Function machine data optimization management system and method based on cloud application

Abstract

The invention discloses a function machine data optimization management system and method based on cloud application, and relates to the technical field of data optimizing.In a function machine, a 4g network is used for being connected with a cloud server, historical tasks of the function machine are extracted, historical task data are analyzed, features are extracted, and a task fingerprint database is generated; the function machine initiates a task request in real time, calculates the similarity of real-time tasks by using task fingerprints, and judges the types of the real-time tasks; performing conflict analysis on the judged real-time tasks, and merging the same tasks obtained by judgment to form a task matrix; the cloud server operates the LSTM model for one time to process the task matrix, and outputs a processing result; and extracting a result public part and a result difference part from an output result, carrying out compression transmission on the two results, and locally recombining task results by the function machine to obtain a final result of the real-time task.

Description

Technical Field

[0001] The present invention relates to the technical field of data optimization, and specifically to a feature phone data optimization management system and method based on cloud applications. Background Art

[0002] With the rapid development of communication technologies, feature phones still occupy a certain market share in people's lives. Although feature phones lag behind smartphones in terms of hardware performance and function richness, their simplicity, ease of use, and long battery life make them irreplaceable in some specific scenarios and user groups. However, the limited local storage and processing capabilities of feature phones pose many challenges to data management. Traditional feature phones can often only store a small amount of data locally and are inefficient in processing complex data tasks, making it difficult to meet users' needs for in-depth data analysis and efficient management. Therefore, in the face of ever-changing application requirements, it is necessary to leverage the powerful computing and storage capabilities of the cloud to expand their functions and application scope. This approach not only eliminates the need for significant hardware upgrades to 4G feature phones but also enables flexible function expansion and dynamic allocation to meet users' changing needs.

[0003] However, when feature phones use cloud servers to process complex tasks, in many cases where there are a large number of online feature phones, the task requests sent to the cloud server are numerous. For similar tasks, the cloud can only process them sequentially, which greatly consumes the computing resources of the cloud server. Therefore, it is crucial to reduce the number of processing times, unify the processing of similar tasks, and ensure the integrity of the specific differentiation results of different feature phones. Summary of the Invention

[0004] The purpose of the present invention is to provide a feature phone data optimization management system and method based on cloud applications to solve the problems raised in the prior art.

[0005] To achieve the above objective, the present invention provides the following technical solutions: A feature phone data optimization management method based on cloud applications, the method comprising the following steps: S100. Use the 4G network in the feature phone to connect to the cloud server, extract the historical tasks of the feature phone, analyze and extract features from the historical task data, construct task fingerprints, and generate a task fingerprint library; Further, the specific steps for constructing task fingerprints are: S101. Use the 4G network in the feature phone to connect to the cloud server, record task logs at the feature phone end, compress and transmit the task data in the task logs to the cloud server, and the cloud server preprocesses the task data, specifically: remove the missing values in the task data, filter the task data using a filter; finally, standardize the task data, and the formula is: ; In the formula, s’ represents the standardized task data, s represents the task data received by the cloud server, sp represents the average value of the task data, and sz represents the standard deviation of the task data; S102. Extract the timestamp of each task data and calculate the duration of each task. The formula is: , where in the formula, △t represents the task duration, t end represents the task end time, and t start represents the task start time; Use the duration of each task as the time feature; Calculate the average value, variance, and peak value of each task data in the time domain as time domain features; In the frequency domain, use the Fourier transform to convert all task data into frequency domain signals s[k], and calculate the amplitude spectrum of the frequency domain signals in each task data. The formula is: ; In the formula, |s[k]| represents the amplitude spectrum of the frequency domain signal, Re(s[k]) represents the real part of the frequency domain signal, and Im(s[k]) represents the imaginary part of the frequency domain signal; Calculate the frequency domain features of each task data. The formula is: , where in the formula, Pt represents the frequency domain feature of each task data, argmax represents extracting the independent variable that makes the function reach the maximum value, and k belongs to 0 to N - 1; S103. Use the mutual information screening method to screen all features in the time domain, frequency domain, and time to obtain the key features of each task, and construct a feature vector V with all key features; Generate task fingerprints using the feature vector of each task. The formula is: ; In the formula, H represents the task fingerprint, MinHsah represents the min - hash algorithm, and LSH represents the locality - sensitive hashing algorithm; Generate a 128 - bit MinHash fingerprint H, generate the task fingerprint H for each type of task, and obtain the task fingerprint library.

