Two-dimensional structure construction method based on time series data

Abstract

The invention discloses a two-dimensional structure construction method based on time series data, and relates to the technical field of data processing, and the method comprises the steps: firstly, carrying out data cleaning, starting to traverse from a specific position of a number sequence, comparing and exchanging adjacent elements, determining the number of data, and setting the range of rejected data; secondly, a two-dimensional structure data matrix is constructed, collected data have different indexes, a data time sequence period can be set according to reality, and the processed data are constructed into a matrix of a specific scale; then data matrix conversion is carried out, calibration values are determined in multiple ways, the matrix is converted into a matrix of a specific type through calculation, data matrixes of different periods are superposed and summed, and the similarity is calculated, so that a proper time sequence period type is determined; and finally, quantifying the two-dimensional structure data matrix, and calculating characteristic values to determine the characteristics of the monitored object. The method provided by the invention breaks through the limitation of traditional one-dimensional analysis, has many advantages in the aspects of data processing and analysis, and can process time sequence data more comprehensively and accurately.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method for constructing a two-dimensional structure based on time series data. Background Art

[0002] With the continuous advancement of modern ecological and environmental observation methods, a large amount of time-series data has gradually accumulated within various industries. These time-series data are like a huge treasure, containing a lot of industry-related information. In-depth analysis of these large amounts of time-series data is the core link for efficiently solving related problems.

[0003] The traditional analysis method for time series monitoring data is mainly based on the high and low values ​​of one dimension. This analysis mode has obvious limitations. For example, it can only look at data from a single dimension such as the size of the value or the high and low of the trend. This is like observing a complex object from a very narrow perspective and only seeing part of the characteristics of the object. Data at this one-dimensional scale is far from enough to describe objective characteristics. It cannot fully and three-dimensionally reflect the complex objective phenomena represented by the data. For example, in the field of ecological environment monitoring, relying solely on one-dimensional numerical analysis may not accurately judge important information such as the comprehensive trend of environmental changes and the relationship between multiple factors. For example, in industrial production monitoring, one-dimensional analysis may not accurately identify the coordinated change characteristics between multiple parameters in the production process. Therefore, in order to break through this limitation and analyze the characteristics of time series data more comprehensively and deeply, a method for constructing a two-dimensional structure based on time series data is urgently needed, which can analyze and process time series data from a richer dimension and provide more powerful data support for solving related industry problems. Summary of the invention

[0004] The purpose of the present invention is to provide a method for constructing a two-dimensional structure based on time series data, so as to solve the problem in the prior art that data on a one-dimensional scale are insufficient in describing the objective characteristics reflected.

[0005] To achieve the above object, the present invention provides a method for constructing a two-dimensional structure based on time series data, comprising the following steps: Step 1: Collect data, clean the collected data, perform preliminary data processing, regularize the data quantity and determine the exclusion range; Step 2: Construct a two-dimensional structure data matrix; determine the data indicators and time series period, and fill in the data to construct the matrix; Step 3: Data matrix conversion: determine the calibration value, convert the matrix and superimpose and sum it to determine the timing cycle type; Step 4: quantify the two-dimensional structure data matrix and calculate the eigenvalues ​​to determine the characteristics of the monitored object.

[0006] Preferably, the process of step 1 is as follows: S11, traverse from the first element of the collected data to the second to last element; S12. For each position of the outer loop, traverse from the current position to the last element of the sequence; S13. In the inner loop, compare two adjacent elements; if the former is smaller than the latter, swap their positions; S14, determining the number of data N through S11-S13; S15. Eliminate data according to the number N of data determined in S14; the lower bound index of the elimination is floor(0.005×N), and the upper bound index is floor(N−0.005×N)−1, where floor represents a rounding down operation.

[0007] Preferably, the process of constructing the two-dimensional structure data matrix in step 2 is as follows: S21. Construct a two-dimensional matrix, including data indicators and data timing cycle The data indicators According to the actual detection situation, it is divided into multiple detection indicators, which represent the dimensions of the monitoring object calibration; the detection indicators give priority to physical indicators, followed by chemical element indicators; data time series cycle It is a time series periodic structure, and the periodic structure is set according to the time period, such as weekly period, monthly period, quarterly period, etc. S22. According to the two-dimensional matrix structure, the data processed in step 1 is filled to form an m×n data matrix; wherein the number of indicator categories is m, and the number of time periods is n.

