Data reasoning method and device, computer equipment and storage medium
By distilling historical data knowledge and incremental distillation of real-time data, combined with fusion analysis of spatiotemporal characteristics, the data inference accuracy problem of wireless access networks in dynamic environments is solved, ensuring high reliability and efficient network resource utilization.
Patent Information
- Application Number
- CN202510351524.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-24
- Publication Date
- 2025-07-18
AI Technical Summary
The data inference operation of existing wireless access networks relies on historical data, making it difficult to adapt to dynamically changing network environments, resulting in a decrease in prediction accuracy.
By distilling the historical data of the target equipment, the basic data is obtained; the real-time data is incrementally distilled to obtain incremental data; the historical data, real-time data and incremental data are fusion analysis to obtain dynamic fusion data; data inference is carried out based on dynamic fusion data, real-time data and basic data to obtain the target operation strategy of the target equipment.
It realizes high-reliability data inference in dynamic network environment, improves inference accuracy and optimizes network resource utilization.
Smart Images
Figure CN120338096A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of wireless communication technologies, and particularly to a data inference method, apparatus, computer device, and storage medium. Background Art
[0002] With the rapid development of radio access networks (RANs), the amount of data in the network environment has increased exponentially, covering multiple key dimensions such as user traffic, radio channel status, and network load.
[0003] However, the data inference operations in existing radio access networks mainly rely on historical data. However, historical data is usually a static data set obtained through long-term collection. Although it can reflect certain network trends, it cannot comprehensively capture the real-time changes in network status. This also leads to difficulty in adapting to the dynamically changing network environment during the data inference process, resulting in a decrease in prediction accuracy. Summary of the Invention
[0004] Based on this, it is necessary to provide a data inference method, apparatus, computer device, and storage medium that can adapt to the dynamically changing network environment and ensure prediction accuracy for the above technical problems.
[0005] In a first aspect, this application provides a data inference method. The method includes:
[0006] Performing knowledge distillation processing on the historical data of the target device to obtain basic data;
[0007] Performing data incremental distillation processing on the real-time data of the target device to obtain incremental data;
[0008] Performing spatio-temporal characteristic fusion analysis on the historical data, the real-time data, and the incremental data to obtain dynamically fused data;
[0009] Performing data inference based on the dynamically fused data, the real-time data, and the basic data to obtain the target operation strategy of the target device.
[0010] In one embodiment, the performing spatio-temporal characteristic fusion analysis on the historical data, the real-time data, and the incremental data to obtain dynamically fused data includes:
[0011] Performing spatio-temporal feature decoupling on the historical data, the real-time data, and the incremental data to obtain the spatial topology feature and the time dynamic feature corresponding to the target device;
[0012] Performing dynamic fusion on the spatial topology feature and the time dynamic feature to obtain dynamically fused data.
[0013] In one embodiment, the dynamic fusion of the spatial topological features and the temporal dynamic features to obtain dynamic fusion data includes:
[0014] Determine a first attention influence weight corresponding to the spatial topological features and a second attention influence weight corresponding to the temporal dynamic features;
[0015] Perform weighted processing on the spatial topological features according to the first attention influence weight to obtain a first weighted value;
[0016] Perform weighted processing on the temporal dynamic features according to the second attention influence weight to obtain a second weighted value;
[0017] Determine the dynamic fusion data according to the first weighted value and the second weighted value.
[0018] In one embodiment, the data inference based on the dynamic fusion data, the real-time data, and the basic data to obtain the target operation strategy of the target device includes:
[0019] Based on the data analysis module of the Radio Access Network Artificial Intelligence Layer (RANAI Layer), perform data inference on the target device according to the dynamic fusion data, the real-time data, and the basic data to obtain the target operation strategy of the target device.
[0020] In one embodiment, the data inference based on the dynamic fusion data, the real-time data, and the basic data to obtain the target operation strategy of the target device includes:
[0021] Perform data inference on the target device according to the dynamic fusion data to obtain an initial operation strategy of the target device;
[0022] Perform operation correction on the initial operation strategy according to the real-time data and the basic data to obtain the target operation strategy of the target device.
[0023] In one embodiment, the operation correction of the initial operation strategy according to the real-time data and the basic data to obtain the target operation strategy of the target device includes:
[0024] Identify abnormal operation sub-strategies in the initial operation strategy that do not conform to the basic data and the real-time data;
[0025] Perform operation correction on the abnormal operation sub-strategies in the initial operation strategy according to the basic data to obtain the target operation strategy of the target device.