[0006] Connecting to the cloud server through the 4G network can make full use of the powerful computing and storage resources of the cloud, without the need to equip the feature phone locally with a large - capacity storage and a high - performance processor to process and store a large amount of historical task data, reducing the hardware cost and energy consumption of the feature phone. Storing the historical task data in the cloud server is convenient for centralized management and maintenance, and can implement functions such as regular data backup and disaster recovery, improving the security and reliability of the data, and avoiding data loss caused by reasons such as the loss or damage of the feature phone.

[0007] Constructing task fingerprints can generate unique identifiers for each task, which helps to quickly identify and distinguish different tasks. The establishment of a task fingerprint library facilitates the classification, retrieval, and comparison of tasks. For example, similar tasks can be quickly found through task fingerprints, providing a reference and decision-making basis for users. It can also be used to detect abnormal tasks and timely discover potential problems or risks.

[0008] S200. The feature phone initiates a task request in real time. After receiving the task request, the cloud server determines the number of feature phones, calculates the similarity of the real-time task using task fingerprints, and determines the real-time task type; updates the task fingerprint library with the new task type; Further, the specific steps for determining whether it is the same type of task are as follows: S201. All feature phones regularly send heartbeat packets to the cloud at time t1. The heartbeat packet contains the feature phone ID. The cloud server maintains an online device list based on the regular heartbeat packets. When a device does not send a heartbeat packet beyond the regular time t1, the cloud server removes the corresponding feature phone ID from the online device list and obtains the number of feature phones according to the real-time online device list; S202. The online feature phone sends a task request data packet to the cloud server. The data packet format is Ta_request={ID, t_request, F i}, where Ta_request represents the task request data packet, t_request represents the task request time, and F i represents the task key feature; the task key feature is normalized to F i '; the cloud server receives the real-time task requests sent by all feature phones; S203. The cloud server compares the received real-time task requests with each historical task fingerprint, calculates the similarity, and the formula is: ; In the formula, Sim k represents the similarity between the real-time task and the k-th task fingerprint, and C k represents the k-th task fingerprint; a similarity threshold Sy is set artificially, and the similarity is judged using the similarity threshold. When Sim k ≥ Sy, it is judged as the same type of task; otherwise, the real-time task is marked as a new task type; the real-time task types of all online feature phones are judged in turn.

[0009] By judging the number of feature phones, the cloud server can reasonably allocate tasks according to the system resource load, avoiding the situation where some feature phones have too many tasks while some are idle, and improving the overall system operation efficiency and resource utilization rate. After accurately judging the real-time task type, the cloud server can call the corresponding processing module or algorithm to process the task, improving the accuracy and efficiency of task processing. For example, for different types of tasks, different priority settings, data processing methods or resource allocation schemes can be adopted to better meet user needs and system requirements.

[0010] Updating the task fingerprint library with new task types enables the system to continuously learn and adapt to new task patterns and characteristics, enriching the content of the task fingerprint library and improving the system's recognition and processing capabilities for various tasks. As time goes by, the task fingerprint library will become more and more perfect, and the system's judgment and processing of tasks will also become more accurate and efficient.

[0011] S300. Conduct conflict analysis on the judged real-time tasks, and merge the same tasks obtained by judgment to form a task matrix; Furthermore, the specific steps for conducting conflict analysis on the judged real-time tasks are as follows: S301. Count the task types of all real-time tasks to generate a task type distribution set Ta_Distribution = {Type1: n1, Type2: n2,..., TypeK: nK}, where Type1, Type2,..., TypeK represent the 1st, 2nd,..., Kth real-time task types, and n1, n2,..., nK represent the number of real-time tasks in the 1st, 2nd,..., Kth real-time task types; S302. In the same real-time task type, calculate the difference in similarity between each real-time task. The formula is: ; In the formula, Sim c (a, b) represents the difference in similarity between real-time task a and real-time task b, Sim a represents the similarity of real-time task a, and Sim b represents the similarity of real-time task b; Calculate the difference in similarity between all real-time tasks under all real-time task types in the same way, calculate the average value and standard deviation of all differences in similarity, and subtract the standard deviation from the average value to obtain the merging threshold By; When Sim c (a, b) ≤ By, it is judged that real-time task a and real-time task b can be merged. When Sim c (a, b) > By, it is judged that real-time task a and real-time task b cannot be merged; Merge the real-time tasks that are judged to be mergeable into a task matrix.

