AI server intelligent data analysis method based on edge-cloud collaboration

Abstract

The invention discloses an AI server intelligent data analysis method based on edge-cloud collaboration, and relates to the technical field of edge-cloud computing collaboration, edge equipment is deployed at an edge end, original data of each task is collected in real time, the data is preprocessed, task difficulty features are extracted, and a task difficulty feature sequence is formed; and based on the task difficulty feature sequence and a pre-trained AI data analysis model, the task difficulty of the collected task data is evaluated in combination with the current resource condition, and the feasibility of task execution at the edge end is judged. According to the method, equipment is deployed at the edge end and data is preprocessed, so that part of computing tasks are effectively unloaded to the edge end, the computing burden of a cloud AI server is remarkably reduced, the edge end can independently complete simple tasks, complex tasks are uploaded to the cloud according to needs, excessive centralized use of cloud resources is avoided, and the computing efficiency is improved. Therefore, the energy consumption and the operation cost of the cloud server are reduced, and the energy efficiency ratio of the whole system is improved.

Description

Technical Field

[0001] The present invention relates to the field of edge-cloud collaborative technologies, and particularly to an intelligent data analysis method for an AI server based on edge-cloud collaboration. Background Art

[0002] With the rapid development of the Internet and the Internet of Things, the amount of data generated is increasing day by day. Traditional data processing methods can no longer meet the requirements for efficient and real-time data analysis. Various industries, especially in the fields of finance, e-commerce, healthcare, manufacturing, etc., are facing the problem of huge and complex amounts of data. In recent years, the field of artificial intelligence has made many major breakthroughs, providing powerful technical support for intelligent data analysis. Through the high-performance processors, large-capacity memory, and high-speed storage devices equipped in AI intelligent data analysis, it can quickly process massive amounts of data, helping enterprises to mine valuable information from the massive data and providing a scientific basis for enterprise decision-making.

[0003] In the prior art, AI servers need to process massive amounts of data, which requires extremely high computing resources, resulting in huge energy consumption and increasing operating costs. Therefore, how to use edge-cloud collaborative technologies to offload some computing tasks to edge devices, reduce the computing burden on cloud AI servers, and thus reduce the computing pressure and energy consumption of cloud AI servers is the problem to be solved by the present invention. For this purpose, an intelligent data analysis method for an AI server based on edge-cloud collaboration is proposed. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent data analysis method for an AI server based on edge-cloud collaboration to solve the problems raised in the above background art.

[0005] To solve the above technical problems, the technical solution adopted by the present invention is: An intelligent data analysis method for an AI server based on edge-cloud collaboration, comprising the following steps: S1. Deploy edge devices at the edge end, collect the original data of each task in real time, preprocess the data, extract task difficulty features, and form a task difficulty feature sequence; S2. Based on the task difficulty feature sequence and a pre-trained AI data analysis model, evaluate the task difficulty of the collected task data in combination with the current resource status, and judge the feasibility of performing the task at the edge end; S3. Combine the task difficulty evaluation results, divide the tasks into cloud tasks and edge tasks, and allocate and manage different tasks; S4. Formulate an offloading strategy in combination with the computing power of the edge end, network status, and cloud load; S5. The edge end receives the task, quickly processes it using local computing resources, obtains a preliminary result, and uploads it to the cloud; S6. The cloud receives the task processing results uploaded by the edge side, analyzes the received offloading tasks, and then integrates the task processing results of the edge side and the cloud to generate a final data analysis report.

[0006] A further improvement of the technical solution of the present invention lies in that: the S1 specifically includes: Based on the application scenario requirements, plan the types, quantities, and deployment locations of the edge devices required by the edge side, select edge devices with corresponding computing, storage, and communication capabilities, deploy the edge devices at the edge side to ensure their normal connection to the system network, initialize the parameters of the edge devices, configure the data collection tasks and frequencies, and make them in a standby state; The deployed edge devices continuously collect the raw data generated by each task according to preset rules and frequencies. The data sources include various sensors, device interfaces, or other data sources, etc. During the collection process, monitor the data quality and mark the missing and abnormal data; Preprocess the collected raw data on the edge devices. The preprocessing includes data formatting, data cleaning, and data normalization. Among them, unify the structure and format of the data through data formatting, remove noise, duplicate data, and unreasonable data points through data cleaning, and adjust the data to a unified dimension range through data normalization. Then, perform feature analysis on the preprocessed data and extract task difficulty features reflecting the task difficulty, including the data volume size, data update frequency, data value density, and computing resource requirements; Integrate the extracted task difficulty features, set reference values for each task difficulty feature according to the preset task division criteria between the edge side and the cloud, and then form a task difficulty feature sequence, and store it in the local storage system of the edge device.

