Time series data preprocessing system and method based on communication network

Abstract

The invention relates to the field of signal processing, in particular to a time sequence data preprocessing system and method based on a communication network, and the system comprises a signal monitoring module, a wave frequency analysis module, a standard fitting module, a sequence processing module and a node scheduling module, the signal monitoring module is used for expanding and storing a time sequence, and the wave frequency analysis module is used for determining a time frequency window; the method comprises the following steps: constructing a standard fitting module, a sequence processing module and a node scheduling module, constructing a cutting sequence and a calculation sub-sequence, outputting a standardized sequence and an error interval by a standard fitting module, filtering a post-sequence sequence, and scheduling an information processing task of a node by a node scheduling module. According to the method, the circulation mode of the time sequence can be predicted more accurately, the integrity and quality of data are improved, abnormal noise in the data is removed, the time sequence is clearer and more interpretable, and the overall information processing efficiency of a communication network is improved.

Description

Technical Field

[0001] The present invention relates to the field of signal processing, and in particular to a time series data preprocessing system and method based on a communication network. Background Art

[0002] A communication network is a system consisting of multiple communication nodes and communication links connecting the nodes. Its purpose is to achieve the transmission and exchange of data information between different nodes. In a communication network, binary time series are often used to transmit signals. A time series is a series of data points arranged in chronological order, reflecting the changes of information over time. Processing time series helps improve data quality and optimize resource allocation between nodes.

[0003] Since the signal is easily mixed with noise during the transmission process, the restored time series signal contains errors when digital sampling, which affects the accuracy of signal transmission. Since the status codes and command statements in the communication network have the same expression form, these errors can be filtered out through signal analysis. However, due to the high instability of the time series, it is difficult for existing processing technologies to accurately distinguish error information.

[0004] In addition, each node in the communication network is composed of computers of different specifications, which essentially have different hardware parameters. There are differences in the processing and adaptability of time series signals. If the time series is not pre-processed, node congestion is likely to occur, affecting the operating efficiency of the communication network. Summary of the invention

[0005] The purpose of the present invention is to provide a time series data preprocessing system and method based on a communication network to solve the problems raised in the above background technology.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a time series data preprocessing system based on a communication network, comprising: a signal monitoring module, a wave frequency analysis module, a standard fitting module, a sequence processing module and a node scheduling module; The signal monitoring module is used to load the monitoring control in the node server of the communication network, obtain the input data and output data of each communication network node, and expand the communication data in a time series, set an information processing device or a signal transmission path at the input and output ports of the node to store the expanded time series, and perform calculation analysis on the time series; The wave frequency analysis module is used to process the expanded time series using a variable window windowing transformation, determine the time frequency window of the time series according to the window interval when the correlation coefficient is the highest, cut the sequence according to the time frequency window to obtain subsequences, perform dimensionless broken line transformation on each subsequence and the previous subsequence, and calculate the composition difference, composition ratio and T-type correlation coefficient of each subsequence according to the information distribution between the subsequences; The standard fitting module is used to calculate the correlation between each subsequence and other subsequences according to the T-type correlation coefficient between the subsequences, retain all sequence sets with correlations less than a threshold to obtain a first reference set, and in the first reference set, discard the set with a sequence number less than the threshold to obtain a second reference set, perform differential fitting on all subsequences in each second reference set, and output a standardized sequence and an observation error interval corresponding to each second reference set; The sequence processing module is used to obtain a post-sequence input or output sequence, cut the sequence according to the time-frequency window, separate the post-sequence subsequence, calculate the correlation between the post-sequence subsequence and each standardized sequence, and when the correlation is within the error interval, perform standardization on the post-sequence to make the post-sequence consistent with the standardized fitting sequence before outputting it, otherwise repeat the time window cutting, and merge the cutting result into the first reference set; The node scheduling module is used to obtain the number of input and output subsequences of each node in the communication network, determine the node's computing speed according to the ratio of the number of input and output subsequences, and perform task-based node management on the communication network according to the node's computing speed, so that the number of input and output sequences of each node maintains a uniform ratio, thereby stabilizing the data processing efficiency of the communication network.

