Data sequence comparison analysis method, device, electronic device and storage medium

Abstract

The present invention relates to a comparative analysis method, device, electronic device and storage medium for data sequences, wherein the comparative analysis method for data sequences comprises: obtaining a first indicator data sequence and a second indicator data sequence for the same performance indicator; generating a sequence to be classified based on the first indicator data sequence and the second indicator data sequence, wherein the sequence to be classified contains difference information between the first indicator data sequence and the second indicator data sequence; inputting the sequence to be classified into a preset comparative analysis model so that the preset comparative analysis model classifies the sequence to be classified to obtain comparative analysis results. The embodiment of the present application can realize intelligent, automated and professional comparative analysis of mobile communication performance indicators, and can replace a large number of business personnel to complete a large amount of daily data comparative analysis work, while ensuring the efficiency and effect of data analysis, saving a large amount of business human resource expenses.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a data sequence comparison and analysis method, device, electronic device and storage medium. Background Art

[0002] Data analysis can be seen everywhere in all walks of life and in daily life, and comparative analysis is one of the most common and important contents in data analysis. The basis of data analysis is statistics.

[0003] In statistics, there are several methods for comparing two numerical sequences to determine whether they are different or similar: difference judgment based on sequence distance measurement, similarity judgment based on sequence cosine similarity, similarity judgment based on Pearson correlation coefficient or Spearman rank correlation coefficient of equal-length sequences, trend similarity judgment based on first-order difference correlation coefficient of sequences, similarity judgment based on ARIMA forecasting model parameters of sequences, similarity judgment of unequal-length sequences based on dynamic time warping algorithm, statistical mean difference judgment of sequences based on t-test (small sample) or Z-test (large sample) under normal distribution assumption, statistical standard deviation difference judgment of sequences based on F-test under normal distribution assumption, statistical difference judgment of sequences based on Wilcoxon rank sum test under non-normal distribution assumption, and so on.

[0004] The performance indicators in the field of mobile communications are numerical series, and the corresponding comparative analysis methods are also based on the basic principles of statistics. However, due to the professionalism and particularity of its business field, when analyzing specific performance indicators, the mathematical model and analysis method to be used should be determined according to the business background, data characteristics, etc. of the performance indicators.

[0005] Data analysis is based on statistics. There are many existing statistical analysis software, including commercial software with complete functions and powerful performance such as SAS, SPSS, Minitab, Statistica, S-plus, GAUSS, Eviews, etc., as well as free and shared open source software such as R language for statistical analysis and Python language with Statsmodels / Scipy / Rpy2 statistical analysis library. The daily office software Excel also provides some basic data statistics functions. Almost all statistical analysis software is committed to supporting rich and diverse visualization of data, so that analysts can make intuitive comparative judgments through visual observation and business experience.

[0006] In the field of communications, analysis of performance indicators is an important daily task for network operation and maintenance and performance optimization, and comparative analysis is the most common content. For example, you can compare the performance indicator differences between network 1 and network 2, compare the performance indicator changes between today and yesterday, compare the performance indicator changes between today and the same day last week, compare the performance indicator changes between this week and last week, compare the performance indicator changes before and after turning on a certain function, compare the performance indicator changes before and after modifying a certain parameter, compare the performance indicator changes before and after upgrading the version, and so on.

[0007] Due to the complexity of communication networks, especially mobile communication networks, which include core networks, transmission networks, and access networks, they involve a variety of hardware equipment types and topological networking, software support functions and parameter configurations, application scenarios, user equipment, service types, perception evaluation, etc. There are many types of related performance indicators, and the number is huge, such as status, access, disconnection, switching, load, rate, delay, packet error, packet loss, retransmission, energy consumption, efficiency, resource overhead, channel quality, scheduling performance, and proprietary indicators of various characteristics (such as characteristic indicators of voice services, characteristic indicators of high-speed mobility, characteristic indicators of space division multiplexing, etc.). The performance indicators have time by default (year / month / day / hour / minute / second / millisecond). The performance indicators of mobile communications are analyzed in terms of time and space (multiple time granularities, such as the entire network / subnet / base station / cell / antenna, etc.), plus the various random factors that affect the performance indicators of the wireless environment in the mobile communication network, such as the distribution and mobility of terminals, network load and changes, signal fading, interference noise, etc., as well as analysis requirements at multiple granularities, such as summary analysis based on day granularity and real-time analysis based on millisecond granularity, macro analysis based on the granularity of the entire network and micro analysis based on the granularity of the cell, etc. Therefore, the performance indicator analysis of mobile communications is not only a large workload, but also a professional job with high technical level and ability requirements. It requires data analysts to have a deep understanding of mobile communication services and to be proficient in statistical analysis methods.

[0008] However, the current comparative analysis of mobile communication performance indicators is a routine action in daily network operation and maintenance and performance optimization. The workload is huge, and it is impossible to adopt it in large numbers or have enough experts who are proficient in both business and statistics to deal with it. Instead, it is done by a large number of ordinary business personnel, based on simple and easy-to-master statistical tools, such as using Excel to calculate the mean, standard deviation, difference in means and change range, difference in standard deviation and change range of two performance indicator sequences to be compared, and other basic statistics, combined with visual observation after visualizing the performance indicators in charts, and relying on the business experience and intuition established by business personnel through their daily business work, to make intuitive comparative judgments. Generally, the more business analysis experience one has, the more accurate the intuitive judgment.

[0009] Therefore, the current dilemma faced by the comparative analysis of mobile communication performance indicators is that the workload of comparative analysis is large, the technical difficulty of comparative analysis is high, and a large number of experts who understand both business and statistics are needed, which is basically impossible to meet in reality. In fact, a large number of ordinary business personnel are used to make intuitive comparative judgments based on simple statistical models, through data visualization display view observations, combined with business experience, to make up for the shortcomings of statistical analysis. However, the accuracy of the judgment is related to the business experience and technical capabilities of the analysts and cannot be effectively guaranteed. Summary of the invention

[0010] In order to solve the above technical problems or at least partially solve the above technical problems, the present application provides a data sequence comparison and analysis method, device, electronic device and storage medium.

[0011] In a first aspect, the present application provides a comparative analysis method for a data sequence, comprising:

[0012] Acquire a first indicator data series and a second indicator data series about the same performance indicator;

[0013] generating a sequence to be classified based on the first indicator data sequence and the second indicator data sequence, wherein the sequence to be classified includes difference information between the first indicator data sequence and the second indicator data sequence;

[0014] The sequence to be classified is input into a preset comparative analysis model, so that the preset comparative analysis model classifies the sequence to be classified to obtain a comparative analysis result.

[0015] Optionally, generating a sequence to be classified based on the first indicator data sequence and the second indicator data sequence includes:

[0016] Preprocessing the first indicator data sequence and the second indicator data sequence respectively to obtain a first equal-length sequence and a second equal-length sequence;

[0017] The first equal-length sequence and the second equal-length sequence are subjected to subtraction and difference processing to obtain the sequence to be classified.

