Method for extracting time sequence trend features of roll and package sorting business indexes
Through wavelet transform and piecewise linear representation of important points, the trend characteristics of the package sorting business time series are extracted, which solves the problem of data nonlinearity and noise interference, and improves fault warning and the controllability of the production process.
Patent Information
- Application Number
- CN202411656998.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-19
- Publication Date
- 2025-01-24
AI Technical Summary
Due to its high-dimensionality, large data volume, nonlinear characteristics, mutations, interruptions and discontinuous points, the time series data of the package sorting business is difficult to effectively analyze and warn, resulting in sorter failures and interruptions in production processes.
Wavelet transform and important point piecewise linear representation are used to extract the trend characteristics of the time series through time series normalization, important point segmentation, wavelet decomposition and reconstruction, piecewise linear representation, global and local factor calculation, and comprehensive factor weighting.
Effectively capture the local characteristics of signals, weaken noise interference, take into account the overall and local information of the time sequence, improve the accuracy of the extraction of time sequence change trends, reduce the amount of indicator data processing in the production process, improve the accuracy of abnormal alarm and the production process of the production process. Controlt.
Smart Images
Figure CN119357639A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of time series trend feature extraction, and more specifically, to a method for extracting time series trend features of roll and package sorting business indicators. Background Art
[0002] The roll and pack logistics system of cigarette manufacturing enterprises involves multiple production links, including scanning of barcodes by the case sealing machine, data transmission by the finished product warehouse scheduling, scanning of barcodes by the sorting machine, barcode transmission by the electric control PLC, and sorting action by the sorting machine. Among the numerous production links, the roll and pack sorting link is currently highly sensitive to abnormalities and prone to failures. The influence of information system-related factors (such as network jitter, high database read and write delay, disk IO delay, etc.) may cause the barcode scanner of the sorting machine to fail to read the barcode normally, and the sorting action cannot be performed, causing cigarettes to flow back in the sorting channel, blocking and interrupting the normal production process.
[0003] In the process of early warning management of the backflow failure of roll and package sorting, it is necessary to conduct real-time analysis of key performance and status indicators, and to warn and judge the occurrence of failures based on the fluctuation state of the time series indicators. The indicator data of the roll and package sorting business is a series of real-time monitoring values arranged according to time, that is, a time series. A time series is a set of ordered data that are arranged in chronological order and are interrelated. Time series data has the characteristics of high dimension and large data volume. Due to the production characteristics of the roll and package sorting business, the key status indicator data presents obvious nonlinear characteristics, and the data has mutations, interruptions and discontinuities according to changes in production equipment and tasks, which poses greater challenges to the analysis and early warning of such time series data.
[0004] Therefore, there is an urgent need for a method to extract time series trend features of roll and package sorting business indicators. Summary of the invention
[0005] The purpose of the present invention is to provide a method for extracting trend features of time series of roll and package sorting business indicators to solve the problems in the above-mentioned prior art, which can better capture the local characteristics of the signal, effectively weaken the noise interference, take into account the global and local information of the time series, and realize the effective extraction of the changing trend of the time series.
[0006] The present invention provides a method for extracting time series trend features of roll and package sorting business indicators, which includes:
[0007] Step S1, time series normalization: read the time series, remove outliers in a given time series, and represent the time series after removing outliers with symbols to obtain an initial time series;
[0008] Step S2, obtaining an initial important point sequence: segmenting the original time series based on the definition of important points to obtain an initial important point sequence;
[0009] Step S3, reconstructing the initial time series: based on the initial time series in step S1, performing wavelet decomposition and reconstruction to obtain a reconstructed time series;
[0010] Step S4, reconstructing the important point sequence: segmenting the reconstructed time series in step S3 based on the piecewise linear representation of the important points to obtain a reconstructed important point sequence;
[0011] Step S5, calculating the global factor: based on the reconstructed important point sequence in step S4, the initial important point sequence in step S2 and the initial time sequence in step S1, comparing the important points in the reconstructed important point sequence that are the same as those in the initial important point sequence, to obtain the global factor value of the important point;
[0012] Step S6, calculating the local factor: calculating the distance from the important point to the endpoint in the important point sequence and the reconstructed important point sequence according to the orthogonal distance;
[0013] Step S7, calculating the comprehensive factor: weighting the global factor value in step S5 and the local factor value in step S6 to calculate the comprehensive factor value;
[0014] Step S8, obtaining the final important point sequence: based on the comprehensive factor value in step S7, sorting the important point sequence, selecting segmentation points from the sorted important point sequence to obtain the final important point sequence;
[0015] Step S9, trend feature extraction: Use straight line segments to connect the final important point sequence in step S8 to obtain a linear representation of the time series, and complete the trend feature extraction.