[0007] Further, the signal monitoring module includes: a port control unit and a central processing unit; The port control unit is arranged in the input and output ports of the node server and is used to monitor the time series of the input and output signals; The central processing unit is used to provide computing power support for time series analysis through a terminal data computing chip or a feedback path.

[0008] Furthermore, the wave frequency analysis module includes: a windowing transformation unit, a time-frequency cutting unit and a composition analysis unit; The windowing transformation unit is used to select a window function according to the data information entropy, and perform discrete wavelet transformation on the time series based on the window function; The time-frequency cutting unit is used to calculate the time delay of the window function when the discrete wavelet transform function takes the maximum value, and cut the sequence with the time-frequency window to obtain a subsequence; The composition analysis unit is used to calculate the composition difference, composition ratio and T-type correlation coefficient function of each subsequence according to the information distribution between the subsequences.

[0009] Further, the standard fitting module includes: an association grouping unit and a differential fitting unit; The association grouping unit is used to calculate the association degree of the T-type association coefficients between the subsequences, and group the subsequences according to the calculation results; The differential fitting unit is used to screen the grouping results, remove invalid groups and non-significant groups, and obtain a second reference set.

[0010] Further, the sequence processing module includes: a signal filtering unit and a sequence fusion unit; The signal filtering unit is used to cut the post-sequence and filter the post-sequence according to the corresponding standard sequence information; The sequence fusion unit is used to fuse the filtered post-order subsequences, restore them to time series, and then send them to the communication node for processing.

[0011] Further, the node scheduling module includes: a node testing unit, a task management unit and an efficiency stabilization unit; The node testing unit is used to calculate the information processing speed of the communication node according to the ratio of the information amount of the input subsequence and the output subsequence in the communication node; The task management unit is used to plan the best task allocation method according to the data volume of the signal to be transmitted and the information processing speed of each node; The efficiency stabilization unit is used to calculate the overall efficiency of the communication network in real time and upload the information volume of the communication network input and output time series in real time.

[0012] The time series data preprocessing method based on the communication network includes the following steps: Step S1. Listen to input and output signals at the node server port, expand the signals in the form of time series, calculate the information entropy of the time series, select a window function according to the information entropy, perform discrete wavelet transform on the time series with the window function as the base frequency, and obtain a discrete wavelet transform function; Step S2. Calculate the time delay of the window function when the discrete wavelet transform function takes the maximum value, record it as the time-frequency window, use the time-frequency window as the cutting time series, obtain the subsequence of fixed code element length, perform dimensionless broken line transformation on each subsequence, and calculate the composition difference, composition ratio and T-type correlation coefficient of each subsequence; Step S3. Calculate the T-type correlation coefficient between each subsequence and other subsequences, classify all subsequences whose mutual correlation is less than the threshold into one category, obtain a first reference set, discard the set with the number of elements less than the threshold, and obtain a second reference set; Step S4. Superimpose and sample the subsequences in each reference set to obtain a standardized sequence and an error interval, cut the subsequent time series according to the time-frequency window, calculate the T-type correlation coefficient between the subsequences of the subsequent time series and each standardized sequence, and when the correlation is within the error interval, perform standardization on each subsequent subsequence; Step S5. The information processing speed of the communication node is represented by the information volume ratio of the input subsequence and the output subsequence in the communication node. Tasks are allocated according to the data volume of the signal to be transmitted and the information processing speed of each node to maximize the overall efficiency of the communication network.