[0026] In a second aspect, the present application also provides a data inference device. The device includes:
[0027] A distillation module, configured to perform knowledge distillation processing on historical data of a target device to obtain basic data;
[0028] An increment module, configured to perform data increment distillation processing on real-time data of the target device to obtain increment data;
[0029] A fusion module, configured to perform spatio-temporal characteristic fusion analysis on the historical data, the real-time data, and the increment data to obtain dynamic fusion data;
[0030] An inference module, configured to perform data inference based on the dynamic fusion data, the real-time data, and the basic data to obtain a target operation strategy for the target device.
[0031] In a third aspect, the present application also provides a computer device. The computer device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:
[0032] Perform knowledge distillation processing on historical data of a target device to obtain basic data;
[0033] Perform data increment distillation processing on real-time data of the target device to obtain increment data;
[0034] Perform spatio-temporal characteristic fusion analysis on the historical data, the real-time data, and the increment data to obtain dynamic fusion data;
[0035] Perform data inference based on the dynamic fusion data, the real-time data, and the basic data to obtain a target operation strategy for the target device.
[0036] In a fourth aspect, the present application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:
[0037] Perform knowledge distillation processing on historical data of a target device to obtain basic data;
[0038] Perform data increment distillation processing on real-time data of the target device to obtain increment data;
[0039] Perform spatio-temporal characteristic fusion analysis on the historical data, the real-time data, and the increment data to obtain dynamic fusion data;
[0040] Perform data inference based on the dynamic fusion data, the real-time data, and the basic data to obtain a target operation strategy for the target device.
[0041] In a fifth aspect, the present application also provides a computer program product. The computer program product includes a computer program which, when executed by a processor, implements the following steps:
[0042] Perform knowledge distillation processing on the historical data of the target device to obtain basic data;
[0043] Perform data incremental distillation processing on the real-time data of the target device to obtain incremental data;
[0044] Perform spatio-temporal characteristic fusion analysis on the historical data, the real-time data, and the incremental data to obtain dynamic fusion data;
[0045] Perform data inference based on the dynamic fusion data, the real-time data, and the basic data to obtain the target operation strategy of the target device.
[0046] The above data inference method, device, computer device, and storage medium obtain basic data by performing knowledge distillation processing on historical data, and obtain incremental data by performing data incremental distillation processing on real-time data; perform spatio-temporal characteristic fusion analysis on the target device based on the historical data, real-time data, and incremental data to obtain dynamic fusion data, and then perform data inference on the target device based on the dynamic fusion data, real-time data, and basic data to obtain the target operation strategy of the target device. According to the above content, it can be seen that during the data inference process of the present application, data incremental distillation processing is performed on real-time data to achieve full-volume and low-latency data acquisition operations, preventing the subsequent data inference accuracy from being affected due to the too low data volume of real-time data. Moreover, by performing data distillation processing on historical data, the acquisition of the historical network trend of the target device is realized; furthermore, it is ensured that the dynamic fusion data for subsequent spatio-temporal characteristic fusion can effectively reflect the actual operation situation of the target device, so that when performing data inference on the target device subsequently, it can be ensured that high reliability is still maintained in a dynamic network environment, effectively improving the inference accuracy and optimizing the network resource utilization rate. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] Figure 1 It is an application environment diagram of a data inference method provided by an embodiment of the present application;
[0048] Figure 2 It is a flowchart of a first data inference method provided by an embodiment of the present application;
[0049] Figure 3 It is a flowchart of a second data inference method provided by an embodiment of the present application;
[0050] Figure 4Schematic flowchart of the third data inference method provided by an embodiment of this application;
[0051] Figure 5 Schematic flowchart of the fourth data inference method provided by an embodiment of this application;
[0052] Figure 6 Structural block diagram of a data inference device provided by an embodiment of this application;
[0053] Figure 7 Internal structure diagram of a computer device in an embodiment. Detailed implementation manners
[0054] In order to make the objectives, technical solutions and advantages of this application clearer and more understandable, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not used to limit this application.
[0055] The data inference method provided by the embodiments of this application can be applied to an application environment as Figure 1 shown. Among them, the terminal 102 communicates with the server 104 through a network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or placed in the cloud or other network servers. By performing knowledge distillation processing on historical data to obtain basic data, and performing data incremental distillation processing on real-time data to obtain incremental data; performing spatio-temporal feature fusion analysis on the target device according to the historical data, real-time data, and incremental data to obtain dynamic fusion data, and then, performing data inference on the target device according to the dynamic fusion data, real-time data, and basic data to obtain the target operation strategy of the target device. Among them, the terminal 102 can be, but is not limited to, various personal computers, laptop computers, smart phones, tablet computers, Internet of Things devices, and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart vehicle-mounted devices, etc. The portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be implemented by an independent server or a server cluster composed of multiple servers.