[0018] Optionally, preprocessing the first indicator data sequence and the second indicator data sequence respectively to obtain a first equal-length sequence and a second equal-length sequence includes:

[0019] Correcting the abnormal values ​​in the first indicator data sequence and the second indicator data sequence respectively to obtain a first corrected sequence and a second corrected sequence;

[0020] Time-aligning the first corrected sequence and the second corrected sequence respectively to obtain a first aligned sequence and a second aligned sequence;

[0021] The first aligned sequence and the second aligned sequence are length-aligned to obtain the first equal-length sequence and the second equal-length sequence, respectively.

[0022] Optionally, performing subtraction difference processing on the first equal-length sequence and the second equal-length sequence to obtain the sequence to be classified includes:

[0023] The first indicator data sequence and the second indicator data sequence are input into a preset subtraction difference formula, and the sequence to be classified is calculated by the preset subtraction difference formula.

[0024] Optionally, the training method of the preset comparative analysis model includes:

[0025] Obtaining first training data corresponding to a plurality of performance indicators and first data labels corresponding thereto;

[0026] Performing supervised training on the initial model using the first training data and the first data label to obtain a pre-trained model;

[0027] Acquire second training data corresponding to a plurality of performance indicators and second data labels corresponding thereto, wherein the second data labels are determined according to expert judgment results;

[0028] The pre-trained model is trained using the second training data and the second data label to obtain the preset comparative analysis model.

[0029] Optionally, obtaining second training data corresponding to a plurality of performance indicators and second data labels corresponding thereto includes:

[0030] Obtaining a third indicator data sequence and a fourth indicator data sequence corresponding to a plurality of the performance indicators;

[0031] For each of the performance indicators, obtaining analysis results based on the third indicator data sequence and the fourth indicator data sequence, wherein the analysis results include expert judgment results and analysis explanation data;

[0032] Combine the third indicator data sequence, the fourth indicator data sequence and the analysis description data corresponding to each performance indicator to obtain second training data corresponding to the performance indicator;

[0033] Determine the expert judgment result corresponding to each performance indicator as a second data label corresponding to the performance indicator;

[0034] The second training data and the second data label corresponding to each performance indicator are obtained.

[0035] Optionally, obtaining first training data corresponding to a plurality of performance indicators and first data labels corresponding thereto includes:

[0036] Acquire a fifth indicator data sequence and a sixth indicator data sequence corresponding to a plurality of the performance indicators;

[0037] Batch labeling the fifth indicator data series and the sixth indicator data series corresponding to the plurality of performance indicators to obtain a first data label corresponding to each performance indicator;

[0038] Combine the fifth indicator data sequence and the sixth indicator data sequence corresponding to each of the performance indicators to obtain first training data corresponding to the performance indicator;

[0039] The first training data and the first data label corresponding to each performance indicator are obtained.

[0040] In a second aspect, the present application provides a data sequence comparison and analysis device, comprising:

[0041] An acquisition module, used to acquire a first indicator data sequence and a second indicator data sequence for the same performance indicator;

[0042] A determination module, configured to generate a sequence to be classified based on the first indicator data sequence and the second indicator data sequence, wherein the sequence to be classified includes difference information between the first indicator data sequence and the second indicator data sequence;

[0043] The classification module is used to input the sequence to be classified into a preset comparative analysis model so that the preset comparative analysis model classifies the sequence to be classified to obtain a comparative analysis result.

[0044] In a third aspect, the present application provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus;

[0045] Memory, used to store computer programs;

[0046] The processor is used to implement any data sequence comparison and analysis method described in the first aspect when executing the program stored in the memory.

[0047] In a fourth aspect, the present application provides a computer-readable storage medium, on which is stored a program for a data sequence comparative analysis method, which, when executed by a processor, implements the steps of any of the data sequence comparative analysis methods described in the first aspect.

[0048] The above technical solution provided by the embodiment of the present application has the following advantages compared with the prior art:

[0049] The method provided in the embodiment of the present application generates a sequence to be classified containing difference information between the first indicator data sequence and the second indicator data sequence corresponding to the performance indicator, and inputs the sequence to be classified into a preset comparative analysis model so that the preset comparative analysis model classifies the differences between the two indicator data sequences to obtain comparative analysis results, thereby realizing intelligent, automated and professional comparative analysis of mobile communication performance indicators, which can replace a large number of business personnel to complete a large amount of daily data comparative analysis work, while ensuring data analysis efficiency and effectiveness, and saving a large amount of business human resource expenses. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0051] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0052] Figure 1 A flowchart of a comparative analysis method of a data sequence provided in an embodiment of the present application;

[0053] Figure 2 for Figure 1 Flow chart of step S102;

[0054] Figure 3 A flowchart of a method for training a preset comparative analysis model provided in an embodiment of the present application;

[0055] Figure 4 A data visualization diagram of sequence 1 corresponding to the number of network users, a performance indicator provided in an embodiment of the present application;

[0056] Figure 5 A data visualization diagram of sequence 2 corresponding to the number of network users, a performance indicator provided in an embodiment of the present application;

[0057] Figure 6 A data visualization diagram of sequence 3 corresponding to the number of network users, a performance indicator provided in an embodiment of the present application;

[0058] Figure 7 A data visualization diagram of sequence 1 corresponding to a performance indicator - network traffic provided in an embodiment of the present application;

[0059] Figure 8A data visualization diagram of sequence 2 corresponding to a performance indicator - network traffic provided in an embodiment of the present application;

[0060] Fig. 9 A data visualization diagram of sequence 3 corresponding to a performance indicator - network traffic provided in an embodiment of the present application;

[0061] Fig.10 A data visualization diagram of sequence 1 corresponding to a performance indicator - network connection rate provided in an embodiment of the present application;

[0062] Fig.11 A data visualization diagram of sequence 2 corresponding to a performance indicator - network connection rate provided in an embodiment of the present application;

[0063] Fig.12 A data visualization diagram of sequence 3 corresponding to a performance indicator - network connection rate provided in an embodiment of the present application;

[0064] Fig.13 A schematic diagram of an exemplary expert judgment result provided in an embodiment of the present application;

[0065] Fig.14 A schematic diagram of an analysis description in an exemplary analysis result provided in an embodiment of the present application;

[0066] Fig.15 A structural diagram of a data sequence comparison and analysis device provided in an embodiment of the present application;

[0067] Fig.16 A structural diagram of a data sequence comparison and analysis system provided in an embodiment of the present application;

[0068] Fig.17 A structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0069] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0070] The current dilemma faced by the comparative analysis of mobile communication performance indicators is that the workload of comparative analysis is large, the technical difficulty of comparative analysis is high, and a large number of experts who understand both business and statistics are required, which is basically impossible to meet in reality. In fact, a large number of ordinary business personnel are used to make intuitive comparative judgments based on simple statistical models, through view observations displayed by data visualization, combined with business experience, to make up for the shortcomings of statistical analysis. However, the accuracy of the judgment is related to the business experience and technical capabilities of the analysts and cannot be effectively guaranteed. In view of the above-mentioned dilemma of comparative analysis of mobile communication performance indicators, inspired by the way in which ordinary business personnel widely make intuitive judgments based on view comparison combined with business experience, and drawing on the deep learning model based on deep neural network that has been very successful and widely used in current artificial intelligence, the embodiment of the present application provides a comparative analysis method, device, electronic device and storage medium for data sequences, which integrates two data comparative analysis methods based on statistical analysis models and based on deep learning models, absorbs and utilizes the respective advantages of statistical analysis and business analysis, and realizes the intelligent, automated and professional comparative analysis of mobile communication performance indicators. While liberating a large number of analysis manpower requirements, it also ensures the accuracy of comparative analysis.