[0008] Preferably, the process of data matrix conversion in step 3 is as follows: S31, select the controllable calibration value Q; S32, calculating the data matrix; and The m×n data matrix is ​​different The indicators are divided by the calibration value, if If it is greater than 1, it is recorded as 1, otherwise it is recorded as 0, so that the m×n data matrix is ​​converted into a binary data matrix composed of 0 / 1; S33, superimpose and sum the data matrices of different periods to form the data matrix of FT index, and calculate the data matrix of different periods. and The similarity with the data matrix of the FT indicator, among which the type of data time series period has the greatest similarity.

[0009] Preferably, the controllable calibration value Q is selected according to one of the following methods: The control calibration value Q is the calculated value of the m×n data matrix transformation and is the threshold for feature extraction of the data matrix; The control calibration value Q is determined by selecting the standard value and the data statistical value; The control calibration value Q is the industry index limit, such as the classification limit given in GB3838-2002, etc. Or the limit of the industry pollutant emission standard; The control calibration value Q is obtained by using the statistical change threshold method; The controllable calibration value Q is obtained by calculation through M0.5; The control calibration value Q is obtained by calculating the average value of M; The control calibration value Q is obtained by taking the value of the limit value threshold Q and then performing the judgment condition.

[0010] Preferably, different cycles and Similarity with the data matrix of the FT indicator The calculation expression is as follows: ; in, represents the covariance of vectors F and T, and Represent the standard deviation of vectors F and T respectively; the similarity value range is [-1,1]. The larger the value, the more similar the two vectors are. The more similarity matrices there are, the more stable the two-dimensional structure data matrix is, and the more accurate the description of the monitored object is.

[0011] Preferably, in step 4, the two-dimensional structure data matrix is ​​quantified, and the characteristic value is calculated to determine the characteristics of the monitoring object. Specifically, the characteristic values ​​are calculated and summed according to the rows and columns of the m×n data matrix respectively; and Calibrate the data matrix features and quantitatively describe the features of the monitored objects.

[0012] Therefore, the present invention adopts the above-mentioned method for constructing a two-dimensional structure based on time series data, which has the following beneficial effects: (1) The two-dimensional structure data matrix constructed by the present invention can analyze time series data from multiple dimensions, overcoming the limitations of traditional one-dimensional analysis. Traditional one-dimensional analysis is based only on numerical values ​​and can only reflect single-dimensional information such as numerical value size or trend. The two-dimensional structure constructed by the present invention can present data features more comprehensively, avoiding the defect of one-dimensional analysis that does not fully describe objective features. (2) By superimposing and summing the data matrices of different periods and calculating the similarity, the most suitable time series period type can be determined; this method can improve the accuracy and stability of data processing, thereby improving the accuracy of the description of the monitored object; (3) The data cleaning step can effectively preprocess the raw data. It can remove interference factors such as outliers that may exist in the raw data, improve data quality, and create good conditions for subsequent two-dimensional structure construction and analysis; (4) The quantization process of the two-dimensional structure data matrix, especially the eigenvalue calculation, can achieve a quantitative description of the characteristics of the monitored object; this helps to more accurately grasp the status of the monitored object and its changing trend.

[0013] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is an overall flow chart of a method for constructing a two-dimensional structure based on time series data according to the present invention; Figure 2 A schematic diagram of a two-dimensional structural matrix according to an embodiment of the present invention; Figure 3 A schematic diagram of calculating eigenvalues ​​of a two-dimensional structure data matrix according to an embodiment of the present invention; Figure 4 It is a quantization schematic diagram of a two-dimensional structure data matrix according to an embodiment of the present invention. DETAILED DESCRIPTION

[0015] The following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. 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.

[0016] See also Figure 1-Figure 4 , a method for constructing a two-dimensional structure based on time series data, comprising the following steps: Step 1: Collect data, clean the collected data, perform preliminary data processing, regularize the data quantity and determine the exclusion range; the specific process is as follows: S11, traverse from the first element of the collected data to the second to last element; S12. For each position of the outer loop, traverse from the current position to the last element of the sequence; S13. In the inner loop, compare two adjacent elements; if the former is smaller than the latter, swap their positions; S14, determining the number of data N through S11-S13; S15. Eliminate data according to the number N of data determined in S14; the lower bound index of the elimination is floor(0.005×N), and the upper bound index is floor(N−0.005×N)−1, where floor represents a rounding down operation.