[0056] In one embodiment, as Figure 2 shown, a data inference method is provided. Taking the method applied to the Figure 1 server 104 as an example, the method includes the following steps:
[0057] S201, perform knowledge distillation processing on the historical data of the target device to obtain basic data.
[0058] It should be noted that before performing knowledge distillation on the historical data of the target device, the historical data of the target device can be preprocessed to reduce redundant data and abnormal data in the historical data. Then, knowledge distillation is performed on the preprocessed historical data to obtain basic data.
[0059] In an embodiment of the present application, if the target device is an automatic delivery robot, when performing knowledge distillation on the historical data of the target device, the following content may be included: The historical data of the automatic delivery robot may include environmental data and driving data; among them, the environmental data may record road conditions at different times (such as traffic congestion sections during the morning rush hour), common obstacles (temporary construction fences, randomly parked electric vehicles), and weather impacts (locations where the road surface is slippery on rainy days); the driving data may record braking force, turning angle, avoidance paths when encountering pedestrians, etc. Analyze the historical data and use data distillation technology to screen out "the most typical scenarios" and "the most representative countermeasures"; for example, the most typical scenarios may include: average traffic flow data at a certain intersection during the morning rush hour, and the critical friction coefficient value of the tile floor on rainy days; the most representative countermeasures may include: when encountering a pedestrian suddenly crossing the road, in 85% of the cases, it is only necessary to fine-tune 30 degrees to the left to avoid, and the maximum climbing angle of a certain ramp is 15 degrees, and the power mode needs to be switched when it exceeds. Finally, the distilled data is aggregated to obtain basic data or a basic data set.
[0060] S202, perform data incremental distillation on the real-time data of the target device to obtain incremental data.
[0061] It should be noted that when it is necessary to perform data incremental distillation on the target device according to the real-time data of the target device, the real-time data can be pre-filtered to obtain the filtered real-time data, and the dynamic features in the real-time data can be extracted to obtain the key features in the real-time data; then, data incremental distillation is performed on the key features to obtain incremental data.
[0062] S203, perform spatio-temporal characteristic fusion analysis on the historical data, real-time data, and incremental data to obtain dynamic fusion data.
[0063] It should be noted that when it is necessary to perform spatio-temporal characteristic fusion analysis on the target device using historical data, real-time data, and incremental data, the spatio-temporal features of the historical data, real-time data, and incremental data can be decoupled respectively to obtain the spatial topological features and time dynamic features corresponding to the target device; then, according to the spatial topological features and time dynamic features, the dynamic fusion data is determined.
[0064] S204, perform data inference based on the dynamic fusion data, real-time data, and basic data to obtain the target operation strategy of the target device.
[0065] It should be noted that when data inference is to be performed on the target device based on dynamic fusion data, real-time data, and basic data, the following content may be included: Based on the RAN AI Layer data analysis module of the radio access network artificial intelligence layer, data inference is performed on the target device according to the dynamic fusion data, real-time data, and basic data to obtain the target operation strategy of the target device.
[0066] As an example, the dynamic fusion data, real-time data, and basic data are all input into the RAN AI Layer data analysis module, and the output result of the RAN AI Layer data analysis module is obtained. This output result is the target operation strategy of the target device.
[0067] As another example, the dynamic fusion data is input into the RAN AI Layer data analysis module, and the output result of the RAN AI Layer data analysis module is obtained. This output result is the initial operation strategy. Furthermore, the initial operation strategy is corrected according to the basic data to obtain the target operation strategy.
[0068] Furthermore, when correcting the initial operation strategy according to the basic data, it can be verified whether the initial operation strategy conforms to the basic data. If it conforms, there is no need to correct the initial operation strategy, and the initial operation strategy is used as the target operation strategy. If it does not conform, the initial operation strategy is corrected according to the basic data to obtain the target operation strategy.
[0069] Further explanation, to ensure that the RAN AI Layer data analysis module can reasonably allocate computing power resources during the process of data inference, an intelligent computing power scheduling mechanism can be preset to be built inside the RAN AI Layer data analysis module, and computing resources are dynamically allocated according to business requirements and the availability of real-time data.
[0070] Specifically, when the real-time data is insufficient, the RAN AI Layer data analysis module can preferentially call the dynamic fusion data and basic data for data inference, and at the same time reasonably allocate CPU (Central Processing Unit) / GPU (Graphics Processing Unit) resources to improve the inference efficiency, reduce the network computing overhead, and ensure the low-latency inference ability.