[0007] A further improvement of the technical solution of the present invention lies in that: the S2 specifically includes: Read the task difficulty feature sequence from the local storage of the edge device. At the same time, load the pre-trained AI data analysis model and obtain the current resource status of the edge side in real time, including CPU usage, remaining memory, and storage space; Input the task difficulty features in the task difficulty feature sequence into the AI data analysis model, analyze and process the input task difficulty features, and output the task difficulty evaluation result of the collected task data. Among them, the AI data analysis model receives the input task difficulty features and outputs a task difficulty score. The task difficulty score is from 1 to 10 points, divides the task difficulty into different task difficulty levels, namely simple tasks, medium-difficulty tasks, and complex tasks, and determines the score range of each task difficulty level; Based on the task difficulty evaluation result, combined with the characteristics and historical data of different types of tasks, judge the feasibility of the edge side to perform tasks.

[0008] A further improvement of the technical solution of the present invention lies in that: the calculation process of the task difficulty score is as follows: Based on the task difficulty feature sequence, calculate the total volume of task data to obtain the data volume size, determine the update frequency of task data, which represents the dynamics of the data, calculate the proportion of valid information in the data to obtain the data value density, determine the degree of demand for computing resources by the task, and determine the reference values of each task difficulty feature; Calculate the ratio of the feature value of each task difficulty feature to its reference value respectively, and square each ratio; Perform a weighted sum of the squared ratios. Among them, the weights of each task difficulty feature are equal, all being 1 / 4. Then calculate the square root of the weighted sum and multiply it by 10 to obtain the final task difficulty score.

[0009] A further improvement of the technical solution of the present invention lies in that: the specific steps of S3 include: According to the task difficulty assessment result, combined with the resource characteristics and processing capacity differences between the cloud and the edge, formulate a standard for dividing tasks into cloud tasks and edge tasks. Considering the requirements of tasks for computing resources, storage resources, as well as the requirements for real-time performance and security, among which, simple tasks are divided into edge tasks, and complex tasks are divided into cloud tasks. For medium-difficulty tasks, tasks with high computing requirements, non-real-time requirements, and relatively low data security sensitivity are divided into cloud tasks, and tasks with low computing requirements, high real-time requirements, and need to be processed quickly locally are divided into edge tasks; According to the formulated division standard, classify all tasks to be processed, assign tasks that meet the characteristics of cloud tasks to the cloud task set, and divide tasks that meet the characteristics of edge tasks into the edge task set. Edge tasks emphasize local processing and quick response to reduce data transmission delay; For the divided cloud tasks and edge tasks, establish a task management system and formulate a task allocation strategy to uniformly manage and monitor cloud tasks and edge tasks.

[0010] A further improvement of the technical solution of the present invention lies in that: the process of formulating the task allocation strategy is as follows: Design and build a task management system with functions of task classification, resource scheduling, and monitoring management, support the collaborative work of the cloud and the edge, and ensure that tasks are allocated to the corresponding computing nodes according to the classification results; According to the classification results of the tasks, formulate a detailed task allocation strategy. Based on the formulated task allocation strategy, allocate cloud tasks and edge tasks to the corresponding computing nodes. Cloud tasks are sent to the cloud server for centralized processing, and edge tasks are allocated to edge devices for local processing. At the same time, monitor the execution status of tasks in real time, and detect and handle abnormal situations during task execution; Unify the monitoring of cloud and edge tasks, collect the execution status, resource usage, and performance metrics of tasks in real time. According to the monitoring data, dynamically adjust the task allocation strategy of each task to optimize the resource utilization efficiency. For tasks with slow execution, reallocate resources according to the actual situation.

[0011] A further improvement of the technical solution of the present invention lies in that: the S4 specifically includes: Monitor the computing power, network status of the edge side, and the load of the cloud in real time, obtain the overall resource status of the current system, ensure the accuracy and real-time nature of the monitoring data, so as to timely reflect the changes in system resources, and conduct a resource requirement assessment for each task to be processed. Combining the task difficulty characteristics, determine the resources required for the task to be executed at the edge side or in the cloud, and ensure that the evaluation results can accurately reflect the actual resource requirements of the task; Formulate a dynamic task offloading strategy according to the computing power, network status of the edge side, and the cloud load. When the resources of the edge side are insufficient or the task complexity exceeds its processing capacity, give priority to offloading the task to the cloud; According to the formulated offloading strategy, offload eligible tasks from the edge side to the cloud. During the task offloading process, monitor the execution status and resource usage of the task in real time, and dynamically adjust the offloading strategy according to the changes in system resources. For the offloaded tasks, reallocate resources according to the actual load of the cloud to ensure that the tasks can be executed efficiently.

