An industrial economic operation intelligent analysis method and system

Abstract

The application relates to the technical field of artificial intelligence, and discloses an industrial economic operation intelligent analysis method and system. The method comprises the following steps: performing data quality evaluation analysis on original data of an industrial chain to obtain a standardized data pool; performing analysis on time effectiveness index data in the standardized data pool to obtain a dynamic feature vector; further performing double-channel joint analysis on the dynamic feature vector and a time weight matrix of the industrial chain to obtain an industrial economic prediction model; performing confidence interval correlation analysis on the industrial economic prediction model in combination with an error distribution matrix to generate a risk early warning signal of the industrial chain and generate a visual decision view; and generating an economic operation analysis report of the industrial chain based on the visual decision view. The application can improve the accuracy of an industrial economic operation analysis result.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and particularly relates to an industrial economic operation intelligent analysis method and system. BACKGROUND

[0002] With the rapid development of industrial economy, the amount of data generated in the process of industrial chain operation increases exponentially, and these data cover multiple links such as production, circulation and consumption, and have characteristics such as multi-source heterogeneity and strong dynamic timeliness.

[0003] The traditional industrial economic operation analysis method mainly relies on manual statistics and static models, usually adopts a single-dimensional statistical method or a simple time series prediction model, lacks dynamic evaluation of data quality, accurate weighting of timeliness indicators and correlation analysis of multi-dimensional risks, and in the prediction modeling stage, fails to effectively fuse dynamic disturbance and long-term time series rules, affecting the robustness of prediction; in the risk warning stage, lacks correlation analysis of error distribution and confidence interval of abnormal fluctuations, and it is difficult to generate accurate warning signals, in addition, the visualization of the prior art is mainly static report, which cannot support interactive analysis of multi-dimensional spatio-temporal data, and restricts the decision-making efficiency.

[0004] Therefore, how to improve the real-time, accuracy and intelligent level of industrial economic operation analysis has become a technical problem to be solved. SUMMARY

[0005] The present application provides an industrial economic operation intelligent analysis method and system, which mainly aims to solve the problem of low efficiency in industrial economic operation intelligent analysis.

[0006] To achieve the above-mentioned purpose, the present application provides an industrial economic operation intelligent analysis method, comprising:

[0007] S1, structuring mapping and time stamp alignment are performed on the original data of the industrial chain to obtain the structured original data of the industrial chain, and data quality evaluation analysis is performed on the structured original data to obtain a standardized data pool of the industrial chain;

[0008] S2, dynamic time decay analysis is performed on the timeliness indicator data in the standardized data pool to obtain a time weight matrix of the standardized data pool, and feature weight adjustment operation is performed on the timeliness indicator data based on the time decay factor to obtain a dynamic feature vector of the standardized data pool;

[0009] S3, double-channel joint analysis is performed on the dynamic feature vector and the time weight matrix of the industrial chain to obtain an industrial economic prediction model of the industrial chain;

[0010] S4, confidence interval correlation analysis is performed on the error distribution matrix of the prediction model and the abnormal fluctuation parameters in the dynamic feature vector, and a risk early warning signal of the industrial chain is generated;

[0011] S5, based on the industry development trend curve in the industrial economic prediction model, multi-dimensional space-time data analysis is performed on the risk early warning signal to generate a visual decision view of the industrial chain, and an economic operation analysis report of the industrial chain is generated based on the visual decision view.

[0012] In a preferred embodiment, the structured mapping of the original data of the industrial chain and the time stamp alignment are performed to obtain the structured original data of the industrial chain, comprising:

[0013] S201, the original data is sorted by data source category to obtain source system label intermediate data of the original data, and the source system label intermediate data is analyzed for format characteristics to obtain formatted source system label intermediate data of the original data;

[0014] S202, the formatted source system label intermediate data is structured mapped based on a preset supply chain metadata template to obtain a standard structured data set of the original data;

[0015] S203, the standard structured data is processed for time sequence alignment to obtain the structured original data of the industrial chain.

[0016] In a preferred embodiment, the data quality evaluation analysis of the structured original data is performed to obtain the standardized data pool of the industrial chain, comprising:

[0017] S301, the quality dimension of the structured original data is calculated to obtain a quality evaluation vector of the industrial chain;

[0018] S302, the dynamic weight of the quality access benchmark value in the standardized data pool is calculated to obtain a dynamic quality threshold of the industrial chain;

[0019] S303, based on the quality evaluation vector and the dynamic quality threshold, the structured original data is processed for hierarchical cleaning and intelligent filling to obtain the standardized data pool of the industrial chain.

[0020] In a preferred embodiment, the dynamic time decay analysis of the time effectiveness index data in the standardized data pool is performed to obtain a time weight matrix of the standardized data pool, comprising:

[0021] The time effectiveness index data in the standardized data pool is extracted for time effectiveness index to obtain a time effectiveness index data set of the standardized data pool;

[0022] A decay coefficient mapping table for the industrial chain is generated based on a preset base time decay coefficient.

[0023] Based on the timeliness index dataset and the decay coefficient mapping table, an exponential decay operation is performed on the timeliness index data to obtain an initial time weight matrix. The expression for the exponential decay operation is as follows:

[0024]

[0025] In the formula, For the first The first data item The initial time weights of each feature, Features The initial weights, This refers to the industry-based attenuation coefficient in the attenuation coefficient mapping table. For the first The time difference between the time point when the data was generated and the time point when the analysis was performed. For the current analysis point in time, For the first The data collection timestamp;

[0026] The initial time weight matrix is ​​subjected to multi-dimensional standardization to obtain the time weight matrix of the standardized data pool, wherein the expression for the multi-dimensional standardization is as follows:

[0027]

[0028] In the formula, For the first Data item number Normalized weights of each feature, For the first The first data item The initial time weights of each feature, For the first The maximum weight value of a feature across all data. For the first Data item number The final standardized weights of each feature, To prevent extremely small constants with a denominator of zero.