[0071] The above data inference method obtains basic data by performing knowledge distillation on historical data and performs data incremental distillation on real-time data to obtain incremental data. It performs spatio-temporal feature fusion analysis on the target device based on historical data, real-time data, and incremental data to obtain dynamic fusion data. Furthermore, it performs data inference on the target device based on the dynamic fusion data, real-time data, and basic data to obtain the target operation strategy of the target device. According to the above content, it can be seen that during the data inference process of this application, incremental distillation processing is performed on real-time data to achieve full-volume and low-latency data acquisition operations, preventing the subsequent data inference accuracy from being affected due to the too low data volume of real-time data. Moreover, by performing data distillation on historical data, the acquisition of the historical network trend of the target device is realized. Furthermore, it ensures that the dynamic fusion data for subsequent spatio-temporal feature fusion can effectively reflect the actual operating conditions of the target device, enabling high reliability to be maintained even in a dynamic network environment during subsequent data inference on the target device, effectively improving the inference accuracy, and optimizing the network resource utilization rate.
[0072] In one embodiment, as Figure 3 shown, when it is necessary to perform spatio-temporal feature fusion analysis on the target device based on historical data, real-time data, and incremental data to obtain dynamic fusion data, the following content may be included:
[0073] S301, perform spatio-temporal feature decoupling on historical data, real-time data, and incremental data to obtain the spatial topological feature and time dynamic feature corresponding to the target device.
[0074] Among them, the spatial topological feature is used to represent the key features with spatial characteristics, such as street and building locations; the time dynamic feature is used to represent the key features with time dynamic characteristics, such as traffic flow affected by the received time, real-time weather, etc.
[0075] It should be noted that when it is necessary to perform spatio-temporal feature decoupling on historical data, real-time data, and incremental data, a first feature decoupling model for spatial topological features and a second feature decoupling model for time dynamic features can be pre-constructed. Then, the spatio-temporal feature decoupling of historical data, real-time data, and incremental data is performed through the first feature decoupling model and the second feature decoupling model to obtain the spatial topological feature and time dynamic feature corresponding to the target device.
[0076] In one embodiment of this application, historical data, real-time data, and incremental data can be input into the first feature decoupling model to achieve feature decoupling of historical data, real-time data, and incremental data through the first feature decoupling model to obtain the spatial topological feature.
[0077] In another embodiment of the present application, historical data, real-time data, and incremental data can also be input into the second feature decoupling model to achieve feature decoupling of the historical data, real-time data, and incremental data through the second feature decoupling model, and obtain time dynamic features.
[0078] Furthermore, sample data can be obtained in advance, and the first initial model and the second initial model can be trained according to the sample data, so as to construct the first feature decoupling model and the second feature decoupling model. Among them, the sample data can include sample historical data, sample real-time data, and sample incremental data; among them, the sample data can be obtained according to the device operation conditions of the reference device. To ensure the decoupling effectiveness of the first feature decoupling model and the second feature decoupling model, it is necessary to ensure that the device type of the reference device is the same as that of the target device.
[0079] Specifically, when training the first feature decoupling model, the sample data can be labeled with spatial topological features, and then the first initial model can be trained according to the sample data labeled with spatial topological features to obtain the first feature decoupling model. When training the second feature decoupling model, the sample data can be labeled with time dynamic features, and then the second initial model can be trained according to the sample data labeled with time dynamic features to obtain the second feature decoupling model.
[0080] S302, dynamically fuse the spatial topological features and the time dynamic features to obtain dynamically fused data.
[0081] It should be noted that when dynamically fusing the spatial topological features and the time dynamic features, it can specifically include the following content: determining the first attention influence weight corresponding to the spatial topological features and the second attention influence weight corresponding to the time dynamic features; performing weighted processing on the spatial topological features according to the first attention influence weight to obtain the first weighted value; performing weighted processing on the time dynamic features according to the second attention influence weight to obtain the second weighted value; determining the dynamically fused data according to the first weighted value and the second weighted value.
[0082] Among them, the first attention influence weight and the second attention influence weight are respectively used to characterize the influence degrees of the spatial topological features and the time dynamic features on the subsequent determination of the target operation strategy.
[0083] In one embodiment of the present application, the influence degrees of the spatial topological features and the time dynamic features on the subsequent determination of the target operation strategy can be respectively judged in advance according to the historical data reasoning situation of the target device, and then the first attention influence weight and the second attention influence weight can be set according to the influence degrees of the spatial topological features and the time dynamic features on the subsequent determination of the target operation strategy.
[0084] In another embodiment of the present application, the first attention influence weight and the second attention influence weight can be set according to the work experience of the operation and maintenance personnel.
[0085] The above data reasoning method realizes obtaining the spatial environment information and time information of the environment where the target device is located by acquiring the spatial topological features and time dynamic features corresponding to the target device; improves the accuracy and stability of data reasoning. Furthermore, according to the first attention influence weight and the second attention influence weight, the spatial topological features and time dynamic features are weighted, realizing the comprehensive consideration of the spatial topological features and time dynamic features, and further improving the effectiveness of subsequent data reasoning.
