Industrial chain strong chain, chain extension and chain supplement analysis method, system and equipment and medium
By integrating time-series data on logical and physical relationships between enterprises, a relational potential energy index is generated, which solves the problem of difficulty in integrating multi-source heterogeneous data in existing technologies. This enables dynamic trend quantitative analysis of relationships between enterprises and supports precise governance of the industrial chain.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-17
- Publication Date
- 2026-04-10
AI Technical Summary
Existing technologies struggle to effectively integrate multi-source heterogeneous time-series data and lack the technical means to quantify and proactively predict the evolution of relationships between enterprises.
By acquiring time-series data on logical and physical relationships between enterprises, confidence time series are calculated, and time-series fusion models such as Kalman filtering algorithms are used for calibration to generate comprehensive confidence evolution curves and calculate relationship potential indicators to analyze the industrial chain.
It enables dynamic trend quantitative analysis of relationships between enterprises, identifies potential connection opportunities and decline risks, and provides data-driven objective basis for decisions on strengthening, extending, and supplementing the industrial chain.
Smart Images
Figure CN121836092A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of data processing and analysis, in particular to an industrial chain strengthening, extending and supplementing analysis method, system, device and medium. BACKGROUND
[0002] The stability and development of the industrial chain is a key indicator of the national economic competitiveness. The formulation and implementation of industrial policies such as "strengthening, extending and supplementing the chain" highly depend on the accurate, dynamic and forward-looking analysis of the correlation between enterprises in the industrial chain.
[0003] In the prior art, industrial chain analysis has gradually evolved from traditional static analysis based on input-output tables or industry reports to data-driven models. For example, some technical solutions use structured data such as equity investment between enterprises, supply chain records or patent citations to construct an industrial chain network map to identify core enterprises and key nodes. These methods have improved the objectivity and timeliness of industrial chain analysis to some extent.
[0004] However, the above prior art solutions generally have a core technical problem in dealing with the dynamic evolution of the relationship between enterprises and predicting future trends: there is a lack of a technical means that can effectively fuse multi-source heterogeneous time series data and extract quantitative indicators with forward-looking significance from them. SUMMARY
[0005] The present application provides an industrial chain strengthening, extending and supplementing analysis method, system, device and medium to solve the problem that the prior art solutions cannot effectively fuse multi-source heterogeneous time series data and extract comprehensive indicators that can quantify and forward-lookingly predict the evolution trend of the relationship between enterprises.
[0006] In view of the above problems, the present application provides an industrial chain strengthening, extending and supplementing analysis method, comprising the following steps: Obtaining a first group of time series data reflecting the logical correlation between enterprises in the industrial chain, and a second group of time series data reflecting the physical correlation between the enterprises; Based on the first group of time series data, a first confidence time series representing the logical correlation strength between the enterprises is calculated; Based on the second group of time series data, a second confidence time series representing the physical correlation strength between the enterprises is calculated; Fusing the first confidence time series and the second confidence time series to generate a comprehensive confidence evolution curve; Based on the comprehensive confidence evolution curve, a relationship potential index representing the evolution trend of the relationship between the enterprises is calculated; According to the relationship potential index, the industrial chain is analyzed for strengthening, extending or supplementing.
[0007] Preferably, the fusion processing step is implemented by a time series fusion model configured to process the time delay and data asynchrony between the first and second confidence time series.
[0008] Preferably, the time series fusion model employs a Kalman filter algorithm or its variants to iteratively calibrate the first and second confidence time series through prediction and update steps to suppress noise and align asynchronous data.
[0009] Preferably, the step of calculating the relationship potential energy indicator includes: determining the value of the comprehensive confidence evolution curve at the current time as a position quantity representing the relationship static strength; performing first-order derivative calculation on the comprehensive confidence evolution curve and determining the calculation result as a velocity quantity representing the relationship change rate; performing second-order derivative calculation on the comprehensive confidence evolution curve and determining the calculation result as an acceleration quantity representing the relationship change trend acceleration; weighting and combining the position quantity, velocity quantity, and acceleration quantity according to preset weight parameters to generate the relationship potential energy indicator.
[0010] Preferably, the preset weight parameters are adaptively configured according to the type of target industrial chain or the result of historical data backtesting to optimize prediction accuracy.
[0011] Preferably, the first set of time series data includes at least one of talent flow data and enterprise technology patent data; and the second set of time series data includes at least one of enterprise geographic spatial distribution data and business change data.