[0029] In a preferred embodiment, the feature weight adjustment operation based on the time decay factor for the timeliness index data includes:

[0030] Based on the time difference value of each of the aforementioned timeliness indicator data, a timeliness indicator vector is generated for the timeliness indicator data, wherein the expression of the timeliness indicator vector is as follows:

[0031]

[0032] In the formula, As a timeliness indicator vector, This represents the time difference between the point when the first data point was generated and the point when the analysis was performed. The time difference between the time point when the second data point was generated and the time point when the analysis was performed. For the first The time difference between the time point when the data was generated and the time point when the analysis was performed. The symbol for the transpose of a matrix or vector;

[0033] Based on the timeliness index vector and the time weight matrix, the timeliness index data is weighted and fused with a preset feature fusion equation set to obtain a dynamic feature vector. The expression of the preset feature fusion equation set is as follows:

[0034]

[0035] In the formula, For dynamic feature vectors, For the total number of features, For feature numbering, For the first The first data item The initial time weights of each feature, The preset initial feature values, For the first The L2 norm of the weight vector of each feature. To prevent extremely small constants with a denominator of zero, For the corrected first The dynamic decay factor corresponding to each feature For the first The dynamic decay factor corresponding to each feature For the first The time difference between the time point when the data was generated and the time point when the analysis was performed. This refers to the industry-based attenuation coefficient in the attenuation coefficient mapping table. The total number of external event types. Number the external event type. For the first The influence coefficient of external events This is the event indicator function.

[0036] In a preferred embodiment, the dynamic feature vector and the time weight matrix of the industrial chain are subjected to double-channel joint analysis to obtain an industrial economic prediction model of the industrial chain, which comprises:

[0037] The dynamic feature vector set is bound and mapped to a main channel to obtain dynamic disturbance monitoring channel data of the industrial chain, the time weight matrix is bound and mapped to a secondary channel to obtain time sequence rule modeling channel data of the industrial chain, and the dynamic disturbance monitoring channel data and the time sequence rule modeling channel data are merged to obtain a double-channel input data set of the industrial chain.

[0038] The dynamic disturbance monitoring channel data and the time sequence rule modeling channel data are subjected to joint feature construction to obtain a double-channel prediction framework of the industrial chain.

[0039] Based on the double-channel input data set, the double-channel prediction framework and a preset loss function, an economic prediction model of the industrial chain is constructed.

[0040] In a preferred embodiment, the error distribution matrix of the prediction model and the abnormal fluctuation parameters in the dynamic feature vector are subjected to confidence interval correlation analysis to generate a risk warning signal of the industrial chain, which comprises:

[0041] The double-channel input data set is input into the economic prediction model to obtain a prediction value of the industrial chain.

[0042] The prediction value and the true data in the industrial chain database are subjected to error analysis to obtain an error distribution matrix of the economic prediction model.

[0043] Based on the statistical characteristics in the error distribution matrix, the error tolerance in the economic prediction model is subjected to random sampling simulation by Monte Carlo simulation to obtain a risk event occurrence probability density function of the economic prediction model.

[0044] Based on the error distribution matrix, Gaussian distribution fitting is performed on the dynamic feature vector to obtain an error confidence interval of the industrial chain.

[0045] The abnormal fluctuation parameters of the dynamic feature vector are extracted to obtain an abnormal fluctuation matrix of the dynamic feature vector.

[0046] The error confidence interval and the abnormal fluctuation matrix are subjected to cross-dimension correlation analysis to obtain an error correlation matrix of the industrial chain.

[0047] The error correlation matrix, the abnormal fluctuation matrix and the risk event occurrence probability density function are input into a Bayesian network model for training to obtain a causal correlation probability matrix.

[0048] Based on the comparison between the data in the causal correlation probability matrix and the preset risk threshold, data exceeding the risk threshold is marked as a path signal to obtain a risk early warning signal of the industrial chain.

[0049] In a preferred embodiment, the multi-dimensional spatio-temporal data analysis of the risk early warning signal generates a visual decision view of the industrial chain, including:

[0050] The geographic coding in the standardized data pool, the time weight matrix, and the risk early warning signal are aligned in the time dimension to obtain a three-dimensional time-aligned early warning signal flow of the industrial chain.

[0051] The three-dimensional time-aligned data are aggregated in the spatial dimension to obtain a spatio-temporal aggregated early warning signal flow of the industrial chain.

[0052] The dynamic feature vector and the spatio-temporal aggregated early warning signal flow are jointly rendered to generate a visual decision view of the industrial chain.

[0053] In a preferred embodiment, the economic operation analysis report of the industrial chain is generated based on the visual decision view, including:

[0054] The risk topology parameters in the visual decision view are quantitatively processed to obtain a structured risk indicator set of the industrial chain.

[0055] The standardized data pool and the structured risk indicator set are input into a preset analysis report template to obtain an initial analysis report of the industrial chain.

[0056] The initial analysis report is compliance-processed to obtain an economic operation analysis report of the industrial chain.

[0057] To solve the above problems, the present application also provides an industrial economic operation intelligent analysis system, which comprises a data standardization processing module, a data feature construction module, a double-channel joint modeling module, a risk confidence correlation early warning module, and a visual decision support module, wherein:

[0058] The data standardization processing module is used for structurally mapping and timestamp aligning the original data of an industrial chain to obtain structured original data of the industrial chain, and performing data quality evaluation analysis on the structured original data to obtain a standardized data pool of the industrial chain.

[0059] The data feature construction module is configured to perform dynamic time attenuation analysis on the timeliness index data in the standardized data pool to obtain a time weight matrix of the standardized data pool, and perform feature weight adjustment operation on the timeliness index data based on a time attenuation factor to obtain a dynamic feature vector of the standardized data pool.

[0060] The double-channel joint modeling module is configured to perform double-channel joint analysis on the dynamic feature vector and the time weight matrix of the industrial chain to obtain an industrial economic prediction model of the industrial chain.

[0061] The risk confidence correlation early warning module is configured to perform confidence interval correlation analysis on an error distribution matrix of the prediction model and an abnormal fluctuation parameter in the dynamic feature vector to generate a risk early warning signal of the industrial chain.

[0062] The visual decision support module is configured to perform multi-dimensional space-time data analysis on the risk early warning signal based on an industry development trend curve in the industrial economic prediction model to generate a visual decision view of the industrial chain, and generate an economic operation analysis report of the industrial chain based on the visual decision view.