[0012] By merging the same type of tasks, redundant tasks can be removed, avoiding the repetitive processing of the same tasks by feature phones or servers, saving computing resources and time, and thus enabling tasks to be processed more quickly and improving overall efficiency. After constructing the task matrix, the resources of feature phones and servers can be uniformly allocated and optimally distributed according to the characteristics and resource requirements of the tasks. The task matrix can intuitively display the classification and distribution of tasks, making task management clearer. Administrators can clearly understand the quantity, status, and interrelationships of various tasks at a glance, facilitating effective task scheduling and management.

[0013] The S400 and the cloud server run the LSTM model once to process the task matrix and output the processing results; in the output results, the common part and the different part of the results are extracted, and the two results are compressed and transmitted; Further, the specific steps for extracting the common part and the different part of the results in the output results are as follows: S401. Input the task matrix into the LSTM model in the cloud server for single processing to output the result matrix, Results = {Y 1 、Y 2 、Y 3 、...、Y d}, Y 1 、Y 2 、Y 3 、...、Y d represents the results of the 1st, 2nd, 3rd,..., dth tasks in the result matrix; calculate the common part of the results, and the formula is: ; In the formula, Y common represents the common part of the results, and Y j represents the result of the jth task in the result matrix; Calculate the different part of the results, and the formula is: ; In the formula, Y diff j represents the different part of the result of the jth task; the cloud server separately compresses the common part and the different part of the results. For the common part of the results, the cloud server performs a single global dissemination to all online feature phones, and all online feature phones receive the common part of the results simultaneously; The cloud server transmits the different part of the results to the online feature phones with corresponding IDs in sequence according to the task request data packets.

[0014] By extracting the common parts and different parts in the result and compressing them, the amount of data to be transmitted can be significantly reduced. In network transmission, the size of the data volume directly affects the transmission speed and bandwidth occupancy. Compressed transmission can accelerate the data transmission speed and reduce the network bandwidth pressure. Especially in the case of unstable network environment or limited bandwidth, it can improve the stability and response speed of the system, ensuring that the processing results can be transmitted to the feature phone or other terminal devices in a timely and efficient manner.

[0015] S500. The feature phone receives the task result sent by the cloud server, and the feature phone reorganizes the task result locally to obtain the final result of the real-time task.

[0016] Further, the specific steps for the feature phone to reorganize the task result locally to obtain the final result of the real-time task are as follows: S501. The online feature phone respectively receives the two parts of the result transmitted by the cloud server. The feature phone decompresses the two parts of the result respectively, and reorganizes the common part and the different part of the result. The formula is: ; In the formula, Y j final represents the final result of the jth task; similarly, all online feature phones decompress and reorganize the common part and the different part to obtain the final result.

[0017] After the receiving end (such as a feature phone) receives the compressed result, it only needs to decompress and simply merge it to restore the complete processing result. Compared with directly receiving a large amount of unprocessed data, this method greatly saves the storage resources and computing resources of the terminal device, reduces the processing burden of the terminal device, enables it to run other tasks more smoothly, and improves the user experience.

[0018] A feature phone data optimization management system based on cloud applications. The feature phone data optimization management system includes a data collection module, a task fingerprint library module, a real-time task type judgment module, a task conflict judgment module, a task processing module, and a result generation module; The data collection module is used to collect the task data in the feature phone history; The task fingerprint library module is used to extract the features of the task data in the history, screen the key features, construct the task fingerprint of each task, and generate a task fingerprint library; The real-time task type judgment module is used to calculate the similarity of the real-time task using the task fingerprint and judge the real-time task type; The task conflict judgment module is used to perform conflict analysis on the judged real-time tasks, merge the same tasks obtained by the judgment, and form a task matrix; The task processing module is used for the cloud server to input the task matrix into the LSTM model for single processing, output the result matrix, and divide the result into a common part and a difference part in the result matrix; The result generation module is used for the online functional machine to receive the common part and the difference part, decompress and reorganize them to obtain the final result.

[0019] The task fingerprint library module includes a feature screening unit and a task fingerprint construction unit; The feature screening unit is used to screen all features in the time domain, frequency domain, and time using the mutual information screening method; The task fingerprint construction unit is used to obtain the task fingerprint through minhash using the screened key features.

[0020] The task conflict judgment module includes a merging judgment unit and a merging unit; The merging judgment unit is used to calculate the similarity difference and judge whether two real-time tasks corresponding to the similarity difference can be merged using the merging threshold; The merging unit is used to merge the real-time tasks that are judged to be mergeable into one task matrix.