[0012] A further improvement of the technical solution of the present invention lies in that: the process of task offloading is as follows: According to the formulated offloading strategy, screen out tasks that meet the offloading conditions, and before starting the offloading operation, conduct a comprehensive analysis of the tasks to clarify the relevant information of the tasks, including task type, data scale, and processing logic. At the same time, establish a communication channel with the cloud to ensure the reliability of data transmission, prepare the protocols and interfaces required for task offloading, and ensure the standardization and compatibility of the offloading process, and then officially start the task offloading process to transmit eligible tasks from the edge side to the cloud; During the task offloading process, monitor the execution status and resource usage of the task in real time, pay attention to the execution status of the task to track the dynamics of task offloading, and dynamically adjust the offloading strategy according to the monitored changes in system resources; For tasks that have been offloaded to the cloud, regularly re-evaluate the resource requirements of the tasks according to the actual load conditions of the cloud for resource reallocation. If the cloud load changes, reallocate the resources required for the tasks according to the current resource status to ensure the execution efficiency of the tasks in the cloud. At the same time, optimize the resource allocation strategy according to the execution situation of the tasks and the resource utilization efficiency to further improve the overall performance and resource utilization rate of the system.

[0013] A further improvement of the technical solution of the present invention lies in that: the S5 specifically includes: The edge side receives the assigned tasks, parses the tasks, and automatically allocates local computing, storage, and network resources according to the resource requirements of the tasks and the current status of the edge devices to ensure that the tasks can obtain sufficient resource support. At the same time, record the detailed information of the tasks and the resource allocation situation; After the task parsing is completed, the edge side comprehensively evaluates the local available computing resources, checks the CPU usage rate, remaining computing power, memory occupancy, available capacity, and the read and write speed and remaining space of the storage device, and allocates resources according to the task requirements and local resource status to reserve computing, memory, and storage space for task processing. At the same time, load the algorithm library and model files required for task processing to ensure that the local environment has the conditions to process the tasks; Using the prepared local resources and algorithm models, the edge side quickly processes the tasks, and during the processing, real-time monitors the execution progress and resource usage of the tasks to ensure that the tasks proceed as expected, and then obtains the preliminary results of the tasks, and formats the preliminary results to meet the requirements of subsequent uploading and cloud processing; The edge side encrypts the generated preliminary results to ensure the security and confidentiality of the data during transmission, uploads the encrypted preliminary results to the cloud. After the upload is completed, the edge side sends feedback information to the task initiator to inform that the preliminary processing of the task has been completed and uploaded to the cloud. At the same time, the edge side records the relevant information of the task processing, including processing time, resource consumption, etc., for subsequent performance analysis and optimization.

[0014] A further improvement of the technical solution of the present invention lies in that: the S6 specifically includes: The cloud receives the task processing results uploaded by the edge side, ensures the integrity and accuracy of the data, decrypts and validates the format of the received data, and stores it in the distributed storage system of the cloud for subsequent analysis and processing. At the same time, record the upload time, data volume, and source information of the tasks; The cloud analyzes the received offloaded tasks, extracts valuable information and patterns from the data, analyzes the execution situation of the tasks, and evaluates the effectiveness and accuracy of the task results to provide a basis for integrating the results; Integrate the task processing results uploaded by the edge side with the analysis results of the offloading tasks by the cloud. Adopt a unified data format and standard to clean, transform, and match data from different sources to ensure data consistency and compatibility. Then analyze the integrated data, compare the results of different processing links, find differences and commonalities, and deeply analyze the impact of each link of task processing on the final result; Based on the results of the comprehensive analysis, the cloud automatically generates a final data analysis report, covering the overall situation of task processing, key findings, potential value, and relevant suggestions. After generating the data analysis report, feedback it to the task initiator, and at the same time store the data analysis report and relevant data in the cloud database for subsequent query, traceability, and further analysis.

[0015] Due to the adoption of the above technical solution, the technical progress achieved by the present invention compared with the prior art is: The present invention provides an intelligent data analysis method for an AI server based on edge-cloud collaboration. By deploying devices at the edge side and preprocessing data, effectively offload some computing tasks to the edge side, significantly reducing the computing burden on the cloud AI server. The edge side can independently complete simple tasks, and complex tasks are uploaded to the cloud as needed, avoiding the over-concentrated use of cloud resources, thereby reducing the energy consumption and operating costs of the cloud server and improving the energy efficiency ratio of the overall system.

[0016] The present invention provides an intelligent data analysis method for an AI server based on edge-cloud collaboration. Combine the computing power of the edge side, network status, and cloud load to formulate an offloading strategy to ensure that tasks are executed in the most suitable environment. When the resources at the edge side are insufficient or the tasks are complex, the tasks are offloaded to the cloud; otherwise, they are retained for processing at the edge side. This flexible resource allocation mechanism avoids the idle and waste of resources, improves the overall utilization rate of system resources, and realizes the optimal allocation of resources. Brief Description of the Drawings

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings.

[0018] Figure 1 It is a schematic diagram of the working process of the present invention.