[0063] Compared with the prior art, the present application has the following beneficial effects:

[0064] 1. The double-channel joint analysis of the dynamic disturbance monitoring channel and the time sequence law modeling channel is adopted, and the model parameters are dynamically adjusted through error distribution matrix and confidence interval correlation analysis, so as to reduce the prediction deviation and improve the stability.

[0065] 2. Based on Monte Carlo simulation and Gaussian distribution fitting, the error tolerance is quantified to generate a risk event occurrence probability density function, so as to improve the reliability of the early warning signal. BRIEF DESCRIPTION OF DRAWINGS

[0066] Figure 1 A flowchart of an industrial economic operation intelligent analysis method provided by an embodiment of the present application is shown in the figure.

[0067] Figure 2 A functional module diagram of an industrial economic operation intelligent analysis system provided by an embodiment of the present application is shown in the figure.

[0068] The implementation, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0069] It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0070] The embodiment of the present application provides an industrial economic operation intelligent analysis method. The execution subject of the industrial economic operation intelligent analysis method includes but is not limited to at least one of electronic devices such as a server and a terminal, which can be configured to execute the method provided by the embodiment of the present application. In other words, the industrial economic operation intelligent analysis method can be executed by software or hardware installed in a terminal device or a server device. The server includes but is not limited to a single server, a server cluster, a cloud server or a cloud server cluster, etc. The server can be a stand-alone server, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content distribution networks (CDN), and basic cloud computing services such as big data and artificial intelligence platforms.

[0071] Referring to Figure 1 Fig. 1 is a flowchart of an industrial economic operation intelligent analysis method provided by an embodiment of the present application. In this embodiment, the industrial economic operation intelligent analysis method includes:

[0072] S1, performing structural mapping and timestamp alignment on original data of an industrial chain to obtain structured original data of the industrial chain, performing data quality evaluation analysis on the structured original data to obtain a standardized data pool of the industrial chain;

[0073] In the embodiment of the present application, the performing structural mapping and timestamp alignment on original data of an industrial chain to obtain structured original data of the industrial chain includes:

[0074] S201, performing data source category sorting on the original data to obtain source system label intermediate data of the original data, performing format feature analysis on the source system label intermediate data to obtain formatted source system label intermediate data of the original data;

[0075] S202, performing structural mapping on the formatted source system label intermediate data based on a preset supply chain metadata template to obtain a standard structured data set of the original data;

[0076] S203, performing time sequence alignment processing on the standard structured data to obtain structured original data of an industrial chain.

[0077] It should be noted that the data source category sorting of the original data is based on the metadata characteristics of the original data to construct a three-dimensional feature space, calculate a data source similarity matrix in the feature space through an improved spectral clustering algorithm, divide the original data stream according to the data source similarity matrix, and generate source system label intermediate data containing a sorting timestamp, a data source confidence and a protocol version number.

[0078] Further, the three-dimensional feature space is composed of a supply chain system identifier, a data format fingerprint and a protocol type code.

[0079] Further, the improved spectral clustering algorithm is as follows:

[0080]

[0081] In the formula, is the data in the first row and the first column of the data source similarity matrix, is the metadata feature vector of the first data, is the metadata feature vector of the first data, is a Gaussian kernel bandwidth parameter, is a protocol type matching degree matrix, and are diagonal matrix elements.

[0082] It should be noted that the format feature analysis refers to first dynamically calling a corresponding preset analysis template and a corresponding feature extraction rule based on the protocol version number in the source system label intermediate data to generate a matrix representing data format characteristics;

[0083] Based on the eigenvalue of the similarity matrix in the spectral clustering sorting process, the matrix representing data format characteristics is normalized to generate a formatted source system label intermediate data containing a protocol feature vector, a normalized timestamp sequence, a normalized confidence index and a format.

[0084] It should be noted that the structured mapping refers to dynamically matching the protocol feature vector in the formatted source system label intermediate data with the field definition in the metadata template based on the attention mechanism to generate a field mapping weight matrix, performing multi-modal field alignment, relationship graph construction and dynamic weight injection on the entity types defined in the metadata template based on the field mapping weight matrix, to obtain a standard structured data set.

[0085] Further, the multi-modal field alignment refers to dynamically selecting a data cleaning rule set using the encoding format identifier in the protocol feature vector to convert the original field value to a standard format that meets the template constraints.

[0086] Furthermore, the relationship graph construction refers to establishing weighted time-space association edges based on standardized timestamp sequences and data source confidence indicators to generate the initial topological structure of the supply chain entity relationship graph.

[0087] In this embodiment of the invention, the step of performing data quality evaluation and analysis on the structured raw data to obtain a standardized data pool for the industry chain includes:

[0088] S301, perform quality dimension calculations on the structured raw data to obtain the quality evaluation vector of the industrial chain;

[0089] S302, dynamically weight the quality access benchmark values ​​in the standardized data pool to obtain the dynamic quality threshold of the industry chain;

[0090] S303, based on the quality evaluation vector and the dynamic quality threshold, the structured raw data is subjected to hierarchical cleaning and intelligent filling processing to obtain a standardized data pool for the industrial chain.

[0091] It should be noted that the expression for calculating the quality dimension is as follows:

[0092]

[0093] In the formula, For the first Quality evaluation vector of each data point This represents the total number of quality dimensions. Number the quality dimensions. For the first Each dimension in time The dynamic weighting coefficients, For the first Feature functions of each dimension For information entropy weighting coefficients, For the information entropy of data, As a weight sensitivity adjustment factor, For dimension The anomaly rate within the time window, For the first The probability of each field value occurring. Total number of fields Number the field.

[0094] It should be noted that the expression for dynamic weight calculation is as follows:

[0095]

[0096] In the formula, a dynamic quality threshold value for time t, a current data pool quality vector mean, a quality vector standard deviation, a standard deviation scaling coefficient, a time decay intensity factor, a decay rate parameter, a quality evaluation vector of the i-th data, a total number of data, a data number, wherein, by default, β = 0.5.

[0097] Further, the hierarchical cleaning and intelligent filling processing is based on a dynamic quality threshold value to perform two-level hierarchical processing on data, wherein the first level is data cleaning, and the second level is data filling.

[0098] Further, the data cleaning refers to directly marking the data with a hash fingerprint and storing it in the database if the quality score of the data is higher than the upper limit of the threshold value.