[0012] Preferably, the step of analyzing the industrial chain includes: outputting the relationship potential energy indicator and the position quantity, velocity quantity, and acceleration quantity as composite attributes of edges in the industrial chain network; identifying emerging connection opportunities or key links at risk of decline in the industrial chain according to the value of the relationship potential energy indicator or the positive or negative characteristics of the velocity quantity.
[0013] An industrial chain strong chain extension chain supplement chain analysis system includes a configuration to perform the above-mentioned industrial chain strong chain extension chain supplement chain analysis method.
[0014] A computer device includes a processor and a memory, the memory stores a computer program, and the computer program is executed by the processor to implement the above-mentioned industrial chain strong chain extension chain supplement chain analysis method.
[0015] A computer readable storage medium, having stored thereon a computer program, the computer program being executed by a processor to implement the industry chain strong chain extension chain complementary chain analysis method.
[0016] The technical scheme provided in the application has at least the following technical effects or advantages: The application fuses two groups of time sequence data of logical association and physical association, and uses a time sequence fusion model to correct asynchrony and noise between data, so that an evolution curve reflecting comprehensive association strength between enterprises is obtained. A relationship potential energy index constructed based on the curve quantitatively calculates relationship strength (a position amount), a change rate (a speed amount) and a change trend (an acceleration amount), so that the evaluation of the relationship between enterprises is improved from a static description to a quantitative analysis of a dynamic trend. The quantitative index can identify potential connection opportunities and recession risks, and provides an objective basis for the decision of a strong chain, an extended chain and a complementary chain based on an evolution trend of data. BRIEF DESCRIPTION OF DRAWINGS
[0017] Figure 1 The industry chain strong chain extension chain complementary chain analysis method flowchart of the application is shown in the figure. Figure 2 The industry chain strong chain extension chain complementary chain analysis system logic architecture diagram is shown in the figure. DETAILED DESCRIPTION
[0018] The above technical scheme will be described in detail below by combining the drawings in the specification and specific embodiments, so that the above technical scheme can be better understood. Obviously, the described embodiments are only part of the embodiments of the application, not all embodiments of the application, and it should be understood that the application is not limited to the example embodiments for explaining the application. Based on the embodiments of the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application. In addition, it should be noted that, for convenience of description, only parts related to the application are shown in the drawings, not all.
[0019] Referring to Figure 1 An industry chain strong chain extension chain complementary chain analysis method, the method comprising the following steps: Obtaining a first group of time sequence data reflecting logical association between enterprises in an industry chain, and a second group of time sequence data reflecting physical association between the enterprises; Based on the first group of time sequence data, a first confidence time sequence representing the logical association strength between the enterprises is calculated; Based on the second group of time sequence data, a second confidence time sequence representing the physical association strength between the enterprises is calculated; The first confidence time sequence and the second confidence time sequence are fused to generate a comprehensive confidence evolution curve. Based on the comprehensive confidence evolution curve, a relationship potential index representing the evolution trend of the relationship between the enterprises is calculated; According to the relationship potential index, the industry chain is analyzed as a strong chain, an extended chain or a complementary chain.
[0020] The present application dynamically infers potential opportunities and risks in the industry chain network through deep analysis of multi-source heterogeneous data, and associates them with the strategic goals of "strong chain, extended chain and complementary chain", thereby providing data-driven decision support for precise governance of the industry chain.
[0021] Referring to Figure 2 To achieve the above-mentioned goal, the industry chain strong chain extended chain complementary chain analysis system of the present application can be divided into three mutually cooperative levels: data acquisition layer, data processing and storage layer, and analysis application layer.
[0022] The data acquisition layer is responsible for obtaining raw data related to industry chain analysis from multiple external data sources.
[0023] The data processing and storage layer includes an industry chain dynamic network inference engine (hereinafter referred to as "inference engine") responsible for converting the raw data into structured knowledge containing dynamic trends for supporting "strong chain, extended chain, complementary chain" analysis.
[0024] The analysis application layer is responsible for visualizing the structured knowledge and finally outputting the candidate target list of "strong chain, extended chain, complementary chain".
[0025] Further, to support the analysis algorithm of the inference engine and ensure that the analysis results can accurately map to the real scene of "strong chain, extended chain, complementary chain", the system pre-constructs and continuously maintains a basic database: a time knowledge graph database in the data processing and storage layer.
[0026] The database is used to store various entities, relationships and their time evolution attributes in the industry chain in a structured manner. The core data model of the database includes: Entity: defines the core entity types in industry chain analysis, mainly including enterprises, technologies, products and talents. Each entity has a globally unique identifier and a set of fields describing its static attributes.
[0027] Relationship: defines the interaction types between entities, mainly including supply, cooperation, competition, possession and employment.