[0071] The data sequence comparison and analysis method, device, electronic device and storage medium provided by the present invention can be applied to the data analysis field of the mobile communication industry, especially the comparison and analysis of mobile communication performance indicators (numerical sequence data), to support business personnel in related industries such as network operation and maintenance, performance optimization, etc. to quickly and efficiently complete a large number of daily performance indicator comparison and analysis work. On the other hand, it can also be applied to the comparison of one-dimensional business data sequences in other industries, especially the comparison of industry data that is highly related to business knowledge and business experience.

[0072] In one embodiment of the present application, a comparative analysis method of a data sequence is provided, such as Figure 1 As shown, the method may include the following steps:

[0073] Step S101, obtaining a first indicator data sequence and a second indicator data sequence for the same performance indicator;

[0074] In the embodiments of the present application, the performance indicator may refer to a performance indicator in the field of mobile communications. For example, the performance indicator may be the number of network users, network traffic volume, network connection rate, network disconnection rate, etc.

[0075] The first indicator data sequence and the second indicator data sequence may be collected under different conditions or before and after the same event node. For example, the first indicator data sequence and the second indicator data sequence may be indicator data sequences of the same performance indicator of different networks, or indicator data sequences of the same performance indicator of different dates, or indicator data sequences of the same performance indicator before and after a certain function is enabled, or indicator data sequences of the same performance indicator before and after a certain parameter is modified, or indicator data sequences of the same performance indicator before and after a version is upgraded, and so on.

[0076] Step S102: Generate a sequence to be classified based on the first indicator data sequence and the second indicator data sequence.

[0077] In the embodiment of the present application, the sequence to be classified includes difference information between the first indicator data sequence and the second indicator data sequence;

[0078] In order to be able to use the preset comparative analysis model to compare the views of two one-dimensional data sequences, a sequence to be classified can be generated based on the first indicator data sequence and the second indicator data sequence, and only the difference information between the first indicator data sequence and the second indicator data sequence is retained. The difference information is the information in the two indicator data sequences that affects the comparative analysis results.

[0079] The present application combines two sequences into one sequence, thus converting the comparison problem of the two sequences into a classification problem of one sequence, thereby achieving the transformation of the problem. In this way, the combined sequence can be classified by adopting a one-dimensional convolutional neural network model, thereby achieving comparative judgment of the two sequences.

[0080] Step S103: input the sequence to be classified into a preset comparative analysis model, so that the preset comparative analysis model classifies the sequence to be classified to obtain a comparative analysis result.

[0081] In the embodiment of the present application, the preset comparative analysis model can adopt a deep learning model, and a deep convolutional neural network is preferentially recommended, which utilizes the advantages of convolutional neural networks in image feature recognition, extraction, and representation to better simulate the way in which business personnel make comparative judgments based on data visualization diagrams, which is more in line with the way in which business personnel make intuitive judgments based on data views combined with business experience. Since the sequence being processed is a one-dimensional sequence, a convolutional neural network based on one-dimensional convolution is used.

[0082] In this step, the one-dimensional sequence to be classified is input into a preset comparative analysis model. The preset comparative analysis model can classify the differences between the two indicator data sequences to obtain comparative analysis results.

[0083] The method provided in the embodiment of the present application generates a sequence to be classified containing difference information between the first indicator data sequence and the second indicator data sequence corresponding to the performance indicator, and inputs the sequence to be classified into a preset comparative analysis model so that the preset comparative analysis model classifies the differences between the two indicator data sequences to obtain comparative analysis results, thereby realizing intelligent, automated and professional comparative analysis of mobile communication performance indicators, which can replace a large number of business personnel to complete a large amount of daily data comparative analysis work, while ensuring data analysis efficiency and effectiveness, and saving a large amount of business human resource expenses.

[0084] In another embodiment of the present application, Figure 2 As shown, step S102 generates a sequence to be classified based on the first indicator data sequence and the second indicator data sequence, including:

[0085] Step S201, preprocessing the first indicator data sequence and the second indicator data sequence respectively to obtain a first equal-length sequence and a second equal-length sequence;

[0086] In this step, in order to facilitate the subsequent subtraction difference processing, the first indicator data sequence and the second indicator data sequence need to be preprocessed first. The preprocessing process includes: outlier processing, time alignment, length alignment and other preprocessing steps, into equal-length sequences of length N.

[0087] In one implementation of the present application, step S201 preprocesses the first indicator data sequence and the second indicator data sequence respectively to obtain a first equal-length sequence and a second equal-length sequence, including:

[0088] 1. Correcting the abnormal values ​​in the first indicator data sequence and the second indicator data sequence respectively to obtain a first corrected sequence and a second corrected sequence;

[0089] 2. Time-aligning the first corrected sequence and the second corrected sequence respectively to obtain a first aligned sequence and a second aligned sequence;

[0090] 3. Perform length alignment on the first alignment sequence and the second alignment sequence to obtain the first equal-length sequence and the second equal-length sequence respectively.

[0091] Exemplarily, the two indicator data sequences Seq1 and Seq2 to be compared become two equal-length sequences of length N after the above-mentioned first three preprocessing steps of outlier processing, time alignment, and length alignment.

[0092] Step S202: performing subtraction and difference processing on the first equal-length sequence and the second equal-length sequence to obtain the sequence to be classified.

[0093] After obtaining the first equal-length sequence and the second equal-length sequence, the two need to be merged. Based on the target task of "comparing the differences between two sequences of the same performance index", the sequences are processed by subtractive difference. By subtracting the difference, the information irrelevant to the comparison is eliminated, or in other words, the information that does not affect the comparison is removed, thereby achieving the simplification of the sequence.

[0094] In one implementation of the present application, performing subtraction difference processing on the first equal-length sequence and the second equal-length sequence to obtain the sequence to be classified includes:

[0095] The first indicator data sequence and the second indicator data sequence are input into a preset subtraction difference formula, and the sequence to be classified is calculated by the preset subtraction difference formula.

[0096] In the embodiment of the present application, the preset subtraction difference formula is:

[0097]

[0098] Exemplarily, two equal-length sequences Seq1(n) and Seq2(n) of length N are input into a preset subtraction difference formula to obtain a new sequence Seq3.

[0099] For example, since the indicator data series of performance indicators is often a time series, the influence of time factors on the series can be eliminated by using the subtraction difference of the preset subtraction difference formula.

[0100] In addition, by combining the two sequences into one, the comparison problem of the two sequences can be turned into a classification problem of one sequence, thus realizing the transformation of the problem. In this way, the merged sequence can be classified by using a deep neural network model to realize the comparison and judgment of the two sequences.