[0013] Further, step S1 includes: Step S11. Loading a monitoring control in a node server of the communication network, monitoring input and output signals, and transmitting the input and output signals to a processing chip through a terminal data operation chip or a feedback path; Step S12. Expand the monitored input and output signals into a time series in the processing chip. The time series is a finite length sequence, and all elements in the sequence are binary. Calculate the information entropy according to the length of the time series, and select a window function from a preloaded function library according to the information entropy. Step S13. Perform discrete wavelet transform on the time series with the window function as the base frequency. The transform function is: ; Among them, f(a, b) is the discrete wavelet transform function, a is the scale parameter, b is the delay parameter, t0 is the sampling interval of the time series, δ(k) is the window function, E(k) is the kth element value in the time series, N is the number of elements in the time series, and k is the sequence number.

[0014] Further, step S2 includes: Step S21. Limit the scale parameter to a period length range of the window function, calculate the maximum value of the discrete wavelet transform function, and output the size of the delay parameter b when the function takes the maximum value, which is recorded as the time-frequency window; Step S22. Use the time-frequency window to cut the time series so that each subsequence contains m elements. If there is a sequence whose length is less than m, it is supplemented with a low-level signal. The subsequences are numbered in chronological order, and the preamble sequence of each subsequence is determined in descending order of the number. The initial subsequence uses the last subsequence as the preamble sequence; Perform dimensionless broken line transformation on each subsequence and calculate the composition difference, composition ratio and T-type correlation coefficient of each subsequence: ; Among them, z represents the composition difference between the subsequence and the previous sequence, z1(x) represents the difference between the xth element and the x+1th element in the previous subsequence, z2(x) represents the difference between the xth element and the x+1th element in the current subsequence, s represents the composition ratio of the subsequence to the previous sequence, min and max represent the maximum and minimum functions respectively, and r represents the T-type correlation coefficient.

[0015] Further, step S3 includes: Step S31. Calculate the T-type correlation coefficients between all subsequences and other sequences, divide the subsequences into a group, make the correlation coefficients between every two sequences in the group smaller than a threshold, store all the groups into a set, and obtain a first reference set; Step S32: All groups in the first reference set whose number of included sequences is less than a threshold are screened out to obtain a second reference set.

[0016] Further, step S4 includes: Step S41. For each group in the second reference set, calculate the average value of each element in all subsequences in the group. If the average value is greater than the sampling standard, it is compiled into a high-level signal. If the average value is less than the sampling standard, it is compiled into a low-level signal. Obtain all compiled signals and combine them into a standardized sequence. The error interval is [-Σ (x=1,m) |z(x)-z0(x)|,Σ (x=1,m) |z(x)-z0(x)|], where z(x) represents the x-th element in the normalized sequence and z0(x) represents the average of the x-th elements of all subsequences in the group; Step S42. When the subsequent time series is transmitted in the communication node, the subsequent time series is cut according to the time-frequency window to obtain the subsequences of the subsequent time series, and the T-type correlation coefficient between each subsequence of the subsequent time series and the standardized sequence is calculated. If the correlation is within the error interval, the subsequence of the subsequent time series is replaced by the standardized sequence, and then restored and output.

[0017] Further, step S5 includes: Step S51. Determine the information processing speed of the communication node according to the ratio of the amount of information of the input subsequence and the output subsequence in the communication node, and identify the speed of each node in the communication network; Step S52. Calculate the amount of information of the data to be processed, and select a scheduling algorithm to distribute traffic among nodes. The scheduling algorithm includes: dynamic weighted scheduling, hybrid scheduling, neural network scheduling, and batch scheduling.

[0018] Compared with the prior art, the beneficial effects achieved by the present invention are: The present invention expands the communication data into a time series, adopts a windowing transformation of a variable window, determines the time-frequency window of the time series, cuts the sequence according to the time-frequency window, performs a dimensionless broken line transformation on each sequence and the previous sequence, and uses the T-type correlation coefficient of the technical sequence to identify long-term trends, cyclical changes and seasonal patterns in the data, and can more accurately predict the cyclical pattern of the time series, thereby improving the integrity and quality of the data.