[0016] The method for extracting trend features of time series of indicators of roll and package sorting business as described above, wherein, preferably, after step S1 and before step S2, the method for extracting trend features of time series of indicators of roll and package sorting business further includes:
[0017] Step S11: Eliminate missing values and outliers in the initial time series.
[0018] As described above, in the method for extracting trend features of time series of roll and package sorting business indicators, preferably, the initial important point sequence in step S2 is composed of the endpoints of the time series and the internal discrete local extreme points.
[0019] In the method for extracting trend features of time series of indicators of roll and package sorting business as described above, preferably, the step S3, reconstructing the initial time series, specifically comprises the following steps:
[0020] Step S31, using the db series wavelet function to decompose the time series k times to perform k-level decomposition, and obtain 1 approximate sequence and k detail sequences;
[0021] Step S32: Use the db series wavelet function to perform Reconstructed, we get The reconstructed time series is: Indicates rounding down;
[0022] Step S33: draw a reconstructed timing diagram.
[0023] In the method for extracting trend features of time series of indicators of roll and package sorting business as described above, preferably, the step S4, reconstructing the important point sequence, specifically includes:
[0024] The PLR-IPS-based method represents the time series in segments so that the fitting error after segmentation is minimized under a preset compression rate.
[0025] In the method for extracting time series trend features of roll and package sorting service indicators as described above, preferably, the step S5, calculating the global factor, specifically includes:
[0026] Set the initial value of the reconstruction parameter θ to , and gradually decreases to 0 with a tolerance of 1. As the reconstruction parameter θ decreases, the components involved in the reconstruction become fewer, resulting in a decrease in the number of important points for reconstruction. Therefore, the global factor will increase. The global factor value is an indicator used to evaluate the overall dimension of the time series.
[0027] In the method for extracting time series trend features of roll and package sorting service indicators as described above, preferably, the step S6, calculating the local factors, specifically includes:
[0028] The distance factor is used to measure the relative difference of important points in a local range. Points with larger distance factors have larger differences with adjacent points and are more likely to be important turning points. Points with smaller distance factors may be interference extreme points. By selecting an appropriate threshold, interference points can be excluded.
[0029] In the method for extracting time series trend features of roll and package sorting service indicators as described above, preferably, the step S7, calculating the comprehensive factor, specifically includes:
[0030] Step S71: Use genetic algorithm to determine model parameters as the weight of distance factor, and select the average fitting error The smallest set of parameter values is used as the final model parameters;
[0031] Step S72: weight the global factor in step S5 and the local factor in step S6 according to the determined model parameters to calculate a comprehensive factor value.
[0032] The method for extracting trend features of time series of indicators of roll and package sorting business as described above, wherein, preferably, after step S8 and before step S9, the method for extracting trend features of time series of indicators of roll and package sorting business further includes:
[0033] Step S81: filter the sorted important point sequence according to the requirements, set the number of segments N, select the first N+1 important points, and obtain the final segmentation point sequence and segmentation position.
