation parameter sequences of each relevant neighborhood task to obtain the relevant task vector; The result evaluation module 43 is used to determine the prediction evaluation result corresponding to the target task based on the target task vector and at least one related task vector. The prediction evaluation result is used to predict the quality of the task processing result in the next task stage of the target task.
[0060] In one possible implementation, the first encoding module 40 is further configured to: The target evaluation parameter sequence is initially encoded by a time encoder to obtain the first candidate task vector. The first candidate task vector is input into the inner product cross neural network layer for internal feature cross learning to obtain the target task vector.
[0061] In one possible implementation, the second encoding module 41 is further configured to: For each relevant neighborhood task, the relevant evaluation parameter sequence is initially encoded by a time encoder to obtain the second candidate task vector; After feature concatenation of each dimension in the second candidate task vector, adaptive weighting is performed to obtain the relevant task vector that highlights the time contribution.
[0062] In one possible implementation, the result evaluation module 42 is further used for: The target task encoding vector is obtained by performing multi-neighbor cross-attention processing on the target task vector and at least one related task vector. The task encoding vector is input into the trained classification network to obtain the prediction and evaluation results corresponding to the target task.
[0063] In one possible implementation, the result evaluation module 42 is further used for: By concatenating the target task vector with at least one related task vector, a global task vector is obtained. The global task vector is processed by multi-neighbor cross-attention and multi-neighbor cross-product calculation to obtain the task encoding vector of the target task.
[0064] In one possible implementation, the device further includes: The scheme determination module is used to determine the corresponding task resource adjustment scheme based on the prediction and evaluation results. The resource configuration module is used to adjust the current task resource configuration corresponding to the target task according to the task resource adjustment plan.
[0065] In one possible implementation, the task resource configuration includes at least one task resource parameter corresponding to a task node, and the task resource adjustment scheme includes at least one adjustment weight corresponding to a task resource parameter. The resource configuration module is further used for: Adjust the current task resource parameters according to the adjustment weight corresponding to each task node.
[0066] The descriptions of the above device embodiments are similar to those of the above method embodiments, and have similar beneficial effects. For technical details not disclosed in the device embodiments of this application, please refer to the descriptions of the method embodiments of this application for understanding.
[0067] It should be noted that, in the embodiments of this application... Figure 4The module division of the time-series task processing device based on the multi-neighborhood hybrid cross-prediction model shown is illustrative and represents only one logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, exist as separate physical units, or be integrated into one unit with two or more units. The integrated units can be implemented in hardware, as software functional units, or a combination of both.
[0068] It should be noted that, in the embodiments of this application, if the above-described methods are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an electronic device to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), magnetic disks, or optical disks. Thus, the embodiments of this application are not limited to any specific hardware and software combination.
[0069] Figure 5 A schematic diagram of an electronic device according to an embodiment of this application is shown. For example... Figure 5 As shown in the figure, this application provides an electronic device, which can be a server, and its internal structure diagram can be as follows. Figure 5 As shown, the electronic device includes a processor 520, a memory, and a transceiver 540 connected via a system bus 510. The processor 520 provides computing and control capabilities. The memory includes a non-volatile storage medium 531 and internal memory 532. The non-volatile storage medium 531 stores an operating system, computer programs, and a database. The internal memory 532 provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium 531. The database stores data. The transceiver 540 communicates with external terminals via a network connection. When the computer program is executed by the processor 520, it implements the methods described above.
[0070] This application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor 520, implements the steps of the method provided in the above embodiments.
[0071] This application provides a computer program product containing instructions that, when run on a computer, cause the computer to perform the steps in the method provided in the above-described method embodiments.
[0072] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the electronic device to which the present application is applied. The specific electronic device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0073] In one possible implementation, the shooting prompting device provided in this application can be implemented as a computer program, which can be configured as follows: Figure 5 The device operates on the electronic device shown. The memory of the electronic device can store various program modules that make up the above-described apparatus. The computer program composed of the various program modules causes the processor 520 to execute the steps of the methods in the various embodiments of this application described in this specification.
[0074] It should be noted that the descriptions of the storage medium and device embodiments above are similar to those of the method embodiments above, and have similar beneficial effects. For technical details not disclosed in the storage medium, storage medium, and device embodiments of this application, please refer to the descriptions of the method embodiments of this application for understanding.
[0075] It should be understood that the phrases "one embodiment," "an embodiment," or "some embodiments" mentioned throughout the specification mean that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, phrases such as "in one possible implementation," "in one embodiment," or "in some embodiments" appearing throughout the specification do not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. It should be understood that in the various embodiments of this application, the sequence numbers of the above-described processes do not imply a sequential order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. The sequence numbers of the above-described embodiments are merely descriptive and do not represent the superiority or inferiority of the embodiments. The descriptions of the various embodiments above tend to emphasize the differences between the various embodiments; their similarities or commonalities can be referred to mutually, and for the sake of brevity, they will not be repeated here.
[0076] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that there can be three kinds of relationships. For example, object A and / or object B can represent three situations: object A exists alone, object A and object B exist simultaneously, and object B exists alone.
[0077] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.