[0019] Figure 2 It is a schematic diagram of the method process of the present invention. Detailed Embodiments

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the protection scope of the present invention.

[0021] Embodiment 1, as Figure 1 , Figure 2 shown, the present invention provides an AI server intelligent data analysis method based on edge-cloud collaboration, including the following steps: S1. Deploy edge devices at the edge side, collect the original data of each task in real time, preprocess the data, extract task difficulty features, form a task difficulty feature sequence, plan the types, quantities, and deployment locations of the edge devices required at the edge side based on the application scenario requirements, select edge devices with corresponding computing, storage, and communication capabilities, deploy the edge devices at the edge side to ensure their normal connection to the system network, initialize the parameters of the edge devices, configure the data collection tasks and frequencies, and make them in a standby state. The deployed edge devices continuously collect the original data generated by each task according to preset rules and frequencies. The data sources include various sensors, device interfaces, or other data sources, etc. During the collection process, monitor the data quality, mark the missing and abnormal data, and preprocess the collected original data on the edge devices. The preprocessing includes data formatting, data cleaning, and data normalization. Among them, the data structure and format are unified through data formatting, noise, duplicate data, and unreasonable data points are removed through data cleaning, and the data is adjusted to a unified dimension range through data normalization. Feature analysis is performed on the preprocessed data to extract task difficulty features reflecting the task difficulty, including the data volume size, data update frequency, data value density, and computing resource requirements. The data volume size is the total volume of the task data, that is, the file size of the task data. The data update frequency is the update frequency of the task data, that is, the dynamicity of the data. The data value density is the proportion of valid information in the data, which measures the importance of the data to the task and is measured by the ratio of the valid data volume to the total data volume. The computing resource requirement is the degree of the task's demand for computing resources. Integrate the extracted task difficulty features, set the reference values of each task difficulty feature according to the preset edge-side and cloud-side task division criteria, and then form a task difficulty feature sequence, which is stored in the local storage system of the edge device; S2. Based on the task difficulty feature sequence and the pre-trained AI data analysis model, combined with the current resource status, evaluate the task difficulty of the collected task data, and judge the feasibility of performing the task at the edge. By evaluating the task requirements, the edge can determine whether it can complete the task independently. Read the task difficulty feature sequence from the local storage of the edge device. At the same time, load the pre-trained AI data analysis model and obtain the current resource status of the edge in real time, including CPU usage, remaining memory, and storage space. Among them, collect a large number of task sample data, including the difficulty features of different types of tasks, label each task sample, and the labeling content includes the actual task difficulty score of the task and the computing resources required to complete the task. Extract features related to task difficulty from the collected task sample data, select the random forest model as the basic architecture to train the AI data analysis model, initialize the parameters of the selected random forest model, input the task difficulty feature data in the training set into the model, obtain the predicted task difficulty evaluation result through the calculation of the model, calculate the loss function value according to the predicted result and the actual labeled task difficulty evaluation label, calculate the gradient through the backpropagation algorithm according to the loss function value, and use the optimization algorithm to update the parameters of the model to minimize the loss function value. Repeat the processes of forward propagation, loss function calculation, backpropagation, and parameter update until the training stop condition is met. Use the test set to evaluate the trained model, calculate the evaluation metrics of the model, and optimize the model according to the model evaluation result. The optimization methods include adjusting the model parameters, increasing or decreasing the number of layers and neurons of the model. Through multiple optimizations and evaluations, obtain the trained AI data analysis model. Finally, deploy the trained AI data analysis model to the edge device so that it can receive the task difficulty feature data in real time and perform task difficulty evaluation. Input the task difficulty features in the task difficulty feature sequence into the AI data analysis model, analyze and process the input task difficulty features, and output the task difficulty evaluation result of the collected task data. Among them, the AI data analysis model receives the input task difficulty features and outputs the task difficulty score. The task difficulty score is from 1 to 10 points. Divide the task difficulty into different task difficulty levels, namely simple tasks, medium-difficulty tasks, and complex tasks, and determine the score range of each task difficulty level. Based on the task difficulty evaluation result, combined with the characteristics and historical data of different types of tasks, judge the feasibility of performing the task at the edge; In addition, the calculation