[0019] The invention calculates the correlation degree according to the T-type correlation coefficient, retains all sequence sets with correlation degrees less than a threshold, and when the subsequent sequence input is within the error interval of the standardized sequence, the subsequent sequence is output after standardized preprocessing. The smoothing technology is used to identify and remove abnormal noise in the data, making the time series clearer and more interpretable.

[0020] The present invention determines the computing speed of the node according to the input-output standardized sequence number of each node and the input-output ratio, and performs task-based node management on the communication network, so that the input-output sequence number of each node maintains a uniform ratio, thereby stabilizing the data processing efficiency of the communication network, better understanding the dynamic information processing state of the communication network, and improving the overall information processing efficiency of the communication network. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1 It is a structural schematic diagram of a time series data preprocessing system based on a communication network of the present invention; Figure 2 It is a schematic diagram of the steps of the time series data preprocessing method based on the communication network of the present invention. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] See also Figure 1 ,The present invention provides a technical solution: a time series data preprocessing system based on a communication network, comprising: a signal monitoring module, a wave frequency analysis module, a standard fitting module, a sequence processing module and a node scheduling module; The signal monitoring module is used to load the monitoring control in the node server of the communication network, obtain the input data and output data of each communication network node, and expand the communication data in a time series, set an information processing device or a signal transmission path at the input and output ports of the node to store the expanded time series, and perform calculation analysis on the time series; The signal monitoring module includes: a port control unit and a central processing unit; The port control unit is arranged in the input and output ports of the node server and is used to monitor the time series of the input and output signals; The central processing unit is used to provide computing power support for time series analysis through a terminal data computing chip or a feedback path.

[0024] The wave frequency analysis module is used to process the expanded time series using a variable window windowing transformation, determine the time frequency window of the time series according to the window interval when the correlation coefficient is the highest, cut the sequence according to the time frequency window to obtain subsequences, perform dimensionless broken line transformation on each subsequence and the previous subsequence, and calculate the composition difference, composition ratio and T-type correlation coefficient of each subsequence according to the information distribution between the subsequences; The wave frequency analysis module includes: a windowing transformation unit, a time-frequency cutting unit and a composition analysis unit; The windowing transformation unit is used to select a window function according to the data information entropy, and perform discrete wavelet transformation on the time series based on the window function; The time-frequency cutting unit is used to calculate the time delay of the window function when the discrete wavelet transform function takes the maximum value, and cut the sequence with the time-frequency window to obtain a subsequence; The composition analysis unit is used to calculate the composition difference, composition ratio and T-type correlation coefficient function of each subsequence according to the information distribution between the subsequences.

[0025] The standard fitting module is used to calculate the correlation between each subsequence and other subsequences according to the T-type correlation coefficient between the subsequences, retain all sequence sets with correlations less than a threshold to obtain a first reference set, and in the first reference set, discard the set with a sequence number less than the threshold to obtain a second reference set, perform differential fitting on all subsequences in each second reference set, and output a standardized sequence and an observation error interval corresponding to each second reference set; The standard fitting module includes: an association grouping unit and a differential fitting unit; The association grouping unit is used to calculate the association degree of the T-type association coefficients between the subsequences, and group the subsequences according to the calculation results; The differential fitting unit is used to screen the grouping results, remove invalid groups and non-significant groups, and obtain a second reference set.

[0026] The sequence processing module is used to obtain a post-sequence input or output sequence, cut the sequence according to the time-frequency window, separate the post-sequence subsequence, calculate the correlation between the post-sequence subsequence and each standardized sequence, and when the correlation is within the error interval, perform standardization on the post-sequence to make the post-sequence consistent with the standardized fitting sequence before outputting it, otherwise repeat the time window cutting, and merge the cutting result into the first reference set; The sequence processing module includes: a signal filtering unit and a sequence fusion unit; The signal filtering unit is used to cut the post-sequence and filter the post-sequence according to the corresponding standard sequence information; The sequence fusion unit is used to fuse the filtered post-order subsequences, restore them to time series, and then send them to the communication node for processing.