[0101] The subtraction difference between the two sequences designed based on the target task (comparing the difference between two sequences of the same performance indicator) in the above preprocessing process eliminates information irrelevant to the comparison, or in other words, removes information that does not affect the comparison, and achieves the simplification of the sequence, so that it can be trained with a smaller model, lower cost, and faster speed. Since mobile communication performance indicators are basically time series, the above preprocessing also eliminates the impact of time factors on the business represented by the sequence, so that the convolutional neural network CNN can also process time series such as mobile communication performance indicators.

[0102] In another embodiment of the present application, Figure 3 As shown, the training method of the preset comparative analysis model includes:

[0103] Step S301, obtaining first training data corresponding to a plurality of performance indicators and first data labels corresponding thereto;

[0104] In order to speed up the training of deep neural networks, a large amount of relatively stable basic training data can be collected and organized in advance in the data subsystem, such as performance indicators that are obviously unchanged during the stable operation of the network, or a stable performance indicator that is randomly split into two indifferent performance indicators, performance indicators that are obviously different during network operation, performance indicators that have been compared and analyzed by business personnel and have been tested and verified by time and the results of comparative analysis, or performance indicators that are obviously unchanged or obviously different through artificial construction, adding noise, simulating data generation, etc. These training data can be easily labeled in batches as the comparison and judgment results of labels, without the need for manual labeling one by one.

[0105] In one embodiment of the present application, step S301 obtains first training data corresponding to a plurality of performance indicators and first data labels corresponding thereto, including:

[0106] 1. Obtaining a fifth indicator data sequence and a sixth indicator data sequence corresponding to a plurality of the performance indicators;

[0107] 2. batch labeling the fifth indicator data series and the sixth indicator data series corresponding to the plurality of performance indicators to obtain a first data label corresponding to each performance indicator;

[0108] 3. Combining the fifth indicator data sequence and the sixth indicator data sequence corresponding to each of the performance indicators to obtain first training data corresponding to the performance indicator;

[0109] 4. Obtain the first training data and the first data label corresponding to each performance indicator.

[0110] In the embodiment of the present application, the first training data corresponding to each performance indicator is a one-dimensional data sequence obtained by performing the preprocessing and subtraction difference processing on the original training data as described in the above embodiment.

[0111] For ease of understanding, the present application embodiment provides several examples of the generation process of the first training data and the first data label:

[0112] Example 1: Compare two series of the performance indicator "number of network users".

[0113] Sequence 1 and Sequence 2 are data from two consecutive weeks. The length of the sequences is 7*24=168, which means 7 days in a week, 24 hours in a day, and 1 statistical sample data in each hour. The comparison result is "no difference".

[0114] Seq1={40.43,37.07,34.81,30.64,29.34,29.76,38.29,51.77,59.12,

[0115] 57.53,59.01,63.04,62.03,59.44,57.9,60.84,61.43,65.46,

[0116] 65.24,64.77,61.34,57.47,51.14,44.1,40.9,35.01,33.01,

[0117] 32.69,31.3,32.98,41.16,54.03,57.79,59.62,59.12,60.24,

[0118] 57.68,55.29,56.9,60.31,63.39,68.59,67.54,69.11,65.12,

[0119] 60.5,54.97,46.74,40.62,35.35,33.39,30.82,29.52,31.25,37.92,50.76,54.62,55.91,59.06,59.71,60.39,58.16,58.68,62.1,

[0120] 61,66.66,65.02,65.74,60.8,58.03,55.22,48.68,40.44,35.05,33.11,30.26,28.94,31.84,38.99,53.02,59.98,56.3,56.38,

[0121] 59.09,57.64,54.81,57.93,60.28,62.72,72.03,72.18,70.67,

[0122] 63.03,58.9,55.05,45.59,38.67,35.08,31.99,30.91,31.24,

[0123] 32.7,41.07,56.46,58.93,56.53,57.83,60.67,61.49,59.16,60.21,64.56,60.62,69.41,72.32,72.67,69.29,67.04,59.3,50.04,

[0124] 43.23,38.43,33.9,30.83,29.89,31.35,40.47,50.33,57.28,

[0125] 57.85,58.24,59.86,59.53,58.95,58.97,60.18,60.72,64.53,

[0126] 62.1,65.47,62.1,58.18,55.11,48.9,42.59,36.07,34.03,31.65,29.07,30.83,38.02,47.09,54.08,57.64,61.14,61.6,61.72,

[0127] 62.49,61.21,63.83,65.04,67.94,67.02,67.85,61.89,59.35,

[0128] 54.7,48.2}

[0129] Seq2={38.67,41.51,31.54,32.81,25.46,30.18,34.72,50.71,60.91,

[0130] 56.55,56.13,60.11,57.44,60.63,58.43,62.56,61.01,66.89,

[0131] 63.34,65.5,62.39,54.24,55.28,47.39,37.02,33.2,33.82,

[0132] 35.02,35.28,29.37,44.7,57.28,56.88,55.63,61.12,62.16,

[0133] 62.38,54.59,60.17,59.76,68.37,70.94,70.79,68.1,70.09,

[0134] 61.32,58.69,44.75,42.43,33.81,36.81,33.73,27.27,28,33.39,55.5,55.13,53.02,57.62,57.41,60.25,57.36,60.81,60.21,58.22,66.81,67.18,68.35,62.36,57.79,59.75,46.7,38.12,36.57,

[0135] 37.13,33.4,30.66,28.33,36.74,54.64,60.22,51.66,54.13,

[0136] 63.94,57.82,54.9,59.78,60.98,58.53,73.08,73.33,68.52,

[0137] 59.17,61.49,55.77,41.56,36.65,39.05,33.23,35.3,34,28.74,36.9,59.67,61.93,59.16,56.02,65.22,60.1,63.19,57.1,67.6,59.38,69.91,67.98,72.79,68.06,67.53,56.9,49.31,42.88,

[0138] 40.91,30.72,34.89,28,26.45,44.35,54.68,53.95,55.95,57.57,62.65,64.32,62.53,63.71,61.81,63.68,60.9,60.09,68.41,

[0139] 59.99,53.51,52.84,49.22,38.4,36.41,37.46,34.34,26.4,

[0140] 33.95,35.09,45.61,53.91,52.73,58.3,61.74,58.42,67.16,

[0141] 62.81,64.86,63.68,71.74,62.32,66.13,63.48,56.9,54.81,

[0142] 44.88}

[0143] Sequence 1 and sequence 2 have met the requirements of no outliers, time alignment, and equal length, so sequence 3 can be calculated according to the preset subtraction difference formula. The label of sequence 3 as training data is 0, which means no difference.