[0099] Data filling refers to using a gradient descent algorithm with total variation regularization constraint to intelligently fill in missing values for data within the threshold interval, and maintaining business logic consistency of the repaired data through a field importance weight matrix.

[0100] S2, performing dynamic time decay analysis on the time-sensitive index data in the standardized data pool to obtain a time weight matrix of the standardized data pool, and performing feature weight adjustment operation on the time-sensitive index data based on a time decay factor to obtain a dynamic feature vector of the standardized data pool;

[0101] In the embodiment of the present application, the dynamic time decay analysis on the time-sensitive index data in the standardized data pool to obtain the time weight matrix of the standardized data pool comprises:

[0102] Performing time-sensitive index extraction on the time-sensitive index data in the standardized data pool to obtain a time-sensitive index data set of the standardized data pool;

[0103] Generating a decay coefficient mapping table of the industrial chain based on a preset basic time decay coefficient;

[0104] Performing exponential decay operation on the time-sensitive index data based on the time-sensitive index data set and the decay coefficient mapping table to obtain an initial time weight matrix, wherein the expression of the exponential decay operation is as follows:

[0105]

[0106] In the formula, is the i-th data in the j-th data set, is the j-th data in the i-th data set, an initial time weight number of the i-th feature, an initial weight of the feature , an industry-based decay coefficient in the decay coefficient mapping table, a time difference between the i-th data generation time point and the analysis time point, the current analysis time point, the current analysis time point, a collection timestamp of the i-th data;

[0107] performing multi-dimensional standardization processing on the initial time weight matrix to obtain a time weight matrix of the standardized data pool, wherein an expression of the multi-dimensional standardization processing is as follows:

[0108]

[0109] In the formula, a normalized weight of the i-th feature of the i-th data, an initial time weight number of the i-th feature of the i-th data, a maximum weight value of the i-th feature in all data, a final standardized weight of the i-th feature of the i-th data, a final standardized weight of the i-th feature of the i-th data, a final standardized weight of the i-th feature of the i-th data, a final standardized weight of the i-th feature of the i-th data, a final standardized weight of the i-th feature of the i-th data, a final standardized weight of the i-th feature of the i-th data, a final standardized weight of the i-th feature of the i-th data, a final standardized weight of the i-th feature of the i-th data, a final standardized weight of the i-th feature of the i-th data.

[0110] It should be noted that the timeliness index extraction refers to filtering out index data containing time-sensitive features from the standardized data pool based on a preset timeliness dimension classification model, quantifying the index data containing time-sensitive features, and generating an initial timeliness index value in combination with the preset basic time decay coefficient, wherein the dimensions include a data generation timestamp, an update frequency, a validity period, and a business scenario time constraint.

[0111] Further, the preset basic time decay coefficient is based on data of a historical economic operation analysis report, extracts timeliness decay rules of key indicators under different industry categories, adopts an exponential decay model to fit a value decay curve of each indicator over time, generates an industry-scenario two-dimensional weight coefficient through an entropy weight method based on a preset business scenario priority and a time sensitivity label defined by a supply chain metadata template, and performs weighted fusion of the industry-scenario two-dimensional weight coefficient and a business rule quantitative value to obtain a timeliness index data set of the standardized data pool, wherein the key indicators include an inventory turnover rate, a price fluctuation period, and a supply-demand response delay.​

[0112] It should be noted that the method for generating the attenuation coefficient mapping table is as follows: a three-level mapping structure is constructed based on the industry classification system to obtain the industry characteristic data. The preset basic time attenuation coefficient is initially allocated according to the industry characteristic data, and the attenuation coefficient mapping table of the industrial chain is generated by combining the timeliness index data extracted from the standardized data pool.

[0113] In this embodiment of the invention, the feature weight adjustment operation based on the time decay factor for the timeliness index data includes:

[0114] Based on the time difference value of each of the aforementioned timeliness indicator data, a timeliness indicator vector is generated for the timeliness indicator data, wherein the expression of the timeliness indicator vector is as follows:

[0115]

[0116] In the formula, As a timeliness indicator vector, This represents the time difference between the point when the first data point was generated and the point when the analysis was performed. The time difference between the time point when the second data point was generated and the time point when the analysis was performed. For the first The time difference between the time point when the data was generated and the time point when the analysis was performed. The symbol for the transpose of a matrix or vector;

[0117] Based on the timeliness index vector and the time weight matrix, the timeliness index data is weighted and fused with a preset feature fusion equation set to obtain a dynamic feature vector. The expression of the preset feature fusion equation set is as follows:

[0118]

[0119] In the formula, For dynamic feature vectors, For the total number of features, For feature numbering, For the first The first data item The initial time weights of each feature, The preset initial feature values, For the first The L2 norm of the weight vector of each feature. To prevent extremely small constants with a denominator of zero, For the corrected first The dynamic decay factor corresponding to each feature For the first a dynamic attenuation factor corresponding to each feature, a first a time difference between the data generation time point and the analysis time point, a basic attenuation coefficient in the attenuation coefficient mapping table, a total number of external event types, an external event type number, a first an influence coefficient of the external event, an event indication function.

[0120] It should be noted that the corrected dynamic attenuation factor is dynamically adjusted based on the real-time monitored external event flow, such as policy release, device fault alarm, market volatility mutation, etc.

[0121] S3, double-channel joint analysis is performed on the dynamic feature vector and the time weight matrix of the industrial chain to obtain an industrial economic prediction model of the industrial chain;

[0122] In the embodiment of the present application, the double-channel joint analysis of the dynamic feature vector and the time weight matrix of the industrial chain to obtain the industrial economic prediction model of the industrial chain comprises:

[0123] The dynamic feature vector set is bound and mapped to the main channel to obtain dynamic disturbance monitoring channel data of the industrial chain, the time weight matrix is bound and mapped to the secondary channel to obtain time sequence rule modeling channel data of the industrial chain, and the dynamic disturbance monitoring channel data and the time sequence rule modeling channel data are merged to obtain a double-channel input data set of the industrial chain;

[0124] Joint feature construction is performed on the dynamic disturbance monitoring channel data and the time sequence rule modeling channel data to obtain a double-channel prediction framework of the industrial chain;

[0125] Based on the double-channel input data set, the double-channel prediction framework and a preset loss function, an economic prediction model of the industrial chain is constructed.