[0027] The node scheduling module is used to obtain the number of input and output subsequences of each node in the communication network, determine the node's computing speed according to the ratio of the number of input and output subsequences, and perform task-based node management on the communication network according to the node's computing speed, so that the number of input and output sequences of each node maintains a uniform ratio, thereby stabilizing the data processing efficiency of the communication network.

[0028] The node scheduling module includes: a node testing unit, a task management unit and an efficiency stabilization unit; The node testing unit is used to calculate the information processing speed of the communication node according to the ratio of the information amount of the input subsequence and the output subsequence in the communication node; The task management unit is used to plan the best task allocation method according to the data volume of the signal to be transmitted and the information processing speed of each node; The efficiency stabilization unit is used to calculate the overall efficiency of the communication network in real time and upload the information volume of the communication network input and output time series in real time.

[0029] like Figure 2 As shown, the time series data preprocessing method based on the communication network includes the following steps: Step S1. Listen to input and output signals at the node server port, expand the signals in the form of time series, calculate the information entropy of the time series, select a window function according to the information entropy, perform discrete wavelet transform on the time series with the window function as the base frequency, and obtain a discrete wavelet transform function; Step S1 includes: Step S11. Loading a monitoring control in a node server of the communication network, monitoring input and output signals, and transmitting the input and output signals to a processing chip through a terminal data operation chip or a feedback path; Step S12. Expand the monitored input and output signals into a time series in the processing chip. The time series is a finite length sequence, and all elements in the sequence are binary. Calculate the information entropy according to the length of the time series, and select a window function from a preloaded function library according to the information entropy. Step S13. Perform discrete wavelet transform on the time series with the window function as the base frequency. The transform function is: ; Among them, f(a, b) is the discrete wavelet transform function, a is the scale parameter, b is the delay parameter, t0 is the sampling interval of the time series, δ(k) is the window function, E(k) is the kth element value in the time series, N is the number of elements in the time series, and k is the sequence number.

[0030] Step S2. Calculate the time delay of the window function when the discrete wavelet transform function takes the maximum value, record it as the time-frequency window, use the time-frequency window as the cutting time series, obtain the subsequence of fixed code element length, perform dimensionless broken line transformation on each subsequence, and calculate the composition difference, composition ratio and T-type correlation coefficient of each subsequence; Step S2 includes: Step S21. Limit the scale parameter to a period length range of the window function, calculate the maximum value of the discrete wavelet transform function, and output the size of the delay parameter b when the function takes the maximum value, which is recorded as the time-frequency window; Step S22. Use the time-frequency window to cut the time series so that each subsequence contains m elements. If there is a sequence whose length is less than m, it is supplemented with a low-level signal. The subsequences are numbered in chronological order, and the preamble sequence of each subsequence is determined in descending order of the number. The initial subsequence uses the last subsequence as the preamble sequence; Perform dimensionless broken line transformation on each subsequence and calculate the composition difference, composition ratio and T-type correlation coefficient of each subsequence: ; Among them, z represents the composition difference between the subsequence and the previous sequence, z1(x) represents the difference between the xth element and the x+1th element in the previous subsequence, z2(x) represents the difference between the xth element and the x+1th element in the current subsequence, s represents the composition ratio of the subsequence to the previous sequence, min and max represent the maximum and minimum functions respectively, and r represents the T-type correlation coefficient.

[0031] Step S3. Calculate the T-type correlation coefficient between each subsequence and other subsequences, classify all subsequences whose mutual correlation is less than the threshold into one category, obtain a first reference set, discard the set with the number of elements less than the threshold, and obtain a second reference set; Step S3 includes: Step S31. Calculate the T-type correlation coefficients between all subsequences and other sequences, divide the subsequences into a group, make the correlation coefficients between every two sequences in the group smaller than a threshold, store all the groups into a set, and obtain a first reference set; Step S32: All groups in the first reference set whose number of included sequences is less than a threshold are screened out to obtain a second reference set.