process of the task difficulty score is as follows: Based on the task difficulty feature sequence, calculate the total volume of task data to obtain the data volume size, determine the update frequency of the task data, which represents the dynamicity of the data, calculate the proportion of valid information in the data to obtain the data value density, determine the degree of demand for computing resources by the task, and determine the reference value of each task difficulty feature. Calculate the ratio of the feature value of each task difficulty feature to its reference value respectively, and square each ratio. Then, perform a weighted sum on the squared ratios. Here, the weights of each task difficulty feature are equal, all being 1 / 4. Furthermore, calculate the square root of the weighted sum and multiply it by 10 to obtain the final task difficulty score; The calculation expression for the task difficulty score is: ; In the formula, is the task difficulty score, with a range of 1 to 10 points. The higher the value, the greater the task difficulty. is the data volume size, representing the total volume of task data. is the reference value of the data volume size, used to standardize the data volume size and determined based on historical data. is the data update frequency, representing the dynamicity of the task data. is the reference value of the data update frequency, used to standardize the data update frequency and determined based on historical data. is the data value density, representing the proportion of valid information in the data. is the reference value of the data value density, used to standardize the data value density and determined based on historical data. is the computing resource requirement, representing the degree of demand for computing resources by the task. is the reference value of the computing resource requirement, used to standardize the computing resource requirement and determined based on historical data. The score range for simple tasks is The score range for medium-difficulty tasks is The score range for complex tasks is ; S3. Based on the task difficulty assessment results, divide the tasks into cloud tasks and edge tasks, and manage the allocation of different tasks. According to the task difficulty assessment results, combined with the resource characteristics and processing capacity differences between the cloud and the edge, formulate the criteria for dividing tasks into cloud tasks and edge tasks. Considering the requirements of tasks for computing resources, storage resources, as well as the requirements for real-time performance and security, divide simple tasks into edge tasks and complex tasks into cloud tasks. For medium-difficulty tasks, divide tasks with high computing requirements, non-real-time requirements, and relatively low sensitivity to data security into cloud tasks, and divide tasks with low computing requirements, high real-time requirements, and the need for rapid local processing into edge tasks. According to the formulated division criteria, classify all tasks to be processed, and assign tasks that meet the characteristics of cloud tasks to the cloud task set. Cloud tasks require powerful computing capabilities and large-scale data storage support and can be centrally processed in the cloud. Assign tasks that meet the characteristics of edge tasks to the edge task set. Edge tasks emphasize local processing and quick response to reduce data transmission latency. For the divided cloud tasks and edge tasks, establish a task management system and formulate a task allocation strategy to uniformly manage and monitor cloud tasks and edge tasks; In addition, the process of formulating the task allocation strategy is as follows: Design and build a task management system with functions of task classification, resource scheduling, and monitoring and management, supporting the collaborative work of the cloud and the edge, ensuring that tasks are allocated to the corresponding computing nodes according to the classification results. According to the classification results of tasks, formulate a detailed task allocation strategy. Among them, for cloud tasks, consider the computing resources and storage capabilities of the cloud to ensure that tasks can be efficiently executed on appropriate servers. For edge tasks, focus on local processing capabilities and preferentially allocate them to edge devices with sufficient resources. According to the formulated task allocation strategy, allocate cloud tasks and edge tasks to the corresponding computing nodes. Cloud tasks are sent to cloud servers for centralized processing, and edge tasks are allocated to edge devices for local processing. At the same time, monitor the execution status of tasks in real time, discover and handle abnormal situations during task execution, uniformly monitor cloud and edge tasks, collect the execution status, resource usage, and performance metrics of tasks in real time, and dynamically adjust the task allocation strategy of each task according to the monitoring data to optimize the resource utilization efficiency. For tasks with slow execution, reallocate resources according to the actual situation; S4. Based on the computing capacity, network status of the edge and the cloud load, formulate an offloading strategy. If the resources of the edge are insufficient or the tasks are complex, offload the tasks to the cloud, make full use of the resource advantages of the edge and the cloud, avoid resource waste, ensure the efficient execution of tasks, and reduce the pressure on the cloud; S5. The edge receives tasks, quickly processes them using local computing resources, obtains preliminary results, and uploads them to the cloud; S6. The cloud receives the task processing results uploaded by the edge side, analyzes the received offloading tasks, and then integrates the task processing results of the edge side and the cloud to generate a final data analysis report.