[0126] It should be noted that the main channel is a dynamic disturbance monitoring channel, and the secondary channel is a time sequence rule modeling channel.

[0127] It should be noted that the joint feature construction is based on a cross-attention mechanism to realize double-channel feature fusion, wherein the arithmetic expression of the cross-attention mechanism is as follows:

[0128]

[0129] In the formula, ​The main and auxiliary channel feature interaction matrix, is a normalization function, is a main channel query vector, is an auxiliary channel query vector, is an auxiliary channel value vector, is a scaling factor.

[0130] Further, based on the predicted value of the economic prediction model, a dynamic weight adjustment algorithm is used to adjust the time weight matrix to obtain a predicted adjustment decay coefficient, and the industry basic decay coefficient in the decay coefficient mapping table is replaced by the predicted adjustment decay coefficient to achieve optimization of the time weight, wherein the dynamic weight adjustment algorithm is as follows:

[0131]

[0132] wherein, is a predicted adjustment decay coefficient, is an industry basic decay coefficient in the decay coefficient mapping table, is a gradient of the loss function to the time weight, is a learning rate, wherein the learning rate is preset to .

[0133] It should be noted that the expression of the preset loss function is as follows:

[0134]

[0135] wherein, is a loss function output value, is a total number of features, is a feature number, is a real economic indicator, is a predicted value, is a loss adjustment parameter, is a variance calculation symbol, is the initial time weight number of the first feature in the first data, is the initial time weight number of the first feature in the first data, is the initial time weight number of the first feature in the first data, is the initial time weight number of the first feature in the first data, is a total number of data, is a data number.

[0136] Further, the loss adjustment parameter is a parameter for dynamically adjusting the balance of the two types of losses according to the industry volatility rate.

[0137] ​S4, performing confidence interval correlation analysis on the error distribution matrix of the prediction model and the abnormal fluctuation parameters in the dynamic feature vector to generate a risk early warning signal of the industrial chain;

[0138] In the embodiment of the application, the confidence interval correlation analysis on the error distribution matrix of the prediction model and the abnormal fluctuation parameters in the dynamic feature vector to generate a risk early warning signal of the industrial chain comprises:

[0139] inputting the double-channel input data set into the economic prediction model to obtain a prediction value of the industrial chain;

[0140] performing error analysis on the prediction value and real data in an industrial chain database to obtain an error distribution matrix of the economic prediction model;

[0141] based on the statistical characteristics in the error distribution matrix, performing random sampling simulation on the error tolerance in the economic prediction model by Monte Carlo simulation to obtain a risk event occurrence probability density function of the economic prediction model;

[0142] based on the error distribution matrix, performing Gaussian distribution fitting on the dynamic feature vector to obtain an error confidence interval of the industrial chain;

[0143] performing abnormal fluctuation parameter extraction on the dynamic feature vector to obtain an abnormal fluctuation matrix of the dynamic feature vector;

[0144] performing cross-dimension correlation analysis on the error confidence interval and the abnormal fluctuation matrix to obtain an error correlation matrix of the industrial chain;

[0145] inputting the error correlation matrix, the abnormal fluctuation matrix and the risk event occurrence probability density function into a Bayesian network model for training to obtain a causal correlation probability matrix;

[0146] based on the data in the causal correlation probability matrix and a preset risk threshold, marking data exceeding the risk threshold as a path signal to obtain a risk early warning signal of the industrial chain.

[0147] It should be noted that the error analysis is performed by constructing an initial error vector through multi-dimensional error index calculation, combining the time decay coefficient and the preset business priority coefficient to generate a dynamically weighted error evaluation result, reorganizing the dynamically weighted error evaluation result according to a three-dimensional structure of time slicing, geographical grid and industry classification to generate an error distribution matrix.

[0148] Further, the calculation of the multi-dimensional error index is based on the point-by-point difference between the predicted value and the true data, and the time dimension error, the space dimension error and the industry subdivision error are calculated synchronously to generate initial error data, and an initial error vector is generated based on the initial error data.

[0149] It should be noted that the specific process of Gaussian distribution fitting is as follows: the error mean and error standard deviation of the dynamic feature vector are calculated based on the error distribution matrix, the error mean and the error standard deviation are weighted by using the time decay coefficient to obtain weighted error parameters, and the error confidence interval of the industrial chain is generated based on the weighted error parameters and preset fitting parameters.

[0150] It should be noted that the cross-dimension correlation analysis is to construct a correlation tensor by using the deviation degree of the error confidence interval and the abnormal fluctuation matrix, and to construct a directed acyclic graph based on Granger causality test to obtain a causal coefficient, and to generate an error correlation matrix based on the correlation tensor and the causal coefficient.

[0151] Further, the Granger causality test is a method for analyzing the causal relationship between variables, which is mainly used for analyzing the causal relationship between economic variables.

[0152] S5, based on the industry development trend curve in the industrial economic prediction model, the multi-dimensional space-time data analysis of the risk early warning signal is performed to generate a visual decision view of the industrial chain, and an economic operation analysis report of the industrial chain is generated based on the visual decision view.

[0153] In the embodiment of the application, the multi-dimensional space-time data analysis of the risk early warning signal to generate the visual decision view of the industrial chain comprises:

[0154] The geographic coding in the standardized data pool, the time weight matrix and the risk early warning signal are time-dimensionally aligned to obtain a three-dimensional time-aligned early warning signal flow of the industrial chain;

[0155] The three-dimensional time-aligned data are spatially dimensionally aggregated to obtain a space-time aggregated early warning signal flow of the industrial chain;

[0156] The dynamic feature vector and the space-time aggregated early warning signal flow are jointly rendered to generate the visual decision view of the industrial chain.

[0157] It should be noted that the spatial dimension aggregation is to weight the spatial relationship topology in the three-dimensional time alignment data and the supply chain metadata template, generate a regional level early warning strength, generate a warning data decay curve based on the regional level early warning strength and the causal time delay parameter in the error correlation matrix, and generate a spatio-temporal aggregation early warning signal flow of the industrial chain based on the data in the data decay curve.