[0032] Step S4. Superimpose and sample the subsequences in each reference set to obtain a standardized sequence and an error interval, cut the subsequent time series according to the time-frequency window, calculate the T-type correlation coefficient between the subsequences of the subsequent time series and each standardized sequence, and when the correlation is within the error interval, perform standardization on each subsequent subsequence; Step S4 includes: Step S41. For each group in the second reference set, calculate the average value of each element in all subsequences in the group. If the average value is greater than the sampling standard, it is compiled into a high-level signal. If the average value is less than the sampling standard, it is compiled into a low-level signal. Obtain all compiled signals and combine them into a standardized sequence. The error interval is [-Σ (x=1,m) |z(x)-z0(x)|,Σ (x=1,m) |z(x)-z0(x)|], where z(x) represents the x-th element in the normalized sequence and z0(x) represents the average of the x-th elements of all subsequences in the group; Step S42. When the subsequent time series is transmitted in the communication node, the subsequent time series is cut according to the time-frequency window to obtain the subsequences of the subsequent time series, and the T-type correlation coefficient between each subsequence of the subsequent time series and the standardized sequence is calculated. If the correlation is within the error interval, the subsequence of the subsequent time series is replaced by the standardized sequence, and then restored and output.

[0033] Step S5. The information processing speed of the communication node is represented by the information volume ratio of the input subsequence and the output subsequence in the communication node. Tasks are allocated according to the data volume of the signal to be transmitted and the information processing speed of each node to maximize the overall efficiency of the communication network.

[0034] Step S5 includes: Step S51. Determine the information processing speed of the communication node according to the ratio of the amount of information of the input subsequence and the output subsequence in the communication node, and identify the speed of each node in the communication network; Step S52. Calculate the amount of information of the data to be processed, and select a scheduling algorithm to distribute traffic among nodes. The scheduling algorithm includes: dynamic weighted scheduling, hybrid scheduling, neural network scheduling, and batch scheduling.

[0035] Embodiment: The time series of the input communication node is [10110110]. After windowing transformation, the time domain window is 3, then the sequence is cut into three subsequences of

[101] ,

[101] and

[100] .

[101] and

[101] are grouped together,

[100] is grouped together, and the group of

[100] is filtered out. When the post-sequence signal input is

[111] , the post-sequence signal is standardized to

[101] output.

[0036] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0037] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A time series data preprocessing method based on a communication network, characterized in that: The method comprises the following steps: Step S1. Listen to input and output signals at the node server port, expand the signals in the form of time series, calculate the information entropy of the time series, select a window function according to the information entropy, perform discrete wavelet transform on the time series with the window function as the base frequency, and obtain a discrete wavelet transform function; Step S2. Calculate the time delay of the window function when the discrete wavelet transform function takes the maximum value, record it as the time-frequency window, use the time-frequency window as the cutting time series, obtain the subsequence of fixed code element length, perform dimensionless broken line transformation on each subsequence, and calculate the composition difference, composition ratio and T-type correlation coefficient of each subsequence; Step S3. Calculate the T-type correlation coefficient between each subsequence and other subsequences, classify all subsequences whose mutual correlation is less than the threshold into one category, obtain a first reference set, discard the set with the number of elements less than the threshold, and obtain a second reference set; Step S4. Superimpose and sample the subsequences in each reference set to obtain a standardized sequence and an error interval, cut the subsequent time series according to the time-frequency window, calculate the T-type correlation coefficient between the subsequences of the subsequent time series and each standardized sequence, and when the correlation is within the error interval, perform standardization on each subsequent subsequence; Step S5. The information processing speed of the communication node is represented by the information volume ratio of the input subsequence and the output subsequence in the communication node. Tasks are allocated according to the data volume of the signal to be transmitted and the information processing speed of each node to maximize the overall efficiency of the communication network.