[0021] The task processing module includes a common part calculation unit and a difference part calculation unit; The common part calculation unit is used to obtain the intersection of all results to get the common part of the results; The difference part calculation unit is used to subtract the common part from each result to get the difference part.

[0022] Compared with the prior art, the beneficial effects of the present invention are: 1. The present invention performs data processing and analysis in the cloud, and the functional machine is mainly responsible for data upload and result display, reducing the local computing burden, thereby improving the running speed and response ability of the functional machine and enhancing the user experience. Users can obtain task-related information more quickly without waiting for the functional machine to process data for a long time.

[0023] 2. The present invention can remove redundant tasks by merging the same type of tasks, avoiding repeated processing of the same tasks by the functional machine or the server, saving computing resources and time, and thus being able to complete task processing more quickly and improving the overall efficiency.

[0024] 3. The present invention can significantly reduce the amount of data to be transmitted by extracting the common part and the difference part in the results and compressing them. Brief Description of the Drawings

[0025] Figure 1 It is a module distribution diagram of a functional machine data optimization management system based on cloud applications of the present invention; Figure 2 Schematic diagram of the steps of a method for optimizing and managing feature phone data based on cloud applications according to the present invention. Specific embodiments

[0026] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0027] Embodiment: As Figure 1 - Figure 2 shown, the present invention provides a technical solution. A method for optimizing and managing feature phone data based on cloud applications, the method comprising the following steps: S100. Use a 4G network in the feature phone to connect to the cloud server, extract the historical tasks of the feature phone, analyze and extract features from the historical task data, construct task fingerprints, and generate a task fingerprint library; The specific steps for constructing task fingerprints are as follows: S101. Use a 4G network in the feature phone to connect to the cloud server, record task logs at the feature phone end, compress and transmit the task data in the task logs to the cloud server, and the cloud server preprocesses the task data, specifically: removing missing values in the task data, filtering the task data using a filter; finally, standardizing the task data, and the formula is: ; In the formula, s’ represents the standardized task data, s represents the task data received by the cloud server, sp represents the average value of the task data, and sz represents the standard deviation of the task data; S102. Extract the timestamp of each task data, calculate the duration of each task, and the formula is: , in the formula, △t represents the task duration, t end represents the task end time, t start represents the task start time; use the duration of each task as a time feature; Calculate the average value, variance, and peak value of each task data in the time domain as time domain features; In the frequency domain, use the Fourier transform to convert all task data into frequency domain signals s[k], and calculate the amplitude spectrum of the frequency domain signals in each task data, and the formula is: ; In the formula, |s[k]| represents the amplitude spectrum of the frequency-domain signal, Re(s[k]) represents the real part of the frequency-domain signal, and Im(s[k]) represents the imaginary part of the frequency-domain signal; Calculate the frequency-domain features of each task data. The formula is: , where Pt represents the frequency-domain features of each task data, argmax represents extracting the independent variable that makes the function reach the maximum value, and k belongs to 0 to N - 1; S103. Use the mutual information screening method to screen all features in the time domain, frequency domain, and time to obtain the key features of each task, and construct the feature vector V with all the key features; generate task fingerprints using the feature vector of each task. The formula is: ; In the formula, H represents the task fingerprint, MinHsah represents the minhash algorithm, and LSH represents the locality-sensitive hashing algorithm; generate a 128-bit MinHash fingerprint H, generate the task fingerprint H for each type of task, and obtain the task fingerprint library.

[0028] Connect to the cloud server through the 4G network, which can make full use of the powerful computing and storage resources of the cloud without the need to equip the feature phone locally with a large-capacity storage and high-performance processor to process and store a large amount of historical task data, reducing the hardware cost and energy consumption of the feature phone. Storing the historical task data in the cloud server is convenient for centralized management and maintenance, and functions such as regular data backup and disaster recovery can be realized, improving the security and reliability of the data and avoiding data loss due to reasons such as the loss or damage of the feature phone.

[0029] Constructing task fingerprints can generate unique identifiers for each task, which helps to quickly identify and distinguish different tasks. The establishment of the task fingerprint library facilitates the classification, retrieval, and comparison of tasks. For example, similar tasks can be quickly found through task fingerprints, providing reference and decision-making basis for users, and can also be used to detect abnormal tasks and timely discover potential problems or risks.