[0028] Timestamp: each relationship instance must be associated with one or more timestamps conforming to international standard format to record the time of its occurrence or duration.
[0029] In one embodiment, the system described in the present application can be deployed and run in a general computer hardware and software environment. On the hardware side, a server cluster with master-slave architecture can be adopted, and a data lake and various types of databases can be configured. On the software side, development can be based on the Linux operating system, using languages such as Python / Java, and related big data and machine learning frameworks.
[0030] The internal structure and specific workflow of the industry chain dynamic network inference engine will be described in detail below.
[0031] Data preparation and dual-channel time series confidence calculation: the inference engine performs a complete relationship inference, starting with data preparation and dual-channel parallel calculation.
[0032] After receiving the raw data transmitted from the data acquisition layer, the inference engine first performs data preparation operations. According to the pre-set entity and relationship definitions in the temporal knowledge graph, the raw data is cleaned, entity recognition and relationship extraction are performed, and all extracted information is converted into event sequences with standardized timestamps. Subsequently, the inference engine assigns all event sequences to two independent parallel processing channels according to the nature of data attributes: logical association data channel and physical association data channel.
[0033] Next, the inference engine performs continuous aggregation calculation on the data of the two channels on the time axis with a preset sliding time window (for example, the window size is set to 90 days, and the sliding step is set to 30 days).
[0034] For the logical association data channel, it aims to assess the strength of association between enterprises based on knowledge, talent, and other non-physical factors. Within each time window, the inference engine iterates through all discrete events in the window, such as a talent flow or a patent transfer. To quantify these events, the inference engine calls a confidence mapping function. In a specific embodiment, the function is implemented as a multi-dimensional lookup table. The input dimensions of the lookup table are {event type, importance level, source credibility}, and the output is a pre-labeled basic confidence score by domain experts. After obtaining the basic scores of all events in the window, the inference engine uses an exponential decay weighted average algorithm for aggregation. The internal working mechanism of this algorithm is to assign a weight to each event in the window; the weight is set as a function that decreases exponentially with the increase of the time interval from the event to the end point of the current time window. Finally, the inference engine calculates the sum of the products of all event basic scores and their respective weights to obtain the instantaneous logical confidence at that moment. As the time window steps forward on the entire time axis, the inference engine generates a first confidence time series for each pair of logically associated enterprises, consisting of the instantaneous confidence values at all times.
[0035] Meanwhile, the inference engine performs parallel processing on the physical association data channel, aiming to evaluate the strength of the association between enterprises in the physical world. For each time window, the inference engine first generates a snapshot of the physical space state at the end of the window based on all relevant data at that time. Then, the inference engine starts a co-location pattern mining algorithm on the snapshot. The internal working mechanism of the algorithm is to calculate an engagement index for any two enterprises. In a specific embodiment, the calculation logic of the index is as follows: first, count the actual observation frequency of the co-occurrence of the two types of enterprises in a specific spatial neighborhood; second, calculate the expected co-occurrence frequency based on the assumption of complete spatial randomness; finally, divide the difference between the actual observation frequency and the expected co-occurrence frequency by the expected co-occurrence frequency to obtain a standardized engagement index that reflects the significance of spatial association. This index is taken as the instantaneous physical association confidence of the time window. Similarly, by continuously generating state snapshots and repeatedly performing spatial analysis on the time axis, the inference engine eventually generates a second confidence time series composed of instantaneous confidence values for each pair of enterprises with physical association.
[0036] After the generation of the time series of the two channels, the inference engine also performs a global minimum-maximum value normalization to linearly map the values of all confidence time series to the unified interval of zero to one. At this point, the inference engine has completed the conversion of the original data into two standardized time series with unified structure, time synchronization, and direct use for the next step of fusion.
[0037] Time series fusion and comprehensive evolution curve generation: After calculating the first and second confidence time series, the inference engine starts the high-precision fusion process. This process aims to solve the technical problems of time delay, data asynchrony, and random noise caused by different data sources of the two time series.
[0038] To achieve this goal, the inference engine internally implements a dedicated time series fusion model. In a specific embodiment, the model is implemented through the technical path based on the Kalman filter algorithm.
[0039] First, the inference engine establishes a linear state space model to describe the evolution of the true relationship strength. In this model, the state variable of the system is defined as the current value of the relationship strength. The system dynamic model is set as a random walk model, which assumes that the true relationship strength at the current time is equal to the true relationship strength at the previous time plus a process noise, and the covariance of the process noise is a configurable system parameter. The observation variable of the system is the first confidence time series value and the second confidence time series value obtained at that time.