2. The method for preprocessing time series data based on a communication network according to claim 1, characterized in that: Step S1 includes: Step S11. Loading a monitoring control in a node server of the communication network, monitoring input and output signals, and transmitting the input and output signals to a processing chip through a terminal data operation chip or a feedback path; Step S12. Expand the monitored input and output signals into a time series in the processing chip. The time series is a finite length sequence, and all elements in the sequence are binary. Calculate the information entropy according to the length of the time series, and select a window function from a preloaded function library according to the information entropy. Step S13. Perform discrete wavelet transform on the time series with the window function as the base frequency. The discrete wavelet transform function is: ； Among them, f(a, b) is the discrete wavelet transform function, a is the scale parameter, b is the delay parameter, t0 is the sampling interval of the time series, δ(k) is the window function, E(k) is the kth element value in the time series, N is the number of elements in the time series, and k is the sequence number.

3. The method for preprocessing time series data based on communication network according to claim 2, characterized in that: Step S2 includes: Step S21. Limit the scale parameter to a period length range of the window function, calculate the maximum value of the discrete wavelet transform function, and output the size of the delay parameter b when the function takes the maximum value, which is recorded as the time-frequency window; Step S22. Use the time-frequency window to cut the time series so that each subsequence contains m elements. If there is a sequence whose length is less than m, it is supplemented with a low-level signal. The subsequences are numbered in chronological order, and the preamble sequence of each subsequence is determined in descending order of the number. The initial subsequence uses the last subsequence as the preamble sequence; Perform dimensionless broken line transformation on each subsequence and calculate the composition difference, composition ratio and T-type correlation coefficient of each subsequence: ； Among them, z represents the composition difference between the subsequence and the previous sequence, z1(x) represents the difference between the xth element and the x+1th element in the previous subsequence, z2(x) represents the difference between the xth element and the x+1th element in the current subsequence, s represents the composition ratio of the subsequence to the previous sequence, min and max represent the maximum and minimum functions respectively, and r represents the T-type correlation coefficient.

4. The method for preprocessing time series data based on communication network according to claim 3, characterized in that: Step S3 includes: Step S31. Calculate the T-type correlation coefficients between all subsequences and other sequences, divide the subsequences into a group, make the correlation coefficients between every two sequences in the group smaller than a threshold, store all the groups into a set, and obtain a first reference set; Step S32. All groups in the first reference set containing sequences whose number is less than a threshold are screened out to obtain a second reference set; Step S4 includes: Step S41. For each group in the second reference set, calculate the average value of each element in all subsequences in the group. If the average value is greater than the sampling standard, it is compiled into a high-level signal. If the average value is less than the sampling standard, it is compiled into a low-level signal. Obtain all compiled signals and combine them into a standardized sequence. The error interval is [-Σ (x=1,m) |z(x)-z0(x)|,Σ (x=1,m) |z(x)-z0(x)|], where z(x) represents the x-th element in the normalized sequence and z0(x) represents the average of the x-th elements of all subsequences in the group; Step S42. When the subsequent time series is transmitted in the communication node, the subsequent time series is cut according to the time-frequency window to obtain the subsequences of the subsequent time series, and the T-type correlation coefficient between each subsequence of the subsequent time series and the standardized sequence is calculated. If the correlation is within the error interval, the subsequence of the subsequent time series is replaced by the standardized sequence, and then restored and output.

5. The method for preprocessing time series data based on communication network according to claim 4, characterized in that: Step S5 includes: Step S51. Determine the information processing speed of the communication node according to the ratio of the amount of information of the input subsequence and the output subsequence in the communication node, and identify the speed of each node in the communication network; Step S52. Calculate the amount of information of the data to be processed, and select a scheduling algorithm to distribute traffic among nodes. The scheduling algorithm includes: dynamic weighted scheduling, hybrid scheduling, neural network scheduling, and batch scheduling.