[0022] Embodiment 2. As Figure 1 、 Figure 2 shown, based on Embodiment 1, the present invention provides a technical solution: Preferably, S4 specifically includes: Real-time monitor the computing power of the edge side, network status, and the load condition of the cloud, obtain the overall resource status of the current system, ensure the accuracy and real-time of the monitored data, so as to timely reflect the changes in system resources, and conduct a resource requirement assessment for each task to be processed. Combine the task difficulty characteristics to determine the resources required for the task to be executed at the edge side or the cloud, ensure that the assessment result can accurately reflect the actual resource requirements of the task. According to the computing power, network status of the edge side and the cloud load, formulate a dynamic task offloading strategy. When the resources of the edge side are insufficient or the task complexity exceeds its processing capacity, give priority to offloading the task to the cloud. At the same time, consider the impact of network status on data transmission, avoid the decline of task execution efficiency caused by network delay or insufficient bandwidth, ensure that the offloading strategy can make full use of resources while reducing the pressure on the cloud. According to the formulated offloading strategy, offload the eligible tasks from the edge side to the cloud. During the task offloading process, real-time monitor the execution status and resource usage of the task, and dynamically adjust the offloading strategy according to the changes in system resources. For the offloaded tasks, reallocate resources according to the actual load condition of the cloud to ensure the efficient execution of the tasks; In addition, the process of task offloading is as follows: According to the formulated uninstallation strategy, tasks that meet the uninstallation conditions are screened out. Before starting the uninstallation operation, a comprehensive analysis of the tasks is carried out to clarify the relevant information of the tasks, including task type, data scale, and processing logic. At the same time, a communication channel with the cloud is established to ensure the reliability of data transmission. The protocols and interfaces required for task uninstallation are prepared to ensure the standardization and compatibility of the uninstallation process. Then, the task uninstallation process is officially started, and the eligible tasks are transmitted from the edge side to the cloud. During the task uninstallation process, the execution status and resource usage of the tasks are monitored in real time. Pay attention to the execution status of the tasks to track the dynamics of task uninstallation. According to the monitored changes in system resources, the uninstallation strategy is dynamically adjusted. Among them, if the resources at the edge side are suddenly relieved during the uninstallation process and the cloud load is high, the number of uninstalled tasks can be appropriately reduced or some uninstallation operations can be suspended, and some tasks can be retained for processing at the edge side. On the contrary, if the resources at the edge side are continuously tense and the cloud has sufficient processing capacity, the task uninstallation speed can be accelerated. By flexibly adjusting the uninstallation strategy, ensure that system resources are reasonably utilized, and avoid situations of resource waste or task execution obstruction. For the tasks that have been uninstalled to the cloud, according to the actual load situation of the cloud, regularly re-evaluate the resource requirements of the tasks for resource reallocation. If the cloud load changes, re-allocate the resources required for the tasks according to the current resource status to ensure the execution efficiency of the tasks in the cloud. At the same time, according to the execution situation of the tasks and the resource usage efficiency, optimize the resource allocation strategy to further improve the overall performance and resource utilization rate of the system; S5 specifically includes: The edge device receives the assigned task, parses the task, and automatically allocates local computing, storage, and network resources based on the resource requirements of the task and the current status of the edge device to ensure that the task can obtain sufficient resource support. At the same time, it records the detailed information of the task and the resource allocation situation. After task parsing is completed, the edge device comprehensively evaluates the available local computing resources, checks the CPU usage rate, remaining computing power, memory occupancy, available capacity, as well as the read / write speed and remaining space of the storage device, and allocates resources according to the task requirements and local resource conditions, reserving computing, memory, and storage space for task processing. At the same time, it loads the algorithm library and model files required for task processing to ensure that the local environment has the conditions to process the task. Using the prepared local resources and algorithm models, the edge device quickly processes the task, and during the processing, it monitors the execution progress and resource usage of the task in real time to ensure that the task proceeds as expected, and then obtains the preliminary result of the task. It formats the preliminary result to meet the requirements of subsequent uploading and cloud processing. The edge device encrypts the generated preliminary result to ensure the security and confidentiality of the data during transmission. It uploads the encrypted preliminary result to the cloud. After the upload is completed, the edge device sends feedback information to the task initiator, informing that the preliminary processing of the task has been completed and uploaded to the cloud. At the same time, the edge device records the relevant information of task processing, including processing time, resource consumption, etc., for subsequent performance analysis and optimization; S6 specifically includes: The cloud receives the task processing results uploaded by the edge side, ensures the integrity and accuracy of the data, decrypts and validates the format of the received data, and stores it in the distributed storage system in the cloud for subsequent analysis and processing. At the same time, the upload time, data volume, and source information of the task are recorded. Among them, the format of the data is verified to ensure that it conforms to the preset specification standards. If the verification passes, the data is stored in the temporary buffer. If the verification fails, a retransmission request is sent to the edge side to request re-uploading the data. The cloud analyzes the received offloading tasks, extracts valuable information and patterns from the data, analyzes the execution of the tasks, evaluates the effectiveness and accuracy of the task results, and provides a basis for integrating the results. Integrate the task processing results uploaded by the edge side with the analysis results of the offloading tasks by the cloud. Adopt unified data formats and standards to clean, transform, and match data from different sources to ensure data consistency and compatibility. Then analyze the integrated data, compare the results of different processing links, find differences and commonalities, and deeply analyze the impact of each link of task processing on the final result. Based on the results of the comprehensive analysis, the cloud automatically generates a final data analysis report covering the overall situation of task processing, key findings, potential value, and relevant suggestions. After generating the data analysis report, it is fed back to the task initiator, and at the same time, the data analysis report and related data are stored in the cloud database for subsequent query, traceability, and further analysis.