[0086] In one embodiment, as Figure 4 shown, when it is necessary to perform data reasoning on the target device based on the dynamically fused data, real-time data, and basic data to obtain the target operation strategy of the target device, the following content may be included:
[0087] S401, perform data reasoning on the target device according to the dynamically fused data to obtain the initial operation strategy of the target device.
[0088] It should be noted that when it is necessary to perform data reasoning on the target device according to the dynamically fused data, the dynamically fused data can be input into the RANAI Layer data analysis module. Furthermore, the operation of performing data reasoning on the target device through the RANAI Layer data analysis module is realized, and the output result of the RANAI Layer data analysis module is obtained, and this output result is the initial operation strategy of the target device.
[0089] S402, perform operation correction on the initial operation strategy according to the real-time data and basic data to obtain the target operation strategy of the target device.
[0090] It should be noted that to further ensure the operation effectiveness and operation stability of the initial operation strategy, it is necessary to ensure that each operation sub-strategy included in the initial operation strategy conforms to the actual situation of the target device. Therefore, when performing operation correction on the initial operation strategy according to the basic data, the abnormal operation sub-strategies that do not conform to the basic data and real-time data in the initial operation strategy can be identified; according to the basic data, the abnormal operation sub-strategies in the initial operation strategy are corrected to obtain the target operation strategy of the target device.
[0091] Furthermore, the dynamically fused data at the current moment can be used as the historical data of the target device at the next moment, so as to realize the collaborative optimization between data distillation and data reasoning, thereby continuously improving the authenticity of the distilled data and the accuracy of model reasoning.
[0092] The above data inference method obtains the initial operation strategy of the target device and corrects the initial operation strategy based on real-time data and basic data, ensuring high reliability in the dynamic network environment when performing data inference on the target device, effectively improving the inference accuracy, and optimizing the network resource utilization rate.
[0093] In one embodiment, as Figure 5 shown, when obtaining the target operation strategy of the target device, it may specifically include the following:
[0094] S501, Perform knowledge distillation processing on the historical data of the target device to obtain basic data.
[0095] S502, Perform data incremental distillation processing on the real-time data of the target device to obtain incremental data.
[0096] S503, Decouple the spatio-temporal features of the historical data, real-time data, and incremental data to obtain the spatial topology features and time dynamic features corresponding to the target device.
[0097] S504, Determine the first attention influence weight corresponding to the spatial topology features and the second attention influence weight corresponding to the time dynamic features.
[0098] S505, Weight the spatial topology features according to the first attention influence weight to obtain the first weighted value.
[0099] S506, Weight the time dynamic features according to the second attention influence weight to obtain the second weighted value.
[0100] S507, Determine the dynamic fusion data according to the first weighted value and the second weighted value.
[0101] S508, Perform data inference on the target device according to the dynamic fusion data to obtain the initial operation strategy of the target device.
[0102] S509, Correct the initial operation strategy according to the real-time data and basic data to obtain the target operation strategy of the target device.
[0103] In one embodiment, if the target device is an intelligent delivery robot, the RANAI Layer data analysis module can be used to perform knowledge distillation processing on the historical data of the intelligent delivery robot to obtain basic data. Moreover, through the RANAI Layer data analysis module and by using the incremental distillation mechanism, data incremental distillation processing is performed on the real-time data and historical data of the target device to obtain incremental data. The RANAI Layer data analysis module decouples the spatio-temporal features of the historical data, real-time data, and incremental data to obtain the spatial topology features and temporal dynamic features corresponding to the target device. Then, dynamic fusion is performed on the spatial topology features and temporal dynamic features to obtain dynamically fused data. Finally, the RANAI Layer data analysis module performs data inference on the target device based on the dynamically fused data, real-time data, and basic data to obtain the target delivery route of the intelligent delivery robot.
[0104] In the above data inference method, basic data is obtained by performing knowledge distillation processing on historical data, and incremental data is obtained by performing data incremental distillation processing on real-time data. Spatio-temporal feature fusion analysis is performed on the target device based on the historical data, real-time data, and incremental data to obtain dynamically fused data. Furthermore, data inference is performed on the target device based on the dynamically fused data, real-time data, and basic data to obtain the target operation strategy of the target device. According to the above content, it can be seen that during the data inference process of this application, incremental distillation processing is performed on real-time data to achieve full-volume and low-latency data acquisition operations, preventing the subsequent data inference accuracy from being affected due to the too low data volume of real-time data. Moreover, by performing data distillation processing on historical data, the acquisition of the historical network trend of the target device is realized. Furthermore, it is ensured that the dynamically fused data for subsequent spatio-temporal feature fusion can effectively reflect the actual operating conditions of the target device, enabling high reliability to be maintained during subsequent data inference on the target device in a dynamic network environment, effectively improving the inference accuracy, and optimizing the network resource utilization rate.