6. A time series data preprocessing system based on a communication network, characterized in that: The system includes the following modules: signal monitoring module, wave frequency analysis module, standard fitting module, sequence processing module and node scheduling module; The signal monitoring module is used to load the monitoring control in the node server of the communication network, obtain the input data and output data of each communication network node, and expand the communication data in a time series, set an information processing device or a signal transmission path at the input and output ports of the node to store the expanded time series, and perform calculation analysis on the time series; The wave frequency analysis module is used to process the expanded time series using a variable window windowing transformation, determine the time frequency window of the time series according to the window interval when the correlation coefficient is the highest, cut the sequence according to the time frequency window to obtain subsequences, perform dimensionless broken line transformation on each subsequence and the previous subsequence, and calculate the composition difference, composition ratio and T-type correlation coefficient of each subsequence according to the information distribution between the subsequences; The standard fitting module is used to calculate the correlation between each subsequence and other subsequences according to the T-type correlation coefficient between the subsequences, retain all sequence sets with correlations less than a threshold to obtain a first reference set, and in the first reference set, discard the set with a sequence number less than the threshold to obtain a second reference set, perform differential fitting on all subsequences in each second reference set, and output a standardized sequence and an observation error interval corresponding to each second reference set; The sequence processing module is used to obtain a post-sequence input or output sequence, cut the sequence according to the time-frequency window, separate the post-sequence subsequence, calculate the correlation between the post-sequence subsequence and each standardized sequence, and when the correlation is within the error interval, perform standardization on the post-sequence to make the post-sequence consistent with the standardized fitting sequence before outputting it, otherwise repeat the time window cutting, and merge the cutting result into the first reference set; The node scheduling module is used to obtain the number of input and output subsequences of each node in the communication network, determine the node's computing speed according to the ratio of the number of input and output subsequences, and perform task-based node management on the communication network according to the node's computing speed, so that the number of input and output sequences of each node maintains a uniform ratio, thereby stabilizing the data processing efficiency of the communication network.

7. The time series data preprocessing system based on communication network according to claim 6, characterized in that: The signal monitoring module includes: a port control unit and a central processing unit; The port control unit is arranged in the input and output ports of the node server and is used to monitor the time series of the input and output signals; The central processing unit is used to provide computing power support for time series analysis through a terminal data computing chip or a feedback path.

8. The time series data preprocessing system based on communication network according to claim 7, characterized in that: The wave frequency analysis module includes: a windowing transformation unit, a time-frequency cutting unit and a composition analysis unit; The windowing transformation unit is used to select a window function according to the data information entropy, and perform discrete wavelet transformation on the time series based on the window function; The time-frequency cutting unit is used to calculate the time delay of the window function when the discrete wavelet transform function takes the maximum value, and cut the sequence with the time-frequency window to obtain a subsequence; The composition analysis unit is used to calculate the composition difference, composition ratio and T-type correlation coefficient function of each subsequence according to the information distribution between the subsequences.

9. The time series data preprocessing system based on communication network according to claim 8, characterized in that: The standard fitting module includes: an association grouping unit and a differential fitting unit; The association grouping unit is used to calculate the association degree of the T-type association coefficients between the subsequences, and group the subsequences according to the calculation results; The differential fitting unit is used to screen the grouping results, remove invalid groups and non-significant groups, and obtain a second reference set; The sequence processing module includes: a signal filtering unit and a sequence fusion unit; The signal filtering unit is used to cut the post-sequence and filter the post-sequence according to the corresponding standard sequence information; The sequence fusion unit is used to fuse the filtered post-order subsequences, restore them to time series, and then send them to the communication node for processing.

10. The time series data preprocessing system based on communication network according to claim 9, characterized in that: The node scheduling module includes: a node testing unit, a task management unit and an efficiency stabilization unit; The node testing unit is used to calculate the information processing speed of the communication node according to the ratio of the information amount of the input subsequence and the output subsequence in the communication node; The task management unit is used to plan the best task allocation method according to the data volume of the signal to be transmitted and the information processing speed of each node; The efficiency stabilization unit is used to calculate the overall efficiency of the communication network in real time and upload the information volume of the communication network input and output time series in real time.

Images