[0030] S200. The feature phone initiates a task request in real time. After the cloud server receives the task request, it judges the number of feature phones, calculates the similarity of the real-time task using the task fingerprint, and judges the real-time task type; updates the task fingerprint library using the new task type; The specific steps to judge whether it is the same type of task are: S201. All feature phones send heartbeat packets to the cloud regularly at time t1. The heartbeat packet contains the feature phone ID. The cloud server maintains the online device list based on the regular heartbeat packets. When a device does not send a heartbeat packet beyond the regular time t1, the cloud server removes the corresponding feature phone ID from the online device list and obtains the number of feature phones according to the real-time online device list; S202. The online functional machine sends a task request data packet to the cloud server. The data packet format is Ta_request = {ID, t_request, F i}, where Ta_request represents the task request data packet, t_request represents the task request time, and F i represents the task key feature; the task key feature is normalized to F i '; the cloud server receives the real-time task requests sent by all functional machines; S203. The cloud server compares the received real-time task requests with each historical task fingerprint, calculates the similarity, and the formula is: ; In the formula, Sim k represents the similarity between the real-time task and the k-th task fingerprint, and C k represents the k-th task fingerprint; a similarity threshold Sy is set artificially, and the similarity is judged using the similarity threshold. When Sim k ≥Sy, it is judged as the same type of task; otherwise, the real-time task is marked as a new task type; the real-time task types of all online functional machines are judged in turn.

[0031] By judging the number of functional machines, the cloud server can reasonably allocate tasks according to the system resource load situation, avoid the situation that some functional machines have too many tasks while some functional machines are idle, and improve the overall system operation efficiency and resource utilization rate; after accurately judging the real-time task type, the cloud server can call the corresponding processing module or algorithm to process the task, improving the accuracy and efficiency of task processing. For example, for different types of tasks, different priority settings, data processing methods or resource allocation schemes can be adopted to better meet user needs and system requirements.

[0032] Updating the task fingerprint library with the new task type can enable the system to continuously learn and adapt to new task patterns and characteristics, enrich the content of the task fingerprint library, and improve the system's recognition and processing ability for various tasks. As time goes by, the task fingerprint library will become more and more perfect, and the system's judgment and processing of tasks will become more and more accurate and efficient.

[0033] S300. Conduct conflict analysis on the judged real-time tasks, and merge the judged same type of tasks to form a task matrix; The specific steps for conducting conflict analysis on the judged real-time tasks are as follows: S301. Count the task types of all real-time tasks and generate a task type distribution set Ta_Distribution = {Type1: n1, Type2: n2,..., TypeK: nK}, where Type1, Type2,..., TypeK represent the 1st, 2nd,..., Kth real-time task types, and n1, n2,..., nK represent the number of real-time tasks in the 1st, 2nd,..., Kth real-time task types; S302. In the same real-time task type, calculate the difference in similarity between each pair of real-time tasks. The formula is: ; In the formula, Sim c (a, b) represents the difference in similarity between real-time task a and real-time task b, Sim a represents the similarity of real-time task a, and Sim b represents the similarity of real-time task b; Calculate the difference in similarity between all real-time tasks under all real-time task types in the same way, calculate the average value and standard deviation of all the differences in similarity, and subtract the standard deviation from the average value to obtain the merging threshold By; When Sim c (a, b) ≤ By, it is determined that real-time task a and real-time task b can be merged. When Sim c (a, b) > By, it is determined that real-time task a and real-time task b cannot be merged; Merge the real-time tasks that can be determined to be merged into a task matrix.

[0034] By merging the same tasks, redundant tasks can be removed, avoiding the functional machine or server from processing the same task repeatedly, saving computing resources and time, and thus being able to complete task processing more quickly and improving the overall efficiency. After forming the task matrix, the resources of the functional machine and the server can be uniformly allocated and optimally distributed according to the characteristics and resource requirements of the tasks. The task matrix can intuitively display the classification and distribution of tasks, making task management clearer. The administrator can clearly understand the quantity, status, and mutual relationship of various tasks at a glance, facilitating effective task scheduling and management.

[0035] S400. The cloud server runs the LSTM model once to process the task matrix and outputs the processing result; Extract the common part and the different part of the result from the output result, and perform compressed transmission on the two results; The specific steps to extract the common part and the different part of the result from the output result are as follows: S401. Input the task matrix into the LSTM model in the cloud server for single processing, and output the result matrix, Results = {Y 1 、Y 2 、Y3 ,..., Y d}, Y 1 , Y 2 , Y 3 ,..., Y d represent the results of the 1st, 2nd, 3rd,..., d-th tasks in the result matrix; calculate the common part of the results, and the formula is: ; In the formula, Y common represents the common part of the results, and Y j represents the result of the j-th task in the result matrix; Calculate the different part of the results, and the formula is: ; In the formula, Y diff j represents the different part of the result of the j-th task; the cloud server separately compresses the common part and the different part of the results. For the common part of the results, the cloud server makes a global broadcast to all online functional machines at one time, and all online functional machines receive the common part of the results simultaneously; The cloud server transmits the different part of the results to the online functional machines corresponding to the IDs in sequence according to the task request data packets.