[0034] The invention provides a method for extracting trend features of time series of indicators of roll and package sorting business. The method extracts trend features of time series based on wavelet transform and important points, makes full use of the characteristic that wavelet transform can provide information in time domain and frequency domain at the same time, can better capture local features of signals, effectively weaken noise interference, proposes a comprehensive factor evaluation method based on distance factor and global factor, takes into account global and local information of time series, and realizes effective extraction of time series change trend; makes full use of time series data collected in the existing production environment for analysis, reduces implementation difficulty through data feature extraction method, has low requirements on computing power, thus has small hardware investment in processing equipment, is simple to operate, and has high efficiency in fault feature screening; effectively reduces the processing amount of indicator data continuously generated in the production process through dimensionality reduction and denoising of key indicator data, and improves the accuracy of abnormal alarm; abnormal small fluctuations can be discovered in time through the final trend chart, better matches the nonlinear characteristics of indicators of the roll and package production process, and improves the controllability of the production process; can be directly applied to the existing production line, the indicator time series data are all monitoring data of the production-related environment, does not involve the sensor installation of the original earlier production line and the real-time transmission of production line data, avoids production line transformation, and reduces installation costs. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be further described below with reference to the accompanying drawings, in which:
[0036] Figure 1 A flow chart of an embodiment of a method for extracting time series trend characteristics of roll and package sorting service indicators provided by the present invention;
[0037] Figure 2 Schematic diagram for normalized time series and important point marking;
[0038] Figure 3 This is a schematic diagram of decomposing and reconstructing the original time series using wavelet function;
[0039] Figure 4It is a schematic diagram of the important point sequence after reconstruction based on PLR-IPS;
[0040] Figure 5 It is a schematic diagram of the ranking of important points based on comprehensive factors;
[0041] Figure 6 This is a schematic diagram of the final trend linear representation. DETAILED DESCRIPTION
[0042] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. The description of the exemplary embodiments is merely illustrative and is in no way intended to limit the present disclosure and its application or use. The present disclosure can be implemented in many different forms and is not limited to the embodiments described herein. These embodiments are provided to make the present disclosure thorough and complete and to fully express the scope of the present disclosure to those skilled in the art. It should be noted that unless otherwise specifically stated, the relative arrangement of the components and steps, the composition of the materials, the numerical expressions and the numerical values set forth in these embodiments should be interpreted as being merely exemplary and not as limiting.
[0043] The words "first", "second" and similar words used in this disclosure do not indicate any order, quantity or importance, but are only used to distinguish different parts. The words "include" or "comprises" and similar words mean that the elements before the word include the elements listed after the word, and do not exclude the possibility of including other elements. "Up", "down" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.
[0044] In the present disclosure, when a specific component is described as being located between a first component and a second component, there may or may not be an intermediate component between the specific component and the first component or the second component. When a specific component is described as being connected to other components, the specific component may be directly connected to the other components without an intermediate component, or may not be directly connected to the other components but have an intermediate component.
[0045] All terms (including technical terms or scientific terms) used in the present disclosure have the same meanings as those understood by ordinary technicians in the field to which the present disclosure belongs, unless otherwise specifically defined. It should also be understood that terms defined in general dictionaries, such as general dictionaries, should be interpreted as having meanings consistent with their meanings in the context of the relevant technology, and should not be interpreted in an idealized or extremely formal sense, unless explicitly defined herein.
[0046] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered part of the specification.
[0047] With the continuous development and progress of information technology, the amount and complexity of time series data are increasing day by day, and are widely used in the fields of manufacturing, finance, network security, health care, climate, etc. The main research of time series includes data preprocessing, data segmentation, similarity measurement, feature extraction, dimension reduction, classification, clustering and rule recognition.
[0048] Trend characteristics are one of the most important information in time series, which reveals the changing trend and pattern characteristics of time series. Therefore, feature extraction of time series has always been the main focus of time series research. In order to capture the trend characteristics of time series, researchers have proposed a time series feature representation method, which can reduce the dimension of time series while retaining its main trend characteristics. Currently, common time series data feature representation methods include frequency domain representation, autoregressive model, symbolic aggregation approximation method, empirical mode decomposition method and piecewise linear representation. Traditional time series processing methods such as Fourier transform are usually based on the distance between points and cannot well describe the dynamic characteristics of time series data. In contrast, piecewise linear representation (PLR) has the advantages of time multi-analysis, intuitive form and high dimensionality reduction, which is more in line with the discontinuous and intermittent monotonicity of roll and package sorting business indicators in different intervals for time series trend analysis. Traditional PLR algorithms, such as the bottom-up piecewise linearization method and the time series PAA piecewise representation algorithm, all use a single fitting error as the basis for segmenting the original data, and their piecewise fitting and trend analysis results are not ideal. Moreover, in the production process of cigarette manufacturing companies, the reflux of sorting business requires efficient and rapid response to early warning or shorten the shutdown time. Therefore, it is necessary to reduce the computational complexity of the algorithm and improve the fitting accuracy, and a time series piecewise linear representation method based on trend turning points, feature points and key points has emerged.