[0105] It should be understood that although each step in the flowcharts involved in the above embodiments is displayed in sequence according to the indication of the arrows, these steps do not necessarily need to be executed in the order indicated by the arrows. Unless clearly stated in this article, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages. These steps or stages do not necessarily need to be executed at the same time, but can be executed at different times. The execution order of these steps or stages does not necessarily need to be sequential, but can be executed alternately or in turn with at least a part of other steps or steps or stages in other steps.
[0106] Based on the same inventive concept, an embodiment of the present application further provides a data inference device for implementing the data inference method involved above. The solution provided by this device for solving problems is similar to the solution described in the above method. Therefore, the specific limitations in one or more embodiments of the data inference device provided below can refer to the limitations on the data inference method in the foregoing, and will not be repeated here.
[0107] In one embodiment, as Figure 6 shown, a data inference device is provided, including: a distillation module 10, an increment module 20, a fusion module 30, and an inference module 40, where:
[0108] The distillation module 10 is configured to perform knowledge distillation processing on the historical data of the target device to obtain basic data;
[0109] The increment module 20 is configured to perform data increment distillation processing on the real-time data of the target device to obtain increment data;
[0110] The fusion module 30 is configured to perform spatio-temporal characteristic fusion analysis on the historical data, real-time data, and increment data to obtain dynamically fused data;
[0111] The inference module 40 is configured to perform data inference based on the dynamically fused data, real-time data, and basic data to obtain the target operation strategy of the target device.
[0112] In one embodiment, the spatio-temporal characteristics of the historical data, real-time data, and increment data are decoupled to obtain the spatial topology characteristics and time dynamic characteristics corresponding to the target device;
[0113] The spatial topology characteristics and time dynamic characteristics are dynamically fused to obtain dynamically fused data.
[0114] In one embodiment, the first attention influence weight corresponding to the spatial topology characteristics and the second attention influence weight corresponding to the time dynamic characteristics are determined;
[0115] According to the first attention influence weight, the spatial topology characteristics are weighted to obtain a first weighted value;
[0116] According to the second attention influence weight, the time dynamic characteristics are weighted to obtain a second weighted value;
[0117] According to the first weighted value and the second weighted value, the dynamically fused data is determined.
[0118] In one embodiment, based on the data analysis module of the radio access network artificial intelligence layer RANAI Layer, data inference is performed on the target device according to the dynamically fused data, real-time data, and basic data to obtain the target operation strategy of the target device.
[0119] In one embodiment, data inference is performed on a target device based on dynamic fusion data to obtain an initial operation strategy for the target device;
[0120] The initial operation strategy is corrected according to real-time data and basic data to obtain a target operation strategy for the target device.
[0121] In one embodiment, an abnormal operation sub-strategy that does not conform to the basic data and real-time data in the initial operation strategy is identified;
[0122] According to the basic data, the abnormal operation sub-strategy in the initial operation strategy is corrected to obtain a target operation strategy for the target device.
[0123] The above data inference device obtains basic data by performing knowledge distillation processing on historical data, and performs data incremental distillation processing on real-time data to obtain incremental data; performs spatio-temporal characteristic fusion analysis on the target device according to the historical data, real-time data and incremental data to obtain dynamic fusion data, and then, performs data inference on the target device according to the dynamic fusion data, real-time data and basic data to obtain a target operation strategy for the target device. According to the above content, it can be seen that in the process of data inference in this application, incremental distillation processing will be performed on real-time data to achieve full-volume and low-latency data profit-making operations, preventing the subsequent data inference accuracy from being affected due to too low data volume of real-time data, and, by performing data distillation processing on historical data, the acquisition of the historical network trend of the target device is realized; furthermore, it is ensured that the dynamic fusion data for subsequent spatio-temporal characteristic fusion can effectively reflect the actual operation of the target device, so that when performing data inference on the target device subsequently, it can be ensured that it still has high reliability in a dynamic network environment, effectively improving the inference accuracy and optimizing the network resource utilization rate.
[0124] Each module in the above data inference device can be implemented in whole or in part by software, hardware and their combination. The above modules can be embedded in the processor in the computer device in hardware form or be independent of it, or can be stored in the memory in the computer device in software form, so that the processor can call and execute the operations corresponding to the above modules.
[0125] In one embodiment, a computer device is provided. The computer device can be a terminal, and its internal structure diagram can be as Figure 7As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit, and an input device. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface, the display unit, and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it realizes a data inference method. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the computer device housing, or an external keyboard, touchpad, or mouse, etc.