[0040] Subsequently, the inference engine iteratively performs filtering calculation in time steps. At each time step, it first performs a prediction step. The inference engine predicts the prior value of the real relationship strength at the current time step based on the state optimal estimate at the last time step.
[0041] Next, the inference engine performs an update step. It obtains two actual observations at the current time step. To deal with the problem of data asynchrony, if there is no new observation for a channel at the current time step, the inference engine can use the observation at the last time step, but will increase the observation noise covariance R of the channel accordingly to reduce the weight of the stale information in this update. In a specific embodiment, the increase can be implemented by multiplying the basic observation noise covariance of the channel by a penalty coefficient proportional to the time step length of the data delay. For example, if the data has been delayed for one time step, the corresponding covariance value can be multiplied by a coefficient of 2; if delayed for two time steps, multiplied by a coefficient of 4. In this way, the more stale the data, the lower its influence in the fusion calculation. After obtaining valid observations, the inference engine calculates the difference between the observations and the predicted value. Then, it calculates the Kalman gain based on the covariance of the system prediction and the covariance of the two observations. Finally, the inference engine uses the Kalman gain and the observation difference to correct the prior value obtained in the prediction step to obtain the optimal estimate of the real relationship strength at the current time step.
[0042] Through this "prediction-update" iterative loop, the inference engine can adaptively give higher weight to the observation channel with smaller noise (i.e., smaller observation noise covariance) and effectively handle asynchronous data.
[0043] Finally, after filtering calculation over the entire time span, the time series composed of the state optimal estimate values at all times, i.e., the comprehensive confidence evolution curve, is obtained. This curve is smoother in shape than the original two input curves and integrates the effective information of the two channels, and is the data basis for the next step of relationship potential calculation.
[0044] The calculation link of the relationship potential index: after generating the comprehensive confidence evolution curve, the inference engine starts the calculation process of the relationship potential index.
[0045] First, to ensure the stability and reliability of subsequent numerical calculations, the inference engine will first call a data smoothing preprocessing module for the input comprehensive confidence evolution curve. In a specific embodiment, the module can use a Savitzky-Golay filter. The filter performs least squares fitting of data with a high-order polynomial within a sliding window to filter out high-frequency noise while retaining its true trend information.
[0046] After smoothing, the inference engine starts computing three metrics. The first metric is the position quantity, which is the value of the smoothed curve at the current analysis time point. It is used to quantitatively represent the static strength of the relationship between the two enterprises at the current time point.
[0047] The second and third metrics are obtained through numerical differentiation. In a specific embodiment, the numerical differentiation can use the central difference method to obtain higher calculation accuracy. The inference engine obtains the first-order numerical differentiation by calculating the smoothed curve at the current time point and its adjacent data points before and after it, and determines it as the velocity quantity, which is used to represent the current change trend of the relationship strength (a positive value indicates strengthening, and a negative value indicates weakening) and the speed. Then, the inference engine continues to perform second-order numerical differentiation on the curve to calculate the acceleration at the current time point, and determines it as the acceleration quantity, which is used to represent the momentum of the change trend of the relationship strength (i.e., whether the momentum of strengthening or weakening is accelerating or slowing down).
[0048] After obtaining the position quantity, the velocity quantity, and the acceleration quantity, the inference engine calculates the final relationship potential index through a pre-set weighted combination model. The implementation of the model is as follows: multiply the above three metrics by their respective configurable weight coefficients, and then sum the three products to generate a comprehensive index that contains both the current situation of the relationship and its future evolution trend.
[0049] To improve the adaptability of the model, the inference engine also supports adaptive configuration of the weight parameters. In an optional embodiment, the configuration can be realized through historical data backtracking test. The inference engine automatically searches and determines a set of optimal weight combinations through cross-validation or Bayesian optimization algorithms with a pre-set optimization goal (e.g., maximizing the prediction accuracy of future key cooperation events). In another embodiment, the system can call different weight configuration schemes from a pre-set weight library according to the industry chain type input by the analyst.
[0050] Industry chain analysis and decision support link: After calculating the relationship potential index, the inference engine enters the final link of converting technical insights into decision support information.
[0051] The inference engine first performs the construction of the composite edge attribute. That is, a structured "relationship potential" object is constructed. The object is defined as a data structure containing four key fields, which store the position quantity, the velocity quantity, the acceleration quantity, and the comprehensive relationship potential index of the relationship at the current time point. The inference engine takes this "relationship potential" object as the core attribute of the edge connecting the two enterprise nodes in the graph.