[0023] As described above, it is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed by this application can easily think of changes or substitutions, which should all be covered within the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claimed rights.

Claims

1. An intelligent data analysis method for an AI server based on edge-cloud collaboration, characterized in that It includes the following steps: S1. Deploy edge devices at the edge side, collect the original data of each task in real time, preprocess the data, extract task difficulty features, and form a task difficulty feature sequence; S2. Based on the task difficulty feature sequence and a pre-trained AI data analysis model, evaluate the task difficulty of the collected task data in combination with the current resource status, and judge the feasibility of performing tasks at the edge side; S3. Combine the task difficulty evaluation results, divide the tasks into cloud tasks and edge tasks, and allocate and manage different tasks; S4. Develop an offloading strategy in combination with the edge-side computing power, network status, and cloud load; S5. The edge side receives the tasks, quickly processes them using local computing resources, obtains preliminary results, and uploads them to the cloud; S6. The cloud receives the task processing results uploaded by the edge side, analyzes the received offloaded tasks, and then integrates the task processing results of the edge side and the cloud to generate a final data analysis report.

2. The intelligent data analysis method of an AI server based on edge-cloud collaboration according to claim 1, wherein: The specific content of S1 includes: Based on the application scenario requirements, plan the types, quantities, and deployment locations of the edge devices required at the edge side, deploy the edge devices at the edge side, initialize the edge device parameters, configure the data collection tasks and frequencies, and make them in a standby state; The deployed edge devices continuously collect the original data generated by each task according to preset rules and frequencies, and monitor the data quality during the collection process, and mark the missing and abnormal data; Preprocess the collected original data on the edge device. The preprocessing includes data formatting, data cleaning, and data normalization, and perform feature analysis on the preprocessed data to extract task difficulty features reflecting task difficulty, including data volume size, data update frequency, data value density, and computing resource requirements; Integrate the extracted task difficulty features, set the reference values of each task difficulty feature according to the preset edge-side and cloud task division criteria, and then form a task difficulty feature sequence, and store it in the local storage system of the edge device.

3. An intelligent data analysis method for an AI server based on edge-cloud collaboration according to claim 2, characterized in that: The specific content of S2 includes: Read the task difficulty feature sequence from the local storage of the edge device. At the same time, load the pre-trained AI data analysis model, and obtain the current resource status of the edge side in real time, including CPU usage, remaining memory, and storage space; Input the task difficulty features in the task difficulty feature sequence into the AI data analysis model, analyze and process the input task difficulty features, and output the task difficulty evaluation result of the collected task data. Among them, the AI data analysis model receives the input task difficulty features, outputs a task difficulty score, the task difficulty score is 1-10 points, divides the task difficulty into different task difficulty levels, namely simple tasks, medium-difficulty tasks, and complex tasks, and determines the score range of each task difficulty level; Based on the task difficulty evaluation result, judge the feasibility of performing tasks at the edge side in combination with the characteristics and historical data of different types of tasks.

4. An intelligent data analysis method for an AI server based on edge-cloud collaboration according to claim 3, characterized in that: The calculation process of the task difficulty score is: Based on the task difficulty feature sequence, calculate the total volume of task data to obtain the data volume size, determine the update frequency of task data, which represents the dynamic nature of the data, calculate the proportion of valid information in the data to obtain the data value density, determine the degree of demand for computing resources of the task, and determine the reference values of each task difficulty feature; Calculate the ratio of the feature value of each task difficulty feature to its reference value respectively, and square each ratio; Perform a weighted sum of the squared ratios. Among them, the weights of each task difficulty feature are equal, all being 1 / 4. Then calculate the square root of the weighted sum, and multiply by 10 to obtain the final task difficulty score.

5. The intelligent data analysis method of an AI server based on edge-cloud collaboration according to claim 3, characterized in that: The specific content of S3 includes: Based on the task difficulty assessment results, combined with the resource characteristics and processing capacity differences between the cloud and the edge, formulate the criteria for dividing tasks into cloud tasks and edge tasks. Considering the comprehensive requirements of tasks for computing resources, storage resources, as well as the requirements for real-time performance and security, simple tasks are divided into edge tasks, and complex tasks are divided into cloud tasks. For medium-difficulty tasks, tasks with high computing requirements, non-real-time requirements, and relatively low sensitivity to data security are divided into cloud tasks, while tasks with low computing requirements, high real-time requirements, and the need for local and rapid processing are divided into edge tasks; According to the formulated division criteria, classify all tasks to be processed, and classify tasks that meet the characteristics of cloud tasks into the cloud task set, and classify tasks that conform to the characteristics of edge tasks into the edge task set; For the divided cloud tasks and edge tasks, establish a task management system and formulate a task allocation strategy to uniformly manage and monitor cloud tasks and edge tasks.