[0126] Those skilled in the art can understand that Figure 7 the structure shown in
[0127] is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0128] Perform knowledge distillation processing on the historical data of the target device to obtain basic data;
[0129] Perform data incremental distillation processing on the real-time data of the target device to obtain incremental data;
[0130] Perform spatio-temporal characteristic fusion analysis on the historical data, real-time data, and incremental data to obtain dynamically fused data;
[0131] Perform data inference based on the dynamically fused data, real-time data, and basic data to obtain the target operation strategy of the target device.
[0132] In one embodiment, when the processor executes the computer program, the following steps are also implemented:
[0133] Decouple the spatio-temporal features of historical data, real-time data, and incremental data to obtain the spatial topology features and time dynamic features corresponding to the target device;
[0134] Dynamically fuse the spatial topology features and time dynamic features to obtain dynamically fused data.
[0135] In one embodiment, when the processor executes the computer program, the following steps are further implemented:
[0136] Determine the first attention influence weight corresponding to the spatial topology features and the second attention influence weight corresponding to the time dynamic features;
[0137] According to the first attention influence weight, perform weighted processing on the spatial topology features to obtain a first weighted value;
[0138] According to the second attention influence weight, perform weighted processing on the time dynamic features to obtain a second weighted value;
[0139] Determine the dynamically fused data according to the first weighted value and the second weighted value.
[0140] In one embodiment, when the processor executes the computer program, the following steps are further implemented:
[0141] Based on the radio access network artificial intelligence layer RANAI Layer data analysis module, perform data inference on the target device according to the dynamically fused data, real-time data, and basic data to obtain the target operation strategy of the target device.
[0142] In one embodiment, when the processor executes the computer program, the following steps are further implemented:
[0143] Perform data inference on the target device according to the dynamically fused data to obtain the initial operation strategy of the target device;
[0144] Perform operation correction on the initial operation strategy according to the real-time data and basic data to obtain the target operation strategy of the target device.
[0145] In one embodiment, when the processor executes the computer program, the following steps are further implemented:
[0146] Identify the abnormal operation sub-strategies in the initial operation strategy that do not conform to the basic data and real-time data;
[0147] According to the basic data, perform operation correction on the abnormal operation sub-strategies in the initial operation strategy to obtain the target operation strategy of the target device.
[0148] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:
[0149] Perform knowledge distillation processing on the historical data of the target device to obtain basic data;
[0150] Perform data incremental distillation processing on the real-time data of the target device to obtain incremental data;
[0151] Perform spatio-temporal feature fusion analysis on the historical data, real-time data, and incremental data to obtain dynamically fused data;
[0152] Perform data inference based on the dynamically fused data, real-time data, and basic data to obtain the target operation strategy of the target device.
[0153] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:
[0154] Perform spatio-temporal feature decoupling on the historical data, real-time data, and incremental data to obtain the spatial topological features and temporal dynamic features corresponding to the target device;
[0155] Perform dynamic fusion on the spatial topological features and temporal dynamic features to obtain dynamically fused data.
[0156] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:
[0157] Determine the first attention influence weight corresponding to the spatial topological features and the second attention influence weight corresponding to the temporal dynamic features;
[0158] Perform weighted processing on the spatial topological features according to the first attention influence weight to obtain a first weighted value;
[0159] Perform weighted processing on the temporal dynamic features according to the second attention influence weight to obtain a second weighted value;
[0160] Determine the dynamically fused data according to the first weighted value and the second weighted value.
[0161] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:
[0162] Based on the radio access network artificial intelligence layer RANAI Layer data analysis module, perform data inference on the target device according to the dynamically fused data, real-time data, and basic data to obtain the target operation strategy of the target device.
[0163] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:
[0164] Perform data inference on the target device according to the dynamically fused data to obtain the initial operation strategy of the target device;
[0165] Perform operation correction on the initial operation strategy according to real-time data and basic data to obtain the target operation strategy of the target device.
[0166] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:
[0167] Identify abnormal operation sub-strategies in the initial operation strategy that do not conform to the basic data and real-time data;
[0168] According to the basic data, perform operation correction on the abnormal operation sub-strategies in the initial operation strategy to obtain the target operation strategy of the target device.
[0169] In one embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the following steps are implemented:
[0170] Perform knowledge distillation processing on the historical data of the target device to obtain basic data;
[0171] Perform data incremental distillation processing on the real-time data of the target device to obtain incremental data;
[0172] Perform spatio-temporal characteristic fusion analysis on the historical data, real-time data, and incremental data to obtain dynamically fused data;
[0173] Perform data inference based on the dynamically fused data, real-time data, and basic data to obtain the target operation strategy of the target device.