[0017] Step 2: Construct a two-dimensional structure data matrix; determine the data indicators and time series period, fill in the data to construct the matrix; the specific process is as follows: S21. Construct a two-dimensional matrix, including data indicators and data timing cycle The data indicators According to the actual detection situation, it is divided into multiple detection indicators, which represent the dimensions of the monitoring object calibration; the detection indicators give priority to physical indicators, followed by chemical element indicators; data time series cycle It is a time series periodic structure, and the periodic structure is set according to the time period, such as weekly period, monthly period, quarterly period, etc. S22. According to the two-dimensional matrix structure, the data processed in step 1 is filled to form an m×n data matrix; wherein the number of indicator categories is m, and the number of time periods is n.

[0018] Step 3: Data matrix conversion: determine the calibration value, convert the matrix and superimpose and sum it to determine the timing cycle type; the specific process is as follows: S31, selecting a controllable calibration value Q; wherein the controllable calibration value Q is selected according to one of the following methods: The control calibration value Q is the calculated value of the m×n data matrix transformation and is the threshold for feature extraction of the data matrix; The control calibration value Q is determined by selecting the standard value and the data statistical value; The control calibration value Q is the industry index limit, such as the classification limit given in GB3838-2002, etc. Or the limit of the industry pollutant emission standard; The control calibration value Q is obtained by using the statistical change threshold method; for example: A certain indicator data X in the indicator, with a length of n (time period length), forms distribution data ; The controllable calibration value Q is obtained by M0.5 calculation; specifically: Sort by smallest to largest: ; When n is an odd number: ; When n is an even number: ; The control calibration value Q is obtained by calculating the average value of M; the expression is as follows: ; The control calibration value Q is obtained by taking the value of the limit value threshold Q and then performing the judgment condition. The expression is as follows: .

[0019] S32, calculating the data matrix; and The m×n data matrix is ​​different The indicators are divided by the calibration value, if If it is greater than 1, it is recorded as 1, otherwise it is recorded as 0, so that the m×n data matrix is ​​converted into a binary data matrix composed of 0 / 1; S33, superimpose and sum the data matrices of different periods to form the data matrix of FT index, and calculate the data matrix of different periods. and The similarity with the data matrix of the FT indicator, among which the type of data time series period has the greatest similarity. and Similarity with the data matrix of the FT indicator The calculation expression is as follows: ; in, represents the covariance of vectors F and T, and Represent the standard deviation of vectors F and T respectively; the similarity value range is [-1,1]. The larger the value, the more similar the two vectors are. The more similarity matrices there are, the more stable the two-dimensional structure data matrix is, and the more accurate the description of the monitored object is.

[0020] Step 4: quantify the two-dimensional structure data matrix and calculate the eigenvalues ​​to determine the characteristics of the monitored object; specifically, calculate and sum the eigenvalues ​​according to the rows and columns of the m×n data matrix; and Calibrate the data matrix features and quantitatively describe the features of the monitored objects.

[0021] Example Get the original data, as shown in Table 1: Table 1 Original data table ;

[0022] The obtained raw data is cleaned, as shown in Table 2: Table 2 Data table after cleaning ;

[0023] Select 20 days as the time period and construct a two-dimensional data matrix, as shown in Table 3: Table 3 Two-dimensional data matrix ;

[0024] According to the COD standard limit of GB3838, which is 20, the data matrix in 2024 is converted as shown in Table 4: Table 4 Two-dimensional data matrix after transformation in 2024 ;

[0025] Following the same steps above, the two-dimensional data matrix after transformation in 2023 is shown in Table 5, and the two-dimensional data matrix after transformation in 2022 is shown in Table 6: Table 5 Two-dimensional data matrix after transformation in 2023 ;

[0026] Table 6 Two-dimensional data matrix after transformation in 2022 ;

[0027] The data matrix of Table 7 is obtained by superimposing and summing the data matrices of multiple years: Table 7 Two-dimensional data matrix after summation ;

[0028] The average similarity of the multiple data matrices is: 0.63, which means that the 20-day cycle is relatively stable for the construction of data feature structure.

[0029] The characteristic value calculation is shown in Table 8: Table 8 Eigenvalue calculation table ;

[0030] Therefore, the present invention adopts the above-mentioned method for constructing a two-dimensional structure based on time series data, firstly performs data cleaning, starts traversal from a specific position of the sequence, compares and adjusts adjacent elements, determines the total amount of data, and clarifies the data range to be eliminated; then constructs a two-dimensional structure data matrix, the data has a variety of collection indicators, the time series period can be set as needed, and the processed data is constructed into a matrix of corresponding scale; then performs data matrix conversion, the calibration value has a variety of determination methods, and the matrix is ​​converted into a specific form through calculation, and the data matrices of different periods are superimposed and summed and the similarity is calculated to determine the appropriate time series period; finally, the two-dimensional structure data matrix is ​​quantified, and the characteristic value is calculated to clarify the characteristics of the monitored object. This method overcomes the defects of traditional one-dimensional analysis, has multiple advantages in data processing and analysis, can more effectively process time series data, more comprehensively and accurately mine data value, and enhance the ability to analyze and grasp the monitored object.