[0036] By extracting the common part and the different part of the results and compressing them, the amount of data to be transmitted can be significantly reduced. In network transmission, the size of the data volume directly affects the transmission speed and bandwidth occupancy. Compressed transmission can accelerate the data transmission speed and reduce the network bandwidth pressure. Especially in the case of unstable network environment or limited bandwidth, it can improve the stability and response speed of the system, ensuring that the processing results can be transmitted to the functional machines or other terminal devices in a timely and efficient manner.

[0037] S500. The functional machine receives the task results sent by the cloud server, and the functional machine reorganizes the task results locally to obtain the final real-time task results.

[0038] The specific steps for the functional machine to reorganize the task results locally to obtain the final real-time task results are as follows: S501. The online functional machine separately receives the two parts of the results transmitted by the cloud server. The functional machine separately decompresses the two parts of the results and reorganizes the common part and the different part of the results. The formula is: ; In the formula, Y j final represents the final result of the j-th task; similarly, all online functional machines decompress and reorganize the common part and the different part to obtain the final results.

[0039] After the receiving end (such as a feature phone) receives the compressed result, it only needs to decompress and simply merge it to restore the complete processing result. Compared with directly receiving a large amount of unprocessed data, this method greatly saves the storage resources and computing resources of the terminal device, reduces the processing burden of the terminal device, enables it to run other tasks more smoothly, and improves the user experience.

[0040] A feature phone data optimization management system based on cloud applications, the feature phone data optimization management system includes a data acquisition module, a task fingerprint library module, a real-time task type judgment module, a task conflict judgment module, a task processing module, and a result generation module; The data acquisition module is used to collect task data in the feature phone history; The task fingerprint library module is used to extract the features of the task data in the history, screen the key features, construct the task fingerprint of each task, and generate a task fingerprint library; The real-time task type judgment module is used to calculate the similarity of the real-time task using the task fingerprint and judge the real-time task type; The task conflict judgment module is used to perform conflict analysis on the judged real-time tasks, merge the same tasks obtained by the judgment, and form a task matrix; The task processing module is used for the cloud server to input the task matrix into the LSTM model for single processing, output a result matrix, and divide the result into a common part and a difference part in the result matrix; The result generation module is used for the online feature phone to receive the common part and the difference part, decompress and reorganize them to obtain the final result.

[0041] The task fingerprint library module includes a feature screening unit and a task fingerprint construction unit; The feature screening unit is used to screen all features in the time domain, frequency domain, and time using the mutual information screening method; The task fingerprint construction unit is used to obtain the task fingerprint through minhash using the screened key features.

[0042] The task conflict judgment module includes a merging judgment unit and a merging unit; The merging judgment unit is used to calculate the similarity difference and judge whether two real-time tasks corresponding to the similarity difference can be merged using the merging threshold; The merging unit is used to merge the real-time tasks that can be judged to be merged into a task matrix.

[0043] The task processing module includes a common part calculation unit and a difference part calculation unit; The common part calculation unit is used to obtain the intersection of all results to get the common part of the results; The difference part calculation unit is used to subtract the common part from each result to obtain the difference part.

[0044] Embodiment: In a rural residence, there are 5 feature phones connected to a cloud server. A task fingerprint database is constructed based on historical task data, and the obtained task types include game types, audio-video types, and document types. The 5 feature phones send task request data packets to the cloud server. The tasks of feature phones 1-5 are downloading music, online translation, game archiving, watching videos, and online music listening respectively. After receiving the task requests, the cloud server calculates the similarity and determines each real-time task type. Among the audio-video types, there are downloading music, watching videos, and online music listening. Calculate the similarity difference, and determine that watching videos and online music listening can be combined to generate a task matrix; the cloud server processes to obtain a result matrix. Extract the common part of the result as audio parsing, and the difference part as video parsing and sound quality improvement. They are respectively compressed and transmitted to the feature phones, and the feature phones reorganize the results to obtain the final result.