[0052] Based on the dynamic network graph containing the composite edge attributes, the inference engine can perform a series of automated analysis applications. For example, to realize the identification of emerging connection opportunities, the inference engine can filter out all the edges with positive speed and positive acceleration according to the rules. To realize the early warning of the recession risk link, the inference engine can highlight all the edges with high position value and negative speed according to the rules.
[0053] Finally, the inference engine links the above-identified "opportunities" and "risks" with the targets of "strong chain, extension chain and complementary chain". The inference engine automatically classifies and outputs all the identified "emerging connection opportunities" as the candidate target list of "complementary chain" or "extension chain". At the same time, the inference engine automatically classifies and outputs all the identified "recession risk links" as the list of key attention objects of "strong chain" work.
[0054] Through the whole set of automated processes, the inference engine finally converts its technological innovation into data-driven decision support for the macro strategy of "industrial chain strong chain, extension chain and complementary chain".
[0055] The preferred embodiments of the present application disclosed above are only used to help explain the present application. The preferred embodiments do not describe all the details, and should not be understood as limiting the present application to these specific embodiments. Obviously, many modifications and variations can be made according to the content of the present application. The present application selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present application, so that those skilled in the art can well understand and utilize the present application. The present application is limited only by the claims and their full scope and equivalents.
Claims
1. A method for analyzing the strengthening, extension, and supplementation of an industrial chain, characterized in that, Includes the following steps: Acquire a first set of time-series data reflecting the logical relationships between enterprises in the industry chain, and a second set of time-series data reflecting the physical relationships between the enterprises; Based on the first set of time series data, a first confidence time series characterizing the strength of the logical association between the enterprises is calculated; Based on the second set of time series data, a second confidence time series characterizing the strength of the physical association between the enterprises is calculated; The first confidence time series and the second confidence time series are fused together to generate a comprehensive confidence evolution curve; Based on the comprehensive confidence evolution curve, a relationship potential index characterizing the evolution trend of the inter-firm relationship is calculated; Based on the aforementioned relational potential energy index, the industrial chain is analyzed to strengthen, extend, or supplement the chain.
2. The method for analyzing the strengthening, extension, and supplementation of the industrial chain as described in claim 1, characterized in that, The fusion processing step is implemented through a time-series fusion model configured to handle the time delay and data asynchrony between the first and second confidence time series.
3. The method for analyzing the strengthening, extension, and supplementation of the industrial chain as described in claim 2, characterized in that, The time series fusion model employs the Kalman filter algorithm or a variant thereof, and iteratively calibrates the first and second confidence time series through prediction and update steps to suppress noise and align asynchronous data.
4. The method for analyzing the strengthening, extension, and supplementation of the industrial chain as described in claim 1, characterized in that, The steps for calculating the relational potential energy index include: The value of the comprehensive confidence evolution curve at the current moment is determined as the positional quantity characterizing the static strength of the relationship; The first derivative of the comprehensive confidence evolution curve is calculated, and the calculation result is determined as the velocity quantity characterizing the rate of change of the relationship. The second derivative of the comprehensive confidence evolution curve is calculated, and the calculation result is determined as an acceleration measure characterizing the acceleration of the relationship change trend. The position, velocity, and acceleration measures are weighted and combined according to preset weight parameters to generate the relational potential energy index.
5. The method for analyzing the strengthening, extension, and supplementation of the industrial chain as described in claim 4, characterized in that, The preset weight parameters are adaptively configured based on the type of the target industry chain or the results of backtesting based on historical data to optimize prediction accuracy.
6. The method for analyzing the strengthening, extension, and supplementation of the industrial chain as described in claim 1, characterized in that, The first set of time-series data includes at least one of talent mobility data and enterprise technology patent data; the second set of time-series data includes at least one of enterprise geospatial distribution data and business registration change data.
7. The method for analyzing the strengthening, extension, and supplementation of the industrial chain as described in claim 1, characterized in that, The steps for analyzing the aforementioned industry chain include: The relationship potential energy index, along with the position quantity, velocity quantity, and acceleration quantity, are output as a composite attribute of the edge in the industrial chain network. Based on the numerical value of the relational potential energy index or the positive or negative characteristics of the velocity quantity, emerging connection opportunities or key links with the risk of decline in the industrial chain can be identified.
8. A supply chain strengthening, extension, and supplementation analysis system, comprising a supply chain strengthening, extension, and supplementation analysis method configured to perform any one of claims 1 to 7.
9. A computer device, comprising a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, it implements the industrial chain strengthening, extension, and supplementation analysis method as described in any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the industrial chain strengthening, extension, and supplementation analysis method as described in any one of claims 1 to 7.