6. The intelligent data analysis method of an AI server based on edge-cloud collaboration according to claim 5, characterized in that: The process of formulating the task allocation strategy is as follows: Design and build a task management system with functions of task classification, resource scheduling, and monitoring management, supporting the collaborative work of the cloud and the edge; According to the classification results of tasks, formulate a detailed task allocation strategy, and according to the formulated task allocation strategy, allocate cloud tasks and edge tasks to the corresponding computing nodes. Cloud tasks are sent to the cloud server for centralized processing, and edge tasks are allocated to edge devices for local processing; Unify the monitoring of cloud and edge tasks, collect the execution status, resource usage, and performance metrics of tasks in real time, and dynamically adjust the task allocation strategy of each task according to the monitoring data. For tasks with slow execution, reallocate resources according to the actual situation.

7. An intelligent data analysis method for an AI server based on edge-cloud collaboration according to claim 1, characterized in that: The specific content of S4 includes: Real-time monitor the computing power, network status of the edge and the load of the cloud, obtain the overall resource status of the current system, and conduct a resource requirement assessment for each task to be processed. Combining the task difficulty features, determine the resources required for the task to be executed at the edge or in the cloud; According to the computing power, network status of the edge and the cloud load, formulate a dynamic task offloading strategy. When the resources at the edge are insufficient or the task complexity exceeds its processing capacity, give priority to offloading the task to the cloud; According to the formulated offloading strategy, tasks that meet the conditions are offloaded from the edge to the cloud. During the task offloading process, the execution status and resource usage of the tasks are monitored in real time, and the offloading strategy is dynamically adjusted according to changes in system resources. For the offloaded tasks, resources are reallocated according to the actual load situation of the cloud.

8. An intelligent data analysis method for an AI server based on edge-cloud collaboration according to claim 7, characterized in that: The process of task offloading is as follows: Based on the formulated offloading strategy, tasks that meet the offloading conditions are screened out. Before starting the offloading operation, a comprehensive analysis of the tasks is carried out to clarify the relevant information of the tasks, including task type, data scale, and processing logic. At the same time, a communication channel with the cloud is established, and the protocols and interfaces required for task offloading are prepared. Then, the task offloading process is officially started, and the tasks that meet the conditions are transmitted from the edge to the cloud. During the task offloading process, the execution status and resource usage of the tasks are monitored in real time, the execution status of the tasks is concerned to track the dynamics of task offloading, and the offloading strategy is dynamically adjusted according to the monitored changes in system resources. For the tasks that have been offloaded to the cloud, according to the actual load situation of the cloud, the resource requirements of the tasks are regularly re-evaluated for resource reallocation. If the cloud load changes, the resources required for the tasks are reallocated according to the current resource status. At the same time, the resource allocation strategy is optimized according to the execution situation and resource usage efficiency of the tasks.

9. An intelligent data analysis method for an AI server based on edge-cloud collaboration according to claim 1, characterized in that: The specific steps of S5 are as follows: The edge device receives the assigned tasks, parses the tasks, and automatically allocates local computing, storage, and network resources according to the resource requirements of the tasks and the current state of the edge device. At the same time, the detailed information of the tasks and the resource allocation situation are recorded. After the task parsing is completed, the edge device comprehensively evaluates the available local computing resources, checks the CPU usage rate, remaining computing power, memory occupancy, available capacity, and the read / write speed and remaining space of the storage device. According to the task requirements and local resource status, resources are allocated to reserve computing, memory, and storage space for task processing. At the same time, the algorithm library and model files required for task processing are loaded. Using the prepared local resources and algorithm models, the edge device quickly processes the tasks. During the processing process, the execution progress and resource usage of the tasks are monitored in real time, and then the preliminary results of the tasks are obtained. The preliminary results are formatted to meet the requirements of uploading and cloud processing. The edge device encrypts the generated preliminary results, uploads the encrypted preliminary results to the cloud. After the upload is completed, the edge device sends feedback information to the task initiator, informing that the preliminary task processing has been completed and uploaded to the cloud. At the same time, the edge device records the relevant information of task processing.

10. An intelligent data analysis method for an AI server based on edge-cloud collaboration according to claim 9, characterized in that: The specific steps of S6 are as follows: The cloud receives the task processing results uploaded by the edge device, decrypts and verifies the format of the received data, and stores it in the distributed storage system of the cloud. At the same time, the upload time, data volume, and source information of the tasks are recorded. The cloud analyzes the received offloaded tasks, extracts valuable information and patterns from the data, analyzes the execution situation of the tasks, and evaluates the effectiveness and accuracy of the task results. Integrate the task processing results uploaded by the edge side with the analysis results of the offloading tasks by the cloud. Adopt a unified data format and standard to clean, transform, and match data from different sources. Then analyze the integrated data, compare the results of different processing links, identify differences and commonalities, and deeply analyze the impact of each link of task processing on the final result; Based on the results of the comprehensive analysis, the cloud automatically generates a final data analysis report, covering the overall situation of task processing, key findings, potential value, and relevant suggestions. After generating the data analysis report, feedback it to the task initiator, and at the same time store the data analysis report and related data in the cloud database.

Images