[0045] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present invention. Any reference signs in the claims should not be regarded as limiting the claimed rights.

Claims

1. A method for optimizing and managing functional machine data based on cloud applications, characterized in that: The method comprises the following steps: S100, using the 4G network to connect to the cloud server in the feature phone, extracting historical tasks of the feature phone, analyzing and extracting features from the historical task data, constructing task fingerprints, and generating a task fingerprint library; S200, the feature machine initiates a task request in real time. After receiving the task request, the cloud server determines the number of feature machines, calculates the similarity of the real-time tasks using the task fingerprint, and determines the type of the real-time task; and updates the task fingerprint library using the new task type; S300, performing conflict analysis on the determined real-time tasks, merging the determined tasks of the same type to form a task matrix; S400, the cloud server runs the LSTM model once to process the task matrix and output the processing result; extracts the common part and the difference part of the result from the output result, and compresses and transmits the two results; S500, the feature machine receives the task results sent by the cloud server, and the feature machine reorganizes the task results locally to obtain the final result of the real-time task.

2. The method for optimizing and managing functional machine data based on cloud applications according to claim 1, characterized in that: The specific steps of constructing the task fingerprint in S100 are: S101, using 4G network to connect to the cloud server in the feature phone, recording the task log on the feature phone, compressing the task data in the task log and transmitting it to the cloud server, the cloud server pre-processes the task data, specifically: removing missing values ​​in the task data, filtering the task data using a filter; finally, standardizing the task data, the formula is: ； In the formula, s' represents the standardized task data, s represents the task data received by the cloud server, sp represents the average value of the task data, and sz represents the standard deviation of the task data; S102, extract the timestamp of each task data, and calculate the duration of each task. The formula is: In the formula, △t represents the task duration, t end Indicates the task end time, t start Indicates the task start time; takes the duration of each task as the time feature; Calculate the mean, variance, and peak value of each task data in the time domain as the time domain features; In the frequency domain, Fourier transform is used to convert all task data into frequency domain signals s[k], and the amplitude spectrum of the frequency domain signal is calculated in each task data. The formula is: ； In the formula, |s[k]| represents the amplitude spectrum of the frequency domain signal, Re(s[k]) represents the real part of the frequency domain signal, and Im(s[k]) represents the imaginary part of the frequency domain signal; Calculate the frequency domain features of each task data using the formula: ,In the formula, Pt represents the frequency domain characteristics of each task data, argmax represents the independent variable that extracts the maximum value of the function, and k belongs to 0 to N-1; S103, using the mutual information screening method to screen all features in the time domain, frequency domain and time, obtain the key features of each task, and construct a feature vector V with all the key features; using the feature vector of each task to generate a task fingerprint, the formula is: ； In the formula, H represents the task fingerprint, MinHsah represents the minimum hash algorithm, and LSH represents the local sensitive hash algorithm; a 128-bit MinHash fingerprint H is generated, and a task fingerprint H for each task is generated to obtain a task fingerprint library.

3. The method for optimizing and managing functional machine data based on cloud applications according to claim 2, characterized in that: The specific steps of determining whether the tasks are the same in S200 are as follows: S201, all energy supply machines send heartbeat packets to the cloud regularly at time t1. The heartbeat packets contain the function machine ID. The cloud server maintains the online device list based on the regular heartbeat packets. When the device fails to send the heartbeat packet beyond the regular time t1, the cloud server removes the corresponding function machine ID from the online device list and obtains the number of function machines based on the real-time online device list; S202, the online function machine sends a task request data packet to the cloud server, the data packet format is Ta_request={ID, t_request, F i }, where Ta_request represents the task request data packet, t_request represents the task request time, F i Represents the key features of the task; normalize the key features of the task to F i '; The cloud server receives real-time task requests sent by all functional machines; S203, the cloud server compares the received real-time task request with each historical task fingerprint and calculates the similarity. The formula is: ； In the formula, Sim k represents the similarity between the real-time task and the kth task fingerprint, C k represents the kth task fingerprint; the similarity threshold is artificially set to Sy, and the similarity is judged by the similarity threshold. k ≥Sy, it is judged as the same type of task; otherwise, the real-time task is marked as a new task type; Determine the real-time task types of all online functional machines in turn.