[0144] Seq3={0.04,-0.06,-0.13,-0.14,0.13,0.14,-0.03,0.04,-0.05,-0.08,0.01,0.05,-0.04,0.01,-0.02,0.06,-0.05,-0.05,0,-0.04,

[0145] 0.06,0.01,-0.02,0.05,-0.09,-0.06,-0.05,-0.14,-0.09,-0.15,0.07,-0.01,-0.05,-0.08,-0.01,-0.05,0.06,-0.01,0.04,-0.07,

[0146] -0.04,0.01,0,-0.06,-0.06,0.01,-0.04,-0.04,0.02,0.07,

[0147] 0.07,0.05,-0.02,-0.03,0.02,0,-0.06,-0.09,0.01,-0.08,

[0148] -0.08,-0.05,0.05,0.07,-0.02,0.02,0.03,0.03,-0.03,0,

[0149] -0.07,-0.02,0.1,0.03,-0.07,0.14,0.05,-0.16,0.09,-0.01,

[0150] 0.07,-0.08,-0.02,0.02,-0.05,0.02,0.08,0.05,-0.01,-0.02,0,-0.03,-0.03,0.01,-0.01,0.02,0.12,0.04,-0.05,-0.04,0.02,-0.05,0.06,-0.05,0.07,0.07,0.02,0.07,0.04,0,

[0151] -0.01,0.03,0.02,0.01,-0.02,0.04,0.02,0.06,0,-0.08,

[0152] 0.07,-0.09,0.07,0,0.11,-0.01,-0.12,-0.08,-0.04,-0.06,

[0153] -0.08,0.07,-0.07,-0.05,-0.03,0.02,-0.02,-0.01,0.01,0.03,0.07,-0.07,0.06,0.08,0.06,0.07,-0.08,0.02,-0.18,-0.03,

[0154] -0.01,0.1,0.05,-0.04,0.06,-0.01,-0.06,0.04,-0.04,0.07,

[0155] 0.04,0.07,0.03,0.07,-0.05,-0.03,0.04,0.01}

[0156] The data visualization diagrams of sequence 1, sequence 2 and sequence 3 corresponding to the performance indicator - number of network users are as follows: Figure 4 , Figure 5 and Figure 6 shown.

[0157] Example 2: Comparing two series of the performance indicator "network traffic".

[0158] Sequence 1 and Sequence 2 are data from two consecutive weeks. The length of each sequence is 7*24=168, and the comparison result is "there is a difference".

[0159] Seq1={0.22,0.24,0.2,0.17,0.28,0.29,0.8,1.09,2,2.8,3.34,3.42,2.15,2.69,2.63,2.59,2.55,2.93,2.38,2.23,1.54,1.38,0.81, 0.44,0.35,0.16,0.27,0.22,0.13,0.27,0.73,1.42,2.39,3.09,2.7 6,3.27,2.44,1.95,2.33,2.85,3.25,2.82,2.41,2.69,2.54,1.51,1 .52,1.15,0.82,0.22,0.17,0.11,0.16,0.33,0.63,1.25,2.16,2.96,3.16,3.05,2.64,1.96,2.63,2.72,2.93,3.3,3.6,3.15,2.58,1.71,1.35,0.69,0.55,0.23,0.45,0.15,0.15,0.53,0.88,1.37,2.65,2.72,2.72,2.86,1.99,1.69,2.52,2.49,2.64,3.6,2.99,2.52,1.88,

[0160] 1.87,1.56,0.6,0.35,0.21,0.19,0.25,0.4,0.42,0.68,1.4,

[0161] 1.98,2.92,2.79,2.57,2.11,1.8,2.26,3.16,2.86,3.22,

[0162] 2.82,2.26,2.13,1.56,1.04,0.59,0.37,0.39,0.31,0.2,

[0163] 0.21,0.35,0.64,1.25,2.1,2.8,2.95,3.04,2.29,2.25,2.42,2.48,2.28,2.3,2.27,2.67,2.53,1.43,1.5,0.91,0.35,0.24,0.17,0.21,0.19,0.39,0.82,1.16,2.2,2.56,2.56,3.01,

[0164] 2.07,2.18,2.05,2.24,2.7,2.26,2.01,2.36,2.11,1.52,

[0165] 1.21,0.53}

[0166] Seq2={0.22,0.26,0.2,0.18,0.29,0.3,0.83,1.13,2.08,2.89,3.44,3.63,2.29,2.87,2.69,2.8,2.58,3.13,2.4,2.4,1.62,1.46,

[0167] 0.84,0.45,0.36,0.18,0.28,0.22,0.13,0.28,0.79,1.52,2.62,3.12,2.87,3.49,2.45,2.02,2.43,3.03,3.49,3.1,

[0168] 2.42,2.78,2.66,1.64,1.66,1.18,0.85,0.24,0.18,0.12,0.17,0.37,0.65,1.3,2.25,3.2,3.32,3.16,2.69,2.07,2.78,2.74,3.17,3.33,3.78,3.46,2.67,1.88,1.43,0.72,0.58,0.24,0.49,0.16,0.15,0.56,0.94,1.42,2.8,2.75,284,

[0169] 2.95,2.15,1.77,2.7,2.63,2.89,3.7,3.04,2.53,1.9,1.97,

[0170] 1.57,0.6,0.37,0.22,0.21,0.27,0.41,0.42,0.69,1.5,2,

[0171] 2.94,2.98,2.6,2.2,1.8,2.38,3.43,3.13,3.45,2.95,2.44,

[0172] 2.21,1.63,1.08,0.59,0.37,0.41,0.32,0.21,0.21,0.37,0.7,1.29,2.24,3.03,3.15,3.11,2.33,2.27,2.46,2.68,2.34,2.44,2.46,2.89,2.62,1 .44,1.59,0.95,0.36,0.24,0.18,0.21,0.19,0.42,0.9,1.26,2.42,2.8,2.77,3.15,2.14,2.39,2.1,2.26,2.88,2.46,2.16,2.47,2.18,1.59,1.32,

[0173] 0.55}

[0174] Sequence 1 and sequence 2 have met the requirements of no outliers, time alignment, and equal length, so sequence 3 can be calculated according to the preset subtraction difference formula. The label of sequence 3 as training data is 1, indicating that there is a difference.

[0175] Seq3={0,0.06,0.02,0.05,0.01,0.04,0.04,0.03,0.04,0.03,0.03,0.06,0.07,0.06,0.02,0.08,0.01,0.06,0.01,0. 08,0.06,0.06,0.04,0.02,0.01,0.09,0.04,0.01,0.03,0.04,0.08,0.07,0.09,0.01,0.04,0.06,0,0.04,0.04,0.06,

[0176] 0.07,0.09,0,0.03,0.05,0.09,0.09,0.02,0.04,0.08,

[0177] 0.01,0.02,0.06,0.09,0.02,0.04,0.04,0.08,0.05,0.04,0.02,0.06,0.06,0.01,0.08,0.01,0.05,0.09,0.04,0.09,0.06,0.05,0.05,0.03,0.0 9,0.05,0.02,0.05,0.07,0.03,0.06,0.01,0.04,0.03,0.08,0.05,0.07,0.05,0.09,0.03,0.02,0,0.01,0.05,0.01,0.01,0.05,0.03,0.06,0.08,

[0178] 0.02,0,0.01,0.06,0.01,0.01,0.07,0.01,0.04,0,0.05,0.08,0.09,0.07,0.04,0.08,0.04,0.05,0.04,0,0.02,

[0179] 0.05,0.06,0.09,0.01,0.05,0.09,0.03,0.07,0.08,0.07,0.02,0.02,0.01,0.02,0.08,0.03,0.06,0.08,0.08,0.03,0,0.06,0.04,0.03,0.02,0.07,0,0.03,0.09,0.1,0.08,0.09,0.09,0.08,0.05,0.04,0.09,0.02,0.01,0.07,0.08,0.07,0.05,0.04,0.05,0.09,0.04}

[0180] Data visualization of sequence 1, sequence 2, and sequence 3 corresponding to the performance indicator - number of network users Figure 7 , Figure 8 and Fig. 9 shown.