[0174] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:
[0175] Perform spatio-temporal feature decoupling on the historical data, real-time data, and incremental data to obtain the spatial topological features and time dynamic features corresponding to the target device;
[0176] Perform dynamic fusion on the spatial topological features and time dynamic features to obtain dynamically fused data.
[0177] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:
[0178] Determine the first attention influence weight corresponding to the spatial topological features and the second attention influence weight corresponding to the time dynamic features;
[0179] According to the first attention influence weight, perform weighted processing on the spatial topological features to obtain a first weighted value;
[0180] According to the second attention influence weight, perform weighted processing on the time dynamic features to obtain a second weighted value;
[0181] Determine the dynamically fused data according to the first weighted value and the second weighted value.
[0182] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:
[0183] Based on the data analysis module of the Radio Access Network Artificial Intelligence Layer (RANAI Layer), perform data inference on the target device according to the dynamically fused data, real-time data, and basic data to obtain the target operation strategy of the target device.
[0184] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:
[0185] Perform data inference on the target device according to the dynamically fused data to obtain the initial operation strategy of the target device;
[0186] Perform operation correction on the initial operation strategy according to the real-time data and basic data to obtain the target operation strategy of the target device.
[0187] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:
[0188] Identify the abnormal operation sub-strategies in the initial operation strategy that do not conform to the basic data and real-time data;
[0189] Perform operation correction on the abnormal operation sub-strategies in the initial operation strategy according to the basic data to obtain the target operation strategy of the target device.
[0190] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data that have been authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with the relevant laws, regulations, and standards of relevant countries and regions.
[0191] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., without limitation.
[0192] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.
[0193] The above embodiments only represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.
Claims
1. A data reasoning method, characterized in that, The method includes: Performing knowledge distillation processing on the historical data of the target device to obtain basic data; Performing data incremental distillation processing on the real-time data of the target device to obtain incremental data; Performing spatio-temporal characteristic fusion analysis on the historical data, the real-time data, and the incremental data to obtain dynamic fusion data; Performing data inference based on the dynamic fusion data, the real-time data, and the basic data to obtain the target operation strategy of the target device.
2. The method according to claim 1, wherein The performing spatio-temporal characteristic fusion analysis on the historical data, the real-time data, and the incremental data to obtain dynamic fusion data includes: Performing spatio-temporal feature decoupling on the historical data, the real-time data, and the incremental data to obtain the spatial topology feature and the time dynamic feature corresponding to the target device; Performing dynamic fusion on the spatial topology feature and the time dynamic feature to obtain dynamic fusion data.
3. The method according to claim 2, wherein The performing dynamic fusion on the spatial topology feature and the time dynamic feature to obtain dynamic fusion data includes: Determining a first attention influence weight corresponding to the spatial topology feature and a second attention influence weight corresponding to the time dynamic feature; Performing weighted processing on the spatial topology feature according to the first attention influence weight to obtain a first weighted value; Performing weighted processing on the time dynamic feature according to the second attention influence weight to obtain a second weighted value; Determining dynamic fusion data according to the first weighted value and the second weighted value.
4. The method according to claim 1, wherein The performing data inference based on the dynamic fusion data, the real-time data, and the basic data to obtain the target operation strategy of the target device includes: Based on the data analysis module of the Radio Access Network Artificial Intelligence Layer (RANAI Layer), performing data inference on the target device according to the dynamic fusion data, the real-time data, and the basic data to obtain the target operation strategy of the target device.
5. The method according to claim 1, wherein The performing data inference based on the dynamic fusion data, the real-time data, and the basic data to obtain the target operation strategy of the target device includes: Performing data inference on the target device according to the dynamic fusion data to obtain the initial operation strategy of the target device; Performing operation correction on the initial operation strategy according to the real-time data and the basic data to obtain the target operation strategy of the target device.
6. The method according to claim 5, characterized in that, The performing operation correction on the initial operation strategy according to the real-time data and the basic data to obtain the target operation strategy of the target device includes: Identifying abnormal operation sub-strategies in the initial operation strategy that do not conform to the basic data and the real-time data; Performing operation correction on the abnormal operation sub-strategies in the initial operation strategy according to the basic data to obtain the target operation strategy of the target device.
7. A data inference device, characterized in that, The device includes: A distillation module for performing knowledge distillation processing on the historical data of the target device to obtain basic data; An incremental module for performing data incremental distillation processing on the real-time data of the target device to obtain incremental data; A fusion module, configured to perform spatio-temporal feature fusion analysis on the historical data, the real-time data, and the incremental data to obtain dynamically fused data; An inference module, configured to perform data inference based on the dynamically fused data, the real-time data, and the basic data to obtain a target operation strategy for the target device.
8. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.
10. A computer program product, comprising a computer program, characterized in that, When this computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.