[0049] The above studies show that the PLR method has a good effect in extracting time series trends, but it belongs to local analysis. The important points based on trend turning points, feature points and key points all depend on the relationship between adjacent points and cannot accurately reflect the overall trend of the time series. In addition, the algorithm is not robust and the results are easily affected by the input parameters.
[0050] like Figure 1 As shown, the method for extracting time series trend features of roll and package sorting service indicators provided in this embodiment specifically includes the following steps during actual execution:
[0051] Step S1, time series normalization: read the time series, remove outliers in a given time series, and represent the time series after removing outliers with symbols to obtain an initial time series.
[0052] In one embodiment of the present invention, the time series of the roll and package sorting service indicators is read to obtain the time and detection value of the detection indicator. There are 9557 data points in the time series. After removing the abnormal values, 7883 data points remain. The first 1000 data points are selected and used (n=1000) Normalized representation;
[0053] Furthermore, after step S1 and before step S2, the method for extracting time series trend features of roll and package sorting service indicators further includes:
[0054] Step S11: Eliminate missing values and outliers in the initial time series.
[0055] Step S2, obtaining an initial important point sequence: segmenting the original time series based on the definition of important points to obtain an initial important point sequence.
[0056] The initial important point sequence is composed of the endpoints of the time series and the internal discrete local extreme points. In one embodiment of the present invention, the endpoints of the time series and the internal discrete local extreme points are defined as the initial important points, and 55 initial important points are obtained. The initial important point sequence is as follows: Figure 2 shown.
[0057] Step S3, reconstructing the initial time series: based on the initial time series in step S1, performing wavelet decomposition and reconstruction to obtain a reconstructed time series.
[0058] In one embodiment of the method for extracting time series trend features of roll and package sorting service indicators of the present invention, step S3 may specifically include:
[0059] Step S31, using the db series wavelet function to decompose the time series k times to perform k-level decomposition, and obtain 1 approximate sequence and k detail sequences.
[0060] In one embodiment of the present invention, the original time series Wavelet decomposition and reconstruction are performed, and the original time series is decomposed into 8 layers using the db2 wavelet function to obtain 1 approximate series and 8 detail series.
[0061] Step S32: Use the db series wavelet function to perform Reconstructed, we get The reconstructed time series is: Indicates rounding down.
[0062] In one embodiment of the present invention, the db2 wavelet function is used to perform 4-layer reconstruction on the decomposed approximate sequence and detail sequence to obtain the reconstructed initial time series .
[0063] Step S33: draw a reconstructed timing diagram.
[0064] The reconstructed initial time series is as follows Figure 3 shown.
[0065] Step S4, reconstructing the important point sequence: segmenting the reconstructed time series in step S3 based on the piecewise linear representation of the important points to obtain a reconstructed important point sequence.
[0066] Specifically, the PLR-IPS method is used to segment the time series, thereby segmenting the time series reconstructed in step S3 to obtain a reconstructed important point sequence, such as Figure 4 As shown. In order to minimize the fitting error after segmentation representation under the preset compression rate. The PLR-IPS method aims to segment the time series and minimize the fitting error under the specified compression rate. The key is to use a suitable evaluation strategy to select segmentation points from the important point sequence to achieve the purpose of reducing the fitting error.
[0067] Step S5, calculate the global factor: based on the reconstructed important point sequence in step S4, the initial important point sequence in step S2 and the initial time sequence in step S1, compare the important points in the reconstructed important point sequence that are the same as those in the initial important point sequence to obtain the global factor value of the important point.
[0068] Specifically, the initial value of the reconstruction parameter θ is set to ( Indicates rounding down), and decreases gradually to 0 with a tolerance of 1. As the reconstruction parameter θ decreases, the components involved in the reconstruction become fewer, resulting in a decrease in the number of important points for reconstruction. Therefore, the global factor will increase. The global factor value is an indicator for evaluating the overall dimension of the time series.
[0069] Step S6, calculating the local factor: calculating the distance from the important point to the endpoint in the important point sequence and the reconstructed important point sequence according to the orthogonal distance.
[0070] Specifically, the distance factor is used to measure the relative difference of important points in a local range. Points with larger distance factors have larger differences with adjacent points and are more likely to be important turning points; points with smaller distance factors may be interference extreme points. By selecting an appropriate threshold, interference points can be excluded.