[0158] It should be noted that the joint rendering refers to displaying the dynamic feature vector and the data in the spatio-temporal aggregation early warning signal in a three-dimensional cubic view through a multi-modal rendering engine.

[0159] In the embodiment of the application, the economic operation analysis report of the industrial chain is generated based on the visual decision view, comprising:

[0160] The risk topology parameters in the visual decision view are quantitatively processed to obtain a structured risk indicator set of the industrial chain.

[0161] The standardized data pool and the structured risk indicator set are input into a preset analysis report template to obtain an initial analysis report of the industrial chain.

[0162] The initial analysis report is compliance-processed to obtain an economic operation analysis report of the industrial chain.

[0163] It should be noted that the analysis report template is obtained by a structured rule engine, multi-modal data fusion and dynamic adaptation mechanism, and the standardized data pool and risk indicators are converted into a decision report meeting business requirements.

[0164] It should be noted that the compliance processing is a data conversion method for converting the initial analysis report into an economic operation analysis report meeting regulations, safety standards and business specifications.

[0165] Compared with the prior art, the present application has the following beneficial effects:

[0166] 1. A dual-channel joint analysis combining a dynamic disturbance monitoring channel and a time sequence law modeling channel is adopted, and through error distribution matrix and confidence interval correlation analysis, model parameters are dynamically adjusted, prediction deviation is reduced, and stability is improved.

[0167] 2. Based on Monte Carlo simulation and Gaussian distribution fitting, error tolerance is quantified, a risk event occurrence probability density function is generated, and the reliability of the early warning signal is improved.

[0168] As shown in Figure 2 Fig. 1 is a functional module diagram of an industrial economic operation intelligent analysis system according to an embodiment of the present application.

[0169] The industrial economic operation intelligent analysis system 100 can be installed in an electronic device. According to the functions implemented, the industrial economic operation intelligent analysis system 100 can include a data standardization processing module 101, a data feature construction module 102, a double-channel joint modeling module 103, a risk confidence correlation early warning module 104, and a visual decision support module 105. The modules disclosed in the present application can also be referred to as units, which refer to a series of computer program segments that can be executed by an electronic device processor and can complete a fixed function, which are stored in the memory of the electronic device.

[0170] In the present embodiment, the functions of each module / unit are as follows:

[0171] The data standardization processing module is configured to perform structured mapping and timestamp alignment on the original data of the industrial chain to obtain structured original data of the industrial chain, and perform data quality evaluation analysis on the structured original data to obtain standardized data pool of the industrial chain.

[0172] The data feature construction module is configured to perform dynamic time decay analysis on the time-sensitive index data in the standardized data pool to obtain a time weight matrix of the standardized data pool, and perform feature weight adjustment operation based on a time decay factor on the time-sensitive index data to obtain a dynamic feature vector of the standardized data pool.

[0173] The double-channel joint modeling module is configured to perform double-channel joint analysis on the dynamic feature vector and the time weight matrix of the industrial chain to obtain an industrial economic prediction model of the industrial chain.

[0174] The risk confidence correlation early warning module is configured to perform confidence interval correlation analysis on the error distribution matrix of the prediction model and the abnormal fluctuation parameters in the dynamic feature vector to generate a risk early warning signal of the industrial chain.

[0175] The visual decision support module is configured to perform multi-dimensional spatio-temporal data analysis on the risk early warning signal based on the industry development trend curve in the industrial economic prediction model to generate a visual decision view of the industrial chain, and generate an economic operation analysis report of the industrial chain based on the visual decision view.

[0176] In several embodiments provided in the present application, it should be understood that the disclosed method and system can be implemented in other ways. For example, the system embodiments described above are only illustrative, for example, the division of the modules is only a logical function division, and another division method can be used in actual implementation.

[0177] The modules described as separate components may or may not be physically separate, and the components displayed as modules may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0178] In addition, each functional module in various embodiments of the application can be integrated in one processing unit, or each unit can exist physically independently, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of hardware plus software functional modules.

[0179] It is obvious for those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application.

[0180] The embodiments of the present application can acquire and process related data based on artificial intelligence technology. Among them, artificial intelligence is to use digital computers or machine controlled by digital computers to simulate, extend and expand human intelligence, perceive environment, obtain knowledge and use knowledge to obtain the best results.