[0181] Example 3: Comparing two series of the performance indicator "network connection rate".

[0182] Sequence 1 and Sequence 2 are data from two consecutive weeks. The length of Sequence 1 and Sequence 2 is 6*24=144, but Sequence 1 is from Tuesday to Sunday, while Sequence 2 is from Monday to Saturday. The comparison result is "no difference".

[0183] Seq1={98.97%,98.98%,98.95%,98.98%,98.98%,98.96%,98.94%,98.93%,98.92%,98.93%,98.86%,98.91%,98.89%,98.92%,98.92%,98.93%,98.88%,98.90%,98.92%,98.93%,98.91%,98.93%,98.92%,98.94%,98.95%,98.94%,98.98%,98.98%,98.98%,98.95%,98.94%,98.92%,98.91%,98.88%,98.90%,98.92%,98.94%,98.88%,98.92%,98.92%,98.87%,98.89%,98.90%,98.91%,98.90%,98.94%,98.91%,98.94%,98.91%,98.97%,98.97%,98.96%,98.99%,98.94%,98.96%,98.94%,98.92%,98.88%,98.92%,98.89%,98.87%,98.89%,98.91%,98.92%,98.91%,98.89%,98.91%,98.91%,98.89%,98.93%,98.91%,98.97%,98.96%,98.95%,98.96%,98.97%,98.94%,98.93%,98.95%,98.93%,98.92%,98.88%,98.92%,98.90%,98.91%,98.90%,98.94%,98.93%,98.91%,98.80%,98.91%,98.88%,98.92%,98.87%,98.93%,98.89%,98.96%,98.96%,98.97%,98.94%,98.91%,98.95%,98.94%,98.89%,98.90%,98.89%,98.92%,98.91%,

[0184] 98.89%,98.92%,98.96%,98.91%,98.93%,98.93%,98.96%,98.92%,98.94%,

[0185] 98.94%,98.63%,98.90%,98.97%,98.89%,98.87%,98.98%,98.97%,98.96%,

[0186] 98.94%,98.91%,98.89%,98.90%,98.95%,98.86%,98.85%,98.89%,98.88%,

[0187] 98.90%,98.89%,98.83%,98.88%,98.91%,98.94%,98.92%,98.85%,98.93%

[0188] ,

[0189] Seq2={98.95%,98.96%,98.97%,99.02%,98.96%,98.95%,98.94%,98.93%,98.90%,

[0190] 98.92%,98.86%,98.87%,98.86%,98.93%,98.88%,98.96%,98.86%,98.92%,

[0191] 98.93%,98.95%,98.88%,98.97%,98.88%,98.93%,98.98%,98.95%,98.99%,

[0192] 99.02%,99.00%,98.90%,98.94%,98.96%,98.88%,98.86%,98.94%,98.88%,

[0193] 98.93%,98.87%,98.95%,98.90%,98.89%,98.93%,98.94%,98.93%,98.93%,

[0194] 98.97%,98.93%,98.94%,98.92%,98.97%,99.00%,99.00%,99.04%,98.98%,

[0195] 99.00%,98.95%,98.89%,98.91%,98.93%,98.91%,98.84%,98.87%,98.94%,

[0196] 98.93%,98.95%,98.88%,98.89%,98.93%,98.87%,98.93%,98.92%,98.92%,

[0197] 98.91%,98.93%,98.94%,98.97%,98.89%,98.89%,98.96%,98.91%,98.95%,

[0198] 98.87%,98.95%,98.85%,98.89%,98.91%,98.89%,98.94%,98.94%,98.77%,

[0199] 98.93%,98.84%,98.89%,98.86%,98.94%,98.91%,98.99%,99.00%,98.94%,

[0200] 98.96%,98.93%,98.99%,98.93%,98.86%,98.86%,98.84%,98.92%,98.91%,

[0201] 98.92%,98.89%,98.99%,98.94%,98.91%,98.91%,99.00%,98.89%,98.89%,

[0202] 98.96%,98.87%,98.86%,98.98%,98.87%,98.89%,98.99%,98.93%,99.00%,

[0203] 98.93%,98.87%,98.89%,98.93%,98.92%,98.87%,98.85%,98.89%,98.87%,

[0204] 98.92%,98.91%,98.81%,98.92%,98.89%,98.94%,98.92%,98.85%,98.95%}

[0205] Sequence 1 and Sequence 2 meet the requirements of no outliers and equal length, but the time is not completely aligned. Since the comparison result is "no difference", according to the process processing requirements, the overlapping and aligned data of Sequence 1 and Sequence 2 from Tuesday to Saturday are intercepted, and Sequence 3 is calculated according to the preset subtraction difference formula. The label of Sequence 3 as training data is 0, which means no difference.

[0206] Seq3={0.03%,-0.01%,0.04%,-0.04%,0.00%,-0.05%,0.00%,0.02%,-0.04%,

[0207] 0.02%,0.04%,0.00%,0.02%,0.01%,0.03%,-0.02%,0.05%,0.00%,

[0208] -0.01％,-0.01％,-0.03％,0.05％,-0.01％,0.03％,-0.04％,-0.01％,-0.02％,

[0209] 0.01％,0.02％,-0.01％,-0.02％,-0.01％,0.03％,0.03％,0.04％,-0.04％,

[0210] 0.05％,0.03％,-0.03％,0.02％,0.00％,0.04％,-0.02％,-0.03％,-0.01％,

[0211] 0.02％,-0.02％,-0.02％,0.03％,0.02％,-0.03％,0.00％,-0.02％,-0.04％,

[0212] 0.05％,0.05％,0.03％,0.04％,0.00％,0.01％,-0.04％,0.04％,0.01％,

[0213] -0.01％,0.02％,0.04％,0.01％,-0.05％,0.03％,0.01％,-0.04％,0.03％,

[0214] -0.03％,0.03％,-0.01％,-0.04％,-0.02％,0.05％,0.01％,-0.04％,0.02％,

[0215] 0.03％,-0.03％,-0.02％,0.01％,0.00％,-0.03％,-0.04％,0.01％,-0.04％,

[0216] 0.05％,-0.01％,-0.01％,0.00％,0.01％,0.02％,0.04％,0.04％,-0.01％,

[0217] -0.03％,-0.01％,0.01％,0.05％,-0.02％,-0.04％,-0.03％,0.04％,-0.01％,

[0218] 0.00％,-0.03％,0.00％,-0.02％,0.01％,-0.03％,0.01％,-0.01％,0.04％,

[0219] 0.04％,-0.01％,-0.04％}

[0220] Data visualization of sequence 1, sequence 2, and sequence 3 corresponding to the performance indicator - number of network users Fig.10 , Fig.11 and Fig.12 shown.

[0221] Example 4: Compare two series of the performance indicator "network drop rate".

[0222] Sequence 1 and Sequence 2 are data from two consecutive weeks. The length of Sequence 1 is 7*24=168, but the length of Sequence 2 is only 1*12=12, that is, it only has half a day's data. Since the number of samples in Sequence 2 is obviously too small, the comparison result is "unable to judge". It is necessary to wait for Sequence 2 to obtain more data for a longer time before an effective comparison can be made.