[0071] Step S7, calculating the comprehensive factor: weighting the global factor value in step S5 and the local factor value in step S6 to calculate the comprehensive factor value.
[0072] In one embodiment of the method for extracting time series trend features of roll and package sorting service indicators of the present invention, step S7 may specifically include:
[0073] Step S71: Use genetic algorithm to determine model parameters as the weight of distance factor, and select the average fitting error The smallest set of parameter values is used as the final model parameters.
[0074] Step S72: weight the global factor in step S5 and the local factor in step S6 according to the determined model parameters to calculate a comprehensive factor value.
[0075] In one embodiment of the present invention, a parameter value of 0.55 is finally selected as the weight value of the final global factor, and a parameter value of 0.45 is selected as the weight value of the distance factor, and finally the comprehensive factor value of each initial important point is calculated.
[0076] Step S8, obtaining the final important point sequence: based on the comprehensive factor value in step S7, the important point sequence is sorted, and segmentation points are selected from the sorted important point sequence to obtain the final important point sequence.
[0077] Specifically, the initial important points are sorted from large to small according to the comprehensive factor values calculated in step S7 to obtain a sorted important point sequence.
[0078] Furthermore, after step S8 and before step S9, the method for extracting time series trend features of roll and package sorting service indicators further includes:
[0079] Step S81: filter the sorted important point sequence according to the requirements, set the number of segments N, select the first N+1 important points, and obtain the final segmentation point sequence and segmentation position.
[0080] In one embodiment of the present invention, the first 40 important points are selected from the important point sequence of the time series after comprehensive sorting to obtain the final segmentation point sequence and segmentation position. The schematic diagram of the important points after sorting and screening based on the comprehensive factors is as follows: Figure 5 shown.
[0081] Step S9, trend feature extraction: Use straight line segments to connect the final important point sequence in step S8 to obtain a linear representation of the time series, and complete the trend feature extraction.
[0082] The feature extraction of time series can reflect the changing trend and pattern characteristics of the time series. In one embodiment of the present invention, the final segmentation point sequence and the segmentation points at adjacent positions in the segmentation position are connected in sequence with straight line segments to obtain a piecewise linear trend representation of the time series. At this time, the trend change of the time series can be intuitively seen, such as Figure 6 shown.
[0083] The method for extracting trend features of time series of indicators of roll and package sorting business provided by the embodiment of the present invention extracts trend features of time series based on wavelet transform and important points, makes full use of the characteristics of wavelet transform that can provide time domain and frequency domain information at the same time, can better capture local features of signals, effectively weaken noise interference, proposes a comprehensive factor evaluation method based on distance factor and global factor, takes into account global and local information of time series, and realizes effective extraction of time series change trend; makes full use of the time series data collected in the existing production environment for analysis, reduces the implementation difficulty through the data feature extraction method, has low requirements on computing power, and therefore has small hardware investment in processing equipment, is simple to operate, and has high efficiency in fault feature screening; effectively reduces the processing volume of indicator data continuously generated in the production process by reducing the dimension and denoising the key indicator data, and improves the accuracy of abnormal alarm; abnormal small fluctuations can be discovered in time through the final trend chart, better matches the nonlinear characteristics of the indicators of the roll and package production process, and improves the controllability of the production process; can be directly applied to the existing production line, the indicator time series data are all monitoring data of the production-related environment, and do not involve the sensor installation and real-time transmission of production line data of the original earlier production line, thus avoiding production line transformation and reducing installation costs.
[0084] So far, various embodiments of the present disclosure have been described in detail. In order to avoid obscuring the concept of the present disclosure, some details known in the art are not described. Based on the above description, those skilled in the art can fully understand how to implement the technical solution disclosed here.
[0085] Although some specific embodiments of the present disclosure have been described in detail by way of examples, it should be understood by those skilled in the art that the above examples are for illustration only and are not intended to limit the scope of the present disclosure. It should be understood by those skilled in the art that the above embodiments may be modified or some technical features may be replaced by equivalents without departing from the scope and spirit of the present disclosure. The scope of the present disclosure is defined by the appended claims.