[0031] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for constructing a two-dimensional structure based on time series data, characterized in that: The following steps are involved: Step 1: Collect data, clean the collected data, perform preliminary data processing, regularize the data quantity and determine the exclusion range; Step 2: Construct a two-dimensional structure data matrix; determine the data indicators and time series period, and fill in the data to construct the matrix; Step 3: Data matrix conversion; Determine the calibration value, transform the matrix and add up the sum to determine the timing cycle type; Step 4: quantify the two-dimensional structure data matrix and calculate the eigenvalues ​​to determine the characteristics of the monitored object.

2. A method for constructing a two-dimensional structure based on time series data according to claim 1, characterized in that: The process for step 1 is as follows: S11, traverse from the first element of the collected data to the second to last element; S12. For each position of the outer loop, traverse from the current position to the last element of the sequence; S13. In the inner loop, compare two adjacent elements; if the former is smaller than the latter, swap their positions; S14, determining the number of data N through S11-S13; S15. Eliminate data according to the number N of data determined in S14; the lower bound index of the elimination is floor(0.005×N), and the upper bound index is floor(N−0.005×N)−1, where floor represents a rounding down operation.

3. The method for constructing a two-dimensional structure based on time series data according to claim 2, characterized in that: The process of constructing the two-dimensional structure data matrix in step 2 is as follows: S21. Construct a two-dimensional matrix, including data indicators and data timing cycle The data indicators According to the actual detection situation, it is divided into multiple detection indicators, which represent the dimensions of the monitoring object calibration; the detection indicators give priority to physical indicators, followed by chemical element indicators; data time series cycle It is a time series periodic structure, and the periodic structure is set according to the time period; S22. According to the two-dimensional matrix structure, the data processed in step 1 is filled to form an m×n data matrix; wherein the number of indicator categories is m, and the number of time periods is n.

4. The method for constructing a two-dimensional structure based on time series data according to claim 3, characterized in that: The process of data matrix conversion in step 3 is as follows: S31, select the controllable calibration value Q; S32, calculating the data matrix; and The m×n data matrix is ​​different The indicators are divided by the calibration value, if If it is greater than 1, it is recorded as 1, otherwise it is recorded as 0, so that the m×n data matrix is ​​converted into a binary data matrix composed of 0 / 1; S33, superimpose and sum the data matrices of different periods to form the data matrix of FT index, and calculate the data matrix of different periods. and The similarity with the data matrix of the FT indicator, among which the type of data time series period has the greatest similarity.

5. The method for constructing a two-dimensional structure based on time series data according to claim 4, characterized in that: Select the controlling calibration value Q according to one of the following methods: The control calibration value Q is the calculated value of the m×n data matrix transformation and is the threshold for feature extraction of the data matrix; The control calibration value Q is determined by selecting the standard value and the data statistical value; The control calibration value Q is the limit value of the industry index or the limit value of the emission standard of the industry pollutant element; The control calibration value Q is obtained by using the statistical change threshold method; The controllable calibration value Q is obtained by calculation through M0.5; The control calibration value Q is obtained by calculating the average value of M; The control calibration value Q is obtained by taking the value of the limit value threshold Q and then performing the judgment condition.

6. A method for constructing a two-dimensional structure based on time series data according to claim 5, characterized in that: Different cycles and Similarity with the data matrix of the FT indicator The calculation expression is as follows: ； in, represents the covariance of vectors F and T, and Represent the standard deviation of vectors F and T respectively; the similarity value range is [-1,1], and the larger the value, the more similar the two vectors are.

7. The method for constructing a two-dimensional structure based on time series data according to claim 6, characterized in that: In step 4, the two-dimensional structure data matrix is ​​quantified, and the characteristic values ​​are calculated to determine the characteristics of the monitored object. Specifically, the characteristic values ​​are calculated and summed according to the rows and columns of the m×n data matrix; from and Calibrate the data matrix features and quantitatively describe the features of the monitored objects.

Images