[0223] Step S302, using the first training data and the first data label to perform supervised training on the initial model to obtain a pre-trained model;

[0224] The first training data and the first data labels collected and organized in advance are used to perform supervised pre-training on the deep neural network to obtain an integrated pre-training model.

[0225] Step S303, obtaining second training data corresponding to a plurality of performance indicators and second data labels corresponding thereto.

[0226] In the embodiment of the present application, the second training data is more complex and more difficult to classify than the first training data, so further judgment by experts is required. Using more complex second training data can increase the robustness of the model. The second data label is determined based on the expert judgment results.

[0227] The human expert group is responsible for receiving and completing the data analysis tasks arranged by the system and providing feedback on the analysis results. The expert analysis results include two parts: one is the expert judgment result, such as no difference or difference, or a more detailed classification judgment; the other is the analysis description, which is used to indicate the analysis method, prompt analysis associations, etc.

[0228] The results of the comparative analysis of two series of a certain performance indicator include, in addition to the expert judgment results, the analysis description. As an important part of the analysis result, the analysis description, together with the judgment result, constitutes the final analysis result. Table 1 is an example of an analysis result:

[0229] Table 1. Standardized format for comparative analysis results of a single performance indicator series

[0230]

[0231] The expert judgment results of expert comparative analysis can adopt simple binary judgment, that is, judging whether there is a difference, or more complex multi-classification judgment, such as distinguishing the degree of difference, no difference, slight difference, significant difference, or adopting more detailed multi-classification judgment, such as judging the significant difference between sequence 2 and sequence 1, and whether the trend is increasing or decreasing. For those that cannot be judged, an "unable to judge" judgment result can be added. The expert judgment results can refer to Fig.13 shown.

[0232] The analysis description in the analysis results can be referred to Fig.14 As shown, analysis instructions can be given, that is, what analysis method to use for analysis, and instructions can also be given, such as considering factors related to business rules, paying attention to related performance indicators, and processing tips for abnormal data with random fluctuations.

[0233] In one embodiment of the present application, step S303 obtains second training data corresponding to a plurality of performance indicators and second data labels corresponding thereto, including:

[0234] 1. Obtaining a third indicator data sequence and a fourth indicator data sequence corresponding to a plurality of the performance indicators;

[0235] 2. For each of the performance indicators, obtaining analysis results based on the third indicator data sequence and the fourth indicator data sequence, wherein the analysis results include expert judgment results and analysis explanation data;

[0236] 3. Combining the third indicator data sequence, the fourth indicator data sequence and the analysis and explanation data corresponding to each performance indicator to obtain second training data corresponding to the performance indicator;

[0237] 4. determining the expert judgment result corresponding to each performance indicator as a second data label corresponding to the performance indicator;

[0238] 5. Obtain second training data and second data labels corresponding to each performance indicator.

[0239] Step S304: train the pre-trained model using the second training data and the second data label to obtain the preset comparative analysis model.

[0240] The pre-trained model is further trained with the second training data and the second data label to obtain a preset comparative analysis model.

[0241] The embodiment of the present application trains the model through two stages. In the pre-training stage, the first training data and the first data labels with lower complexity are used. In the further training stage, the second training data with higher complexity and the second data labels obtained based on the expert judgment results are used. This can increase the richness of the training data and improve the robustness and accuracy of the preset comparative analysis model finally obtained.

[0242] In another embodiment of the present application, Fig.15 As shown, a data sequence comparison and analysis device is also provided, comprising:

[0243] An acquisition module 11 is used to acquire a first indicator data sequence and a second indicator data sequence for the same performance indicator;

[0244] A determination module 12, configured to generate a sequence to be classified based on the first indicator data sequence and the second indicator data sequence, wherein the sequence to be classified includes difference information between the first indicator data sequence and the second indicator data sequence;

[0245] The classification module 13 is used to input the sequence to be classified into a preset comparative analysis model, so that the preset comparative analysis model classifies the sequence to be classified to obtain a comparative analysis result.

[0246] In another embodiment of the present application, a data sequence comparison and analysis system is also provided, such as Fig.16 As shown in the figure, the data sequence comparative analysis system includes three subsystems: control subsystem, data subsystem and analysis subsystem, as well as a human expert group and a large language model similar to ChatGPT (optional) that externally supports this system. Among them, the data subsystem includes a data sample library and three modules: network data collection module, simulation data generation module, and training data summary module; the analysis subsystem includes an analysis method library and three modules: data processing module, data comparison module based on statistical analysis method, and data comparison module based on deep neural network. The data processing module includes five submodules: indicator classification submodule, task scheduling submodule, result processing submodule, analysis method submodule, and training data submodule.

[0247] The main control module of the control subsystem is responsible for the operation of the entire system process, and sends the performance indicators to be compared and analyzed to the analysis subsystem for analysis by notifying the data subsystem.

[0248] The data in the data subsystem includes performance indicators collected from the mobile communication network by the "network data collection module", performance indicators generated by the "simulation data generation module" based on the simulation algorithm, and performance indicators with analysis results collected by the "training data summary module" from the analysis subsystem and returned to the data subsystem. Performance indicators with comparative analysis results are a type of training data that can be used to train deep neural networks, where the judgment results of comparative analysis are the labels of training data. The "network data collection module" and "simulation data generation module" can also output performance indicators that can be batch-labeled with comparative judgment results and used as training data.

[0249] The performance indicators sent from the data subsystem to the analysis subsystem first enter the data processing module, where they are classified by the indicator classification submodule. Then, the task scheduling submodule assigns them to one or more of the three: an external expert group, a data comparison module based on statistical analysis methods, and a data comparison module based on deep neural networks for analysis. The data comparison module based on statistical analysis methods needs to use which statistical method to perform data analysis based on the guidance of the analysis method library in the data processing module, while the data comparison module based on deep neural networks needs to be trained and learned with a large amount of training data first. Therefore, before the analysis method library is fully built and the deep learning model based on deep neural networks is not fully trained, the analysis tasks are mainly assigned to external expert groups.

[0250] The indicator classification submodule classifies the performance indicators based on the business categories of the performance indicators as mentioned above. In addition, the importance, i.e., weight, of different performance indicators in business evaluation is also different. This submodule is used to count and control the proportion of various performance indicators in the training data, ensure the generalization performance of the training, and take into account the balance consistent with the indicator weight.

[0251] The task scheduling submodule arranges the performance indicators sent by the data subsystem to be analyzed to the external expert group or the two internal data comparison modules according to the instructions of the control subsystem. The comparative analysis task of each performance indicator can be assigned to multiple experts, each expert can analyze multiple indicators, and each expert can perform multiple comparative analyses on each indicator and feedback the results of multiple comparative analyses.