Claims
1. A method for extracting time series trend features of roll and package sorting business indicators, characterized in that: include: Step S1, time series normalization: read the time series, remove outliers in a given time series, and represent the time series after removing outliers with symbols to obtain an initial time series; Step S2, obtaining an initial important point sequence: segmenting the original time series based on the definition of important points to obtain an initial important point sequence; Step S3, reconstructing the initial time series: based on the initial time series in step S1, performing wavelet decomposition and reconstruction to obtain a reconstructed time series; Step S4, reconstructing the important point sequence: segmenting the reconstructed time series in step S3 based on the piecewise linear representation of the important points to obtain a reconstructed important point sequence; Step S5, calculating the global factor: based on the reconstructed important point sequence in step S4, the initial important point sequence in step S2 and the initial time sequence in step S1, comparing the important points in the reconstructed important point sequence that are the same as those in the initial important point sequence, to obtain the global factor value of the important point; Step S6, calculating the local factor: calculating the distance from the important point to the endpoint in the important point sequence and the reconstructed important point sequence according to the orthogonal distance; Step S7, calculating the comprehensive factor: weighting the global factor value in step S5 and the local factor value in step S6 to calculate the comprehensive factor value; Step S8, obtaining the final important point sequence: based on the comprehensive factor value in step S7, sorting the important point sequence, selecting segmentation points from the sorted important point sequence to obtain the final important point sequence; Step S9, trend feature extraction: Use straight line segments to connect the final important point sequence in step S8 to obtain a linear representation of the time series, and complete the trend feature extraction.
2. The method for extracting time series trend characteristics of roll and package sorting business indicators according to claim 1 is characterized in that: After step S1 and before step S2, the method for extracting time series trend features of roll and package sorting service indicators further includes: Step S11: Eliminate missing values and outliers in the initial time series.
3. The method for extracting time series trend characteristics of roll and package sorting business indicators according to claim 1 is characterized in that: The initial important point sequence in step S2 consists of the endpoints of the time series and internal discrete local extreme points.
4. The method for extracting time series trend characteristics of roll and package sorting business indicators according to claim 1 is characterized in that: The step S3, reconstructing the initial time series, specifically comprises the following steps: Step S31, using the db series wavelet function to decompose the time series k times to perform k-level decomposition, and obtain 1 approximate sequence and k detail sequences; Step S32: Use the db series wavelet function to perform Reconstructed, we get The reconstructed time series is: Indicates rounding down; Step S33: draw a reconstructed timing diagram.
5. The method for extracting time series trend characteristics of roll and package sorting business indicators according to claim 1 is characterized in that: The step S4, reconstructing the important point sequence, specifically includes: The PLR-IPS-based method represents the time series in segments so that the fitting error after segmentation is minimized under a preset compression rate.
6. The method for extracting time series trend characteristics of roll and package sorting business indicators according to claim 1 is characterized in that: The step S5, calculating the global factor, specifically includes: Set the initial value of the reconstruction parameter θ to , and gradually decreases to 0 with a tolerance of 1. As the reconstruction parameter θ decreases, the components involved in the reconstruction become fewer, resulting in a decrease in the number of important points for reconstruction. Therefore, the global factor will increase. The global factor value is an indicator used to evaluate the overall dimension of the time series.
7. The method for extracting time series trend characteristics of roll and package sorting business indicators according to claim 1 is characterized in that: The step S6, calculating the local factor, specifically includes: The distance factor is used to measure the relative difference of important points in a local range. Points with larger distance factors have larger differences with adjacent points and are more likely to be important turning points. Points with smaller distance factors may be interference extreme points. By selecting an appropriate threshold, interference points can be excluded.
8. The method for extracting time series trend characteristics of roll and package sorting business indicators according to claim 1 is characterized in that: The step S7, calculating the comprehensive factor, specifically includes: Step S71: Use genetic algorithm to determine model parameters as the weight of distance factor, and select the average fitting error The smallest set of parameter values is used as the final model parameters; Step S72: weight the global factor in step S5 and the local factor in step S6 according to the determined model parameters to calculate a comprehensive factor value.
9. The method for extracting time series trend characteristics of roll and package sorting business indicators according to claim 1 is characterized in that: After step S8 and before step S9, the method for extracting time series trend features of roll and package sorting service indicators further includes: Step S81: filter the sorted important point sequence according to the requirements, set the number of segments N, select the first N+1 important points, and obtain the final segmentation point sequence and segmentation position.