[0078] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The embodiments described above are merely illustrative. For example, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple modules or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or modules can be electrical, mechanical, or other forms.
[0079] The modules described above as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules. They may be located in one place or distributed across multiple network units. Some or all of the modules may be selected to achieve the purpose of this embodiment according to actual needs.
[0080] In addition, each functional module in the various embodiments of this application can be integrated into one processing unit, or each module can be a separate unit, or two or more modules can be integrated into one unit; the integrated modules can be implemented in hardware or in the form of hardware plus software functional units.
[0081] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as mobile storage devices, read-only memory (ROM), magnetic disks, or optical disks.
[0082] Alternatively, if the integrated units described above are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an electronic device to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROMs, magnetic disks, or optical disks.
[0083] The methods disclosed in the several method embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments.
[0084] The features disclosed in the several product embodiments provided in this application can be arbitrarily combined without conflict to obtain new product embodiments.
[0085] The features disclosed in the several method or device embodiments provided in this application can be arbitrarily combined without conflict to obtain new method or device embodiments.
[0086] The above description is merely an embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A time-series task processing method based on a multi-neighborhood hybrid cross-prediction model, characterized in that, The method includes: Determine the target evaluation parameter sequence corresponding to the phased target task, and the relevant evaluation parameter sequence corresponding to at least one related neighboring task. The target evaluation parameter sequence includes the task evaluation parameters corresponding to the target task in at least one historical task phase, and the relevant evaluation parameter sequence includes the task evaluation parameters corresponding to the corresponding related neighboring task in at least one historical task phase. Feature extraction is performed on the target evaluation parameter sequence to obtain the target task vector; Feature extraction is performed on the relevant evaluation parameter sequences of each of the relevant neighborhood tasks to obtain the relevant task vectors; The prediction and evaluation result corresponding to the target task is determined based on the target task vector and at least one of the related task vectors. The prediction and evaluation result is used to predict the quality of the task processing result in the next task stage of the target task.
2. The method according to claim 1, characterized in that, The step of extracting features from the target evaluation parameter sequence to obtain the target task vector includes: The target evaluation parameter sequence is initially encoded by a time encoder to obtain the first candidate task vector. The first candidate task vector is input into the inner product cross neural network layer for internal feature cross learning to obtain the target task vector.
3. The method according to claim 1, characterized in that, The step of extracting features from the relevant evaluation parameter sequences of each of the relevant neighborhood tasks to obtain the relevant task vector includes: For each relevant neighborhood task, the relevant evaluation parameter sequence is initially encoded by a time encoder to obtain the second candidate task vector; After feature concatenation of each dimension in the second candidate task vector, adaptive weighting is performed to obtain a relevant task vector that highlights the contribution of time.
4. The method according to claim 1, characterized in that, The step of determining the prediction and evaluation result corresponding to the target task based on the target task vector and at least one of the related task vectors includes: Multi-neighbor cross-attention processing is performed on the target task vector and at least one of the related task vectors to obtain the task encoding vector of the target task. The task encoding vector is input into the trained classification network to obtain the prediction and evaluation results corresponding to the target task.
5. The method according to claim 4, characterized in that, The step of performing multi-neighborhood cross-attention processing based on the target task vector and at least one of the related task vectors to obtain the task encoding vector of the target task includes: By concatenating the target task vector and at least one of the related task vectors, a global task vector is obtained; The global task vector is subjected to multi-neighbor cross-attention processing and multi-neighbor cross-product calculation to obtain the task encoding vector of the target task.
6. The method according to claim 1, characterized in that, The method further includes: Based on the predicted and evaluated results, a corresponding task resource adjustment plan will be determined. Adjust the current task resource configuration corresponding to the target task according to the task resource adjustment scheme.
7. The method according to claim 6, characterized in that, The task resource configuration includes at least one task resource parameter corresponding to a task node, and the task resource adjustment scheme includes at least one adjustment weight corresponding to the task resource parameter. The step of adjusting the current task resource configuration corresponding to the target task according to the task resource adjustment scheme includes: Adjust the current task resource parameters according to the adjustment weight corresponding to each task node.
8. A time-series task processing device based on a multi-neighborhood hybrid cross-prediction model, characterized in that, The device includes: The sequence determination module is used to determine the target evaluation parameter sequence corresponding to the phased target task and the related evaluation parameter sequence corresponding to at least one related neighboring task. The target evaluation parameter sequence includes the task evaluation parameters corresponding to the target task in at least one historical task phase, and the related evaluation parameter sequence includes the task evaluation parameters corresponding to the corresponding related neighboring task in at least one historical task phase. The first encoding module is used to extract features from the target evaluation parameter sequence to obtain the target task vector; The second encoding module is used to extract features from the relevant evaluation parameter sequences of each of the relevant neighborhood tasks to obtain the relevant task vectors. The result evaluation module is used to determine the prediction evaluation result corresponding to the target task based on the target task vector and at least one of the related task vectors. The prediction evaluation result is used to predict the quality of the task processing result in the next task stage of the target task.
9. An electronic device comprising a memory and a processor, the memory storing a computer program executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method according to any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 7.