4. The method for optimizing and managing functional machine data based on cloud applications according to claim 3 is characterized in that: The specific steps of performing conflict analysis on the determined real-time tasks in S300 are: S301. Count the task types of all real-time tasks and generate a task type distribution set Ta_Distribution={Type1: n1, Type2: n2, ..., TypeK: nK}, where Type1, Type2, ..., TypeK represent the 1st, 2nd, ..., Kth real-time task types, and n1, n2, ..., nK represent the number of real-time tasks in the 1st, 2nd, ..., Kth real-time task types; S302: In the same real-time task type, the difference of similarity of each real-time task is calculated, and the formula is: ； In the formula, Sim c (a, b) represents the similarity difference between real-time task a and real-time task b, Sim a Indicates the similarity of real-time task a, Sim b represents the similarity of real-time task b; similarly, the similarity differences of all real-time tasks under all real-time task types are calculated, the average value and standard deviation of all similarity differences are calculated, and the merge threshold By is obtained by subtracting the standard deviation from the average value; When Sim c When (a, b) ≤ By, it is determined that real-time tasks a and b can be merged. c When (a, b)>By, it is determined that real-time task a and real-time task b cannot be merged; the real-time tasks that are determined to be mergeable are merged into one task matrix.

5. The method for optimizing and managing functional machine data based on cloud applications according to claim 4, characterized in that: The specific steps of extracting the common part of the results and the difference part of the results from the output results in S400 are: S401: In the cloud server, the task matrix is ​​input into the LSTM model for single processing, and the result matrix is ​​output, Results={Y1, Y2, Y3, ..., Y d }, Y1, Y2, Y3, ..., Y d Represents the results of the 1st, 2nd, 3rd, ..., dth tasks in the result matrix; the common part of the calculation result is calculated as follows: ； In the formula, Y common Indicates the public part of the result, Y j Represents the jth task result in the result matrix; The difference in the calculation results is calculated as follows: ； In the formula, Y diff j Represents the difference part of the j-th task result; the cloud server compresses the common part of the result and the difference part separately, and the cloud server performs a global propagation of the common part of the result to all online functional machines, and all online functional machines receive the common part of the result at the same time; The cloud server transmits the difference results to the online function machine of the corresponding ID in sequence according to the task request data packet.

6. The method for optimizing and managing functional machine data based on cloud applications according to claim 5, characterized in that: The specific steps of the functional machine in S500 locally reorganizing the task results to obtain the final result of the real-time task are: S501, the online feature machine receives the two parts of the results transmitted by the cloud server respectively, and the feature machine decompresses the two parts of the results respectively, and reassembles the common part and the difference part of the results. The formula is: ； In the formula, Y j final Represents the final result of the jth task; similarly, all online functional machines decompress and reorganize the common part and the difference part to obtain the final result.

7. A functional machine data optimization management system based on cloud application, characterized by: The functional machine data optimization management system includes a data acquisition module, a task fingerprint library module, a real-time task type judgment module, a task conflict judgment module, a task processing module and a result generation module; The data collection module is used to collect task data in the history of the feature machine; The task fingerprint library module is used to extract the features of the historical task data, screen the key features, construct the task fingerprint of each task, and generate the task fingerprint library; The real-time task type determination module is used to calculate the similarity of the real-time tasks using the task fingerprints and determine the real-time task type; The task conflict judgment module is used to perform conflict analysis on the judged real-time tasks, merge the same tasks judged, and form a task matrix; The task processing module is used for the cloud server to input the task matrix into the LSTM model for single processing, output the result matrix, and divide the result into a common part and a difference part in the result matrix; The result generation module is used for an online functional machine to receive the common part and the difference part, decompress and reassemble to obtain the final result.

8. The cloud application-based functional machine data optimization management system according to claim 7, characterized in that: The task fingerprint library module includes a feature screening unit and a task fingerprint construction unit; The feature screening unit is used to screen all features in the time domain, frequency domain and time using a mutual information screening method; The task fingerprint construction unit is used to obtain the task fingerprint by using the filtered key features through minimum hashing.

9. The cloud application-based functional machine data optimization management system according to claim 7, characterized in that: The task conflict judgment module includes a merging judgment unit and a merging unit; The merging judgment unit is used to calculate the similarity difference and use the merging threshold to judge whether two real-time tasks corresponding to the similarity difference can be merged; The merging unit is used to merge the real-time tasks that are determined to be mergeable into a task matrix.

10. The cloud application-based functional machine data optimization management system according to claim 7, characterized in that: The task processing module includes a common part calculation unit and a difference part calculation unit; The common part calculation unit is used to obtain the intersection of all the results to obtain the common part of the results; The difference part calculation unit is used for subtracting the common part from each result to obtain the difference part.

Images