[0181] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. An intelligent analysis method for industrial economy operation, characterized in that, The method comprises: S1, the original data of the industrial chain is structured and mapped and time stamped to obtain structured original data of the industrial chain, and the data quality of the structured original data is evaluated and analyzed to obtain a standardized data pool of the industrial chain; S2, dynamic time decay analysis is performed on the time-sensitive index data in the standardized data pool to obtain a time weight matrix of the standardized data pool, and feature weight adjustment operation is performed on the time-sensitive index data based on a time decay factor to obtain a dynamic feature vector of the standardized data pool; The dynamic time decay analysis on the time-sensitive index data in the standardized data pool to obtain the time weight matrix of the standardized data pool comprises: Time-sensitive index extraction is performed on the time-sensitive index data in the standardized data pool to obtain a time-sensitive index data set of the standardized data pool; A decay coefficient mapping table of the industrial chain is generated based on a preset basic time decay coefficient; Exponential decay operation is performed on the time-sensitive index data based on the time-sensitive index data set and the decay coefficient mapping table to obtain an initial time weight matrix, wherein the expression of the exponential decay operation is as follows: ， In the formula, is the first is the initial time weight of the first characteristic in the first piece of data, is the initial weight of the characteristic, is the industry-based decay coefficient in the decay coefficient mapping table, is the time difference between the time point when the first piece of data is generated and the analysis time point, is the current analysis time point, is the collection timestamp of the first piece of data. Multi-dimensional standardization processing is performed on the initial time weight matrix to obtain the time weight matrix of the standardized data pool, wherein the expression of the multi-dimensional standardization processing is as follows: ， wherein, is the normalized weight of the kth feature of the ith data, is the initial time weight number of the kth feature in the ith data, is the maximum weight value of the kth feature in all data, is the final normalized weight of the kth feature of the ith data, is a very small constant to prevent the denominator from being zero;​​ The feature weight adjustment operation on the time-sensitive index data based on the time decay factor comprises: A time-sensitive index vector of the time-sensitive index data is generated based on the time difference value of each time-sensitive index data, wherein the expression of the time-sensitive index vector is as follows: ， In the formula, is an aging index vector, is the time difference between the first data generation time point and the analysis time point, is the time difference between the second data generation time point and the analysis time point, is the time difference between the third data generation time point and the analysis time point, is the time difference between the fourth data generation time point and the analysis time point, is a matrix vector transposition symbol; The time-sensitive index data and a preset feature fusion equation set are weighted and fused based on the time-sensitive index vector and the time weight matrix to obtain a dynamic feature vector, wherein the expression of the preset feature fusion equation set is as follows: ， In the formula, For dynamic feature vectors, For the total number of features, For feature numbering, For the first The first data item The initial time weights of each feature, The preset initial feature values, For the first The L2 norm of the weight vector of each feature. To prevent extremely small constants with a denominator of zero, This is the dynamic attenuation factor corresponding to the k-th feature after correction. The dynamic decay factor corresponding to the k-th feature. For the first The time difference between the time point when the data was generated and the time point when the analysis was performed. This refers to the industry-based attenuation coefficient in the attenuation coefficient mapping table. The total number of external event types. Number the external event type. Let m be the influence coefficient of the m-th type of external event. This is an event indicator function; S3, double-channel joint analysis is performed on the dynamic feature vector and the time weight matrix of the industrial chain to obtain an industrial economic prediction model of the industrial chain; The double-channel joint analysis on the dynamic feature vector and the time weight matrix of the industrial chain to obtain the industrial economic prediction model of the industrial chain comprises: The dynamic feature vector set is bound and mapped with a main channel to obtain dynamic disturbance monitoring channel data of the industrial chain, the time weight matrix is bound and mapped with a secondary channel to obtain time sequence rule modeling channel data of the industrial chain, and the dynamic disturbance monitoring channel data and the time sequence rule modeling channel data are merged to obtain a double-channel input data set of the industrial chain; Joint feature construction is performed on the dynamic disturbance monitoring channel data and the time sequence rule modeling channel data to obtain a double-channel prediction framework of the industrial chain; An economic prediction model of the industrial chain is constructed based on the double-channel input data set, the double-channel prediction framework, and a preset loss function; S4, confidence interval correlation analysis is performed on the error distribution matrix of the prediction model and the abnormal fluctuation parameters in the dynamic feature vector, and a risk early warning signal of the industrial chain is generated; The confidence interval correlation analysis on the error distribution matrix of the prediction model and the abnormal fluctuation parameters in the dynamic feature vector to generate the risk early warning signal of the industrial chain comprises: The double-channel input data set is input into the economic prediction model to obtain the prediction value of the industrial chain; Error analysis is performed on the prediction value and the true data in the industrial chain database to obtain the error distribution matrix of the economic prediction model; Based on the statistical characteristics in the error distribution matrix, the error tolerance in the economic prediction model is simulated by random sampling using Monte Carlo simulation to obtain the risk event occurrence probability density function of the economic prediction model; Based on the error distribution matrix, Gaussian distribution fitting is performed on the dynamic feature vector to obtain the error confidence interval of the industrial chain; Abnormal fluctuation parameter extraction is performed on the dynamic feature vector to obtain the abnormal fluctuation matrix of the dynamic feature vector; Cross-dimension correlation analysis is performed on the error confidence interval and the abnormal fluctuation matrix to obtain the error correlation matrix of the industrial chain; The error correlation matrix, the abnormal fluctuation matrix and the risk event occurrence probability density function are input into the Bayesian network model for training to obtain the causal correlation probability matrix; Based on the data in the causal correlation probability matrix and the preset risk threshold, data exceeding the risk threshold is marked as a path signal to obtain the risk early warning signal of the industrial chain; S5, based on the industry development trend curve in the industrial economic prediction model, multi-dimensional spatio-temporal data analysis is performed on the risk early warning signal to generate a visual decision view of the industrial chain, and an economic operation analysis report of the industrial chain is generated based on the visual decision view.

2. The method of claim 1, wherein the method comprises: The structured mapping and timestamp alignment of the original data of the industrial chain to obtain the structured original data of the industrial chain comprises: S201, data source category sorting is performed on the original data to obtain source system label intermediate data of the original data, and format feature analysis is performed on the source system label intermediate data to obtain formatted source system label intermediate data of the original data; S202, based on a preset supply chain metadata template, the formatted source system label intermediate data is structured mapped to obtain a standard structured data set of the original data; S203, time sequence alignment processing is performed on the standard structured data to obtain the structured original data of the industrial chain.

3. The intelligent analysis method for industrial economic operation as described in claim 2, characterized in that, The data quality evaluation analysis of the structured original data to obtain the standardized data pool of the industrial chain comprises: S301, quality dimension calculation is performed on the structured original data to obtain a quality evaluation vector of the industrial chain; S302, dynamic weight calculation is performed on the quality access benchmark value in the standardized data pool to obtain a dynamic quality threshold of the industrial chain; S303, based on the quality evaluation vector and the dynamic quality threshold, the structured raw data is graded and cleaned and intelligently filled to obtain the standardized data pool of the industrial chain.

4. The intelligent analysis method for industrial economic operation as described in claim 1, characterized in that, The multi-dimensional space-time data analysis of the risk early warning signal generates a visual decision view of the industrial chain, including: The time dimension alignment of the geographic coding in the standardized data pool, the time weight matrix, and the risk early warning signal is performed to obtain a three-dimensional time alignment early warning signal flow of the industrial chain; The spatial dimension aggregation of the three-dimensional time alignment data is performed to obtain a space-time aggregated early warning signal flow of the industrial chain; The dynamic feature vector and the space-time aggregated early warning signal flow are jointly rendered to generate a visual decision view of the industrial chain.

5. The method of claim 4, wherein the method further comprises: The economic operation analysis report of the industrial chain is generated based on the visual decision view, including: ​ The risk topology parameters in the visual decision view are quantitatively processed to obtain a structured risk indicator set of the industrial chain; The standardized data pool and the structured risk indicator set are input into a preset analysis report template to obtain an initial analysis report of the industrial chain; The initial analysis report is processed to obtain an economic operation analysis report of the industrial chain.