[0252] The result processing submodule mainly completes two functions. One is to summarize the results returned by the performance indicator analysis task. If a single indicator returns multiple analysis results, a voting mechanism is used to determine the final judgment result. The other is to decompose the results of the external expert analysis feedback (including the judgment results and analysis instructions). The judgment results and the corresponding two compared performance indicator sequences are sent to the training data submodule. After preprocessing, the training data is obtained and then sent to the data comparison module based on the deep neural network to perform supervised training on the deep network. The analysis instructions and corresponding performance indicators are sent to the analysis method submodule for clustering and classification of the analysis instructions, or sent to the external large language model. After understanding, refining, summarizing, and summarizing by the large language model, they are fed back to the analysis method submodule. The relevant analysis methods or processing tips that have been determined for each performance indicator are sorted out and finally imported into the analysis method library.

[0253] The analysis method submodule extracts and summarizes the prior business knowledge of each performance indicator, the analysis explanations of experts, etc., and gradually establishes, supplements, and improves the analysis methods and processing tips corresponding to each performance indicator, determines the analysis methods that can be used for the indicator (statistical comparison and / or view comparison, view comparison is the use of data comparison based on deep neural network), which factors affecting the business need to be paid attention to, which related indicators need to be paid attention to, etc. The analysis operation suggestions and opinions of these performance indicators are finally integrated into the analysis method library as a guide or reference for analysis.

[0254] The training data submodule is responsible for generating training data from the comparison data and analysis results input by the result processing submodule and sending them to the data comparison module based on the deep neural network.

[0255] In another embodiment of the present application, an electronic device is provided, including a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus;

[0256] Memory, used to store computer programs;

[0257] The processor is used to implement the data sequence comparison and analysis method described in any of the above method embodiments when executing the program stored in the memory.

[0258] In the electronic device provided by the embodiment of the present invention, the processor generates a sequence to be classified containing difference information between the first indicator data sequence and the second indicator data sequence corresponding to the performance indicator based on the program stored in the memory, and inputs the sequence to be classified into a preset comparative analysis model so that the preset comparative analysis model classifies the differences between the two indicator data sequences to obtain comparative analysis results, thereby realizing intelligent, automated and professional comparative analysis of mobile communication performance indicators, which can replace a large number of business personnel to complete a large amount of daily data comparative analysis work, while ensuring the efficiency and effectiveness of data analysis, saving a large amount of business human resource expenses.

[0259] The communication bus 1140 mentioned in the above electronic device can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The communication bus 1140 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Fig.17 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0260] The communication interface 1120 is used for communication between the above electronic device and other devices.

[0261] The memory 1130 may include a random access memory (RAM) or a non-volatile memory, such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0262] The above-mentioned processor 1110 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.

[0263] In another embodiment of the present application, a computer-readable storage medium is provided, on which a program of a comparative analysis method for data sequences is stored. When the program of the comparative analysis method for data sequences is executed by a processor, the steps of the comparative analysis method for data sequences described in any of the aforementioned method embodiments are implemented.

[0264] It should be noted that, in this article, relational terms such as "first" and "second" 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. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0265] The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A comparative analysis method for data sequences, characterized in that: include: Acquire a first indicator data series and a second indicator data series about the same performance indicator; generating a sequence to be classified based on the first indicator data sequence and the second indicator data sequence, wherein the sequence to be classified includes difference information between the first indicator data sequence and the second indicator data sequence; The sequence to be classified is input into a preset comparative analysis model, so that the preset comparative analysis model classifies the sequence to be classified to obtain a comparative analysis result.

2. The comparative analysis method of data sequence according to claim 1, characterized in that: Generating a sequence to be classified based on the first indicator data sequence and the second indicator data sequence includes: Preprocessing the first indicator data sequence and the second indicator data sequence respectively to obtain a first equal-length sequence and a second equal-length sequence; The first equal-length sequence and the second equal-length sequence are subjected to subtraction and difference processing to obtain the sequence to be classified.

3. The comparative analysis method of data sequence according to claim 2, characterized in that: Preprocessing the first indicator data sequence and the second indicator data sequence respectively to obtain a first equal-length sequence and a second equal-length sequence includes: Correcting the abnormal values ​​in the first indicator data sequence and the second indicator data sequence respectively to obtain a first corrected sequence and a second corrected sequence; Time-aligning the first corrected sequence and the second corrected sequence respectively to obtain a first aligned sequence and a second aligned sequence; The first aligned sequence and the second aligned sequence are length-aligned to obtain the first equal-length sequence and the second equal-length sequence, respectively.

4. The comparative analysis method of data sequence according to claim 2, characterized in that: Subtracting and differencing the first equal-length sequence and the second equal-length sequence to obtain the sequence to be classified includes: The first indicator data sequence and the second indicator data sequence are input into a preset subtraction difference formula, and the sequence to be classified is calculated by the preset subtraction difference formula.

5. The comparative analysis method of data sequence according to claim 1, characterized in that: The training method of the preset comparative analysis model includes: Obtaining first training data corresponding to a plurality of performance indicators and first data labels corresponding thereto; Performing supervised training on the initial model using the first training data and the first data label to obtain a pre-trained model; Acquire second training data corresponding to a plurality of performance indicators and second data labels corresponding thereto, wherein the second data labels are determined according to expert judgment results; The pre-trained model is trained using the second training data and the second data label to obtain the preset comparative analysis model.

6. The comparative analysis method of data sequence according to claim 5, characterized in that: Acquiring second training data corresponding to a plurality of performance indicators and second data labels corresponding thereto, including: Obtaining a third indicator data sequence and a fourth indicator data sequence corresponding to a plurality of the performance indicators; For each of the performance indicators, obtaining analysis results based on the third indicator data sequence and the fourth indicator data sequence, wherein the analysis results include expert judgment results and analysis explanation data; Combine the third indicator data sequence, the fourth indicator data sequence and the analysis description data corresponding to each performance indicator to obtain second training data corresponding to the performance indicator; Determine the expert judgment result corresponding to each performance indicator as a second data label corresponding to the performance indicator; The second training data and the second data label corresponding to each performance indicator are obtained.

7. The comparative analysis method of data sequence according to claim 5, characterized in that: Acquiring first training data corresponding to a plurality of performance indicators and first data labels corresponding thereto, including: Acquire a fifth indicator data sequence and a sixth indicator data sequence corresponding to a plurality of the performance indicators; Batch labeling the fifth indicator data series and the sixth indicator data series corresponding to the plurality of performance indicators to obtain a first data label corresponding to each performance indicator; Combine the fifth indicator data sequence and the sixth indicator data sequence corresponding to each of the performance indicators to obtain first training data corresponding to the performance indicator; The first training data and the first data label corresponding to each performance indicator are obtained.

8. A data sequence comparison and analysis device, characterized in that: include: An acquisition module, used to acquire a first indicator data sequence and a second indicator data sequence for the same performance indicator; A determination module, configured to generate a sequence to be classified based on the first indicator data sequence and the second indicator data sequence, wherein the sequence to be classified includes difference information between the first indicator data sequence and the second indicator data sequence; The classification module is used to input the sequence to be classified into a preset comparative analysis model so that the preset comparative analysis model classifies the sequence to be classified to obtain a comparative analysis result.

9. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, used to implement the data sequence comparison and analysis method described in any one of claims 1 to 7 when executing the program stored in the memory.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program of a comparative analysis method for data sequences, and when the program of the comparative analysis method for data sequences is executed by a processor, the steps of the comparative analysis method for data sequences described in any one of claims 1 to 7 are implemented.

Images