6. An industrial economy operation intelligent analysis system, characterized in that, The system includes a data standardization processing module, a data feature construction module, a double-channel joint modeling module, a risk confidence correlation early warning module, and a visual decision support module, wherein: The data standardization processing module is configured to structure map and timestamp align the raw data of the industrial chain to obtain structured raw data of the industrial chain, and perform data quality evaluation analysis on the structured raw data to obtain a standardized data pool of the industrial chain; The data feature construction module is configured to perform dynamic time decay analysis on the time-sensitive indicator data in the standardized data pool to obtain a time weight matrix of the standardized data pool, and perform feature weight adjustment operation on the time-sensitive indicator data based on a time decay factor to obtain a dynamic feature vector of the standardized data pool; The dynamic time decay analysis on the time-sensitive indicator data in the standardized data pool to obtain the time weight matrix of the standardized data pool includes: Time-sensitive indicator extraction is performed on the time-sensitive indicator data in the standardized data pool to obtain a time-sensitive indicator data set of the standardized data pool; An attenuation coefficient mapping table of the industrial chain is generated based on a preset basic time decay coefficient; Based on the time-sensitive indicator data set and the attenuation coefficient mapping table, an exponential decay operation is performed on the time-sensitive indicator data to obtain an initial time weight matrix, wherein the expression of the exponential decay operation is as follows: ， In the formula, For the first The first data item The initial time weights of each feature, Features The initial weights, This refers to the industry-based attenuation coefficient in the attenuation coefficient mapping table. For the first The time difference between the time point when the data was generated and the time point when the analysis was performed. For the current analysis point in time, For the first The data collection timestamp; The multi-dimensional standardization processing of the initial time weight matrix is performed to obtain the time weight matrix of the standardized data pool, wherein the expression of the multi-dimensional standardization processing is as follows: ， wherein, is the normalized weight of the kth feature of the ith data, is the initial time weight number of the kth feature in the ith data, is the initial time weight number of the kth feature in the ith data, is the initial time weight number of the kth feature in the ith data, is the maximum weight value of the kth feature in all data, is the final normalized weight of the kth feature of the ith data, is a very small constant to prevent the denominator from being zero; The feature weight adjustment operation on the time-sensitive indicator data based on the time decay factor includes: generate a timeliness index vector of the timeliness index data based on a time difference value of each piece of the timeliness index data, where an expression of the timeliness index vector is as follows: ， In the formula, is an aging index vector, is the time difference between the first data generation time point and the analysis time point, is the time difference between the second data generation time point and the analysis time point, is the time difference between the third data generation time point and the analysis time point, is the time difference between the fourth data generation time point and the analysis time point, is a matrix vector transposition symbol; perform feature fusion on the timeliness index data and a preset feature fusion equation group by weight based on the timeliness index vector and the time weight matrix to obtain a dynamic feature vector, where an expression of the preset feature fusion equation group is as follows: ， In the formula, For dynamic feature vectors, For the total number of features, For feature numbering, For the first The first data item The initial time weights of each feature, The preset initial feature values, For the first The L2 norm of the weight vector of each feature. To prevent extremely small constants with a denominator of zero, This is the dynamic attenuation factor corresponding to the k-th feature after correction. The dynamic decay factor corresponding to the k-th feature. For the first The time difference between the time point when the data was generated and the time point when the analysis was performed. This refers to the industry-based attenuation coefficient in the attenuation coefficient mapping table. The total number of external event types. Number the external event type. Let m be the influence coefficient of the m-th type of external event. This is an event indicator function; perform double-channel joint analysis on the dynamic feature vector and the time weight matrix of the industrial chain to obtain an industrial economic prediction model of the industrial chain. perform double-channel joint analysis on the dynamic feature vector and the time weight matrix of the industrial chain to obtain an industrial economic prediction model of the industrial chain, including: bind and map the dynamic feature vector set to a main channel to obtain dynamic disturbance monitoring channel data of the industrial chain, bind and map the time weight matrix to a secondary channel to obtain time sequence rule modeling channel data of the industrial chain, and merge the dynamic disturbance monitoring channel data and the time sequence rule modeling channel data to obtain a double-channel input data set of the industrial chain; perform joint feature construction on the dynamic disturbance monitoring channel data and the time sequence rule modeling channel data to obtain a double-channel prediction framework of the industrial chain; construct an economic prediction model of the industrial chain based on the double-channel input data set, the double-channel prediction framework, and a preset loss function; perform confidence interval correlation analysis on an error distribution matrix of the prediction model and an abnormal fluctuation parameter in the dynamic feature vector to generate a risk early warning signal of the industrial chain. perform confidence interval correlation analysis on an error distribution matrix of the prediction model and an abnormal fluctuation parameter in the dynamic feature vector to generate a risk early warning signal of the industrial chain, including: input the double-channel input data set into the economic prediction model to obtain a predicted value of the industrial chain; perform error analysis on the predicted value and real data in an industrial chain database to obtain an error distribution matrix of the economic prediction model; based on statistical characteristics in the error distribution matrix, perform random sampling simulation on error tolerance in the economic prediction model by using Monte Carlo simulation to obtain a risk event occurrence probability density function of the economic prediction model; based on the error distribution matrix, perform Gaussian distribution fitting on the dynamic feature vector to obtain an error confidence interval of the industrial chain; perform abnormal fluctuation parameter extraction on the dynamic feature vector to obtain an abnormal fluctuation matrix of the dynamic feature vector; perform cross-dimension correlation analysis on the error confidence interval and the abnormal fluctuation matrix to obtain an error correlation matrix of the industrial chain; input the error correlation matrix, the abnormal fluctuation matrix, and the risk event occurrence probability density function into a Bayesian network model for training to obtain a cause-effect correlation probability matrix; and perform cross-dimension correlation analysis on the error confidence interval and the abnormal fluctuation matrix to obtain an error correlation matrix of the industrial chain. Based on comparison between data in the causal correlation probability matrix and a preset risk threshold, data exceeding the risk threshold is marked as a path signal to obtain a risk early warning signal of the industrial chain; The visual decision support module is configured to perform multi-dimensional space-time data analysis on the risk early warning signal based on an industry development trend curve in the industrial economic prediction model, generate a visual decision view of the industrial chain, and generate an economic operation analysis report of the industrial chain based on the visual decision view.

Images