Enterprise digital capability evaluation method and system based on multi-source data fusion

Abstract

The present application relates to the field of digital capability evaluation, and more particularly to a kind of enterprise digital capability evaluation method and system based on multi-source data fusion.The method comprises the following steps: obtaining the multi-source heterogeneous data of target enterprise, carrying out digital behavior track analysis, and constructing digital behavior matrix;Business topology dependency modeling is carried out based on digital behavior matrix, and inter-business dependency network is constructed;Collect the multi-dimensional operating results of enterprise, carry out digital value correlation analysis on inter-business dependency network, and construct value conversion efficiency curve;According to the value conversion efficiency curve, carry out disturbance injection simulation, extract digital value evolution track;Based on digital value evolution track, the elasticity response index of different business fields of enterprise is calculated, and comprehensive digital capability evaluation is carried out, and the capability evaluation report is output.The present application can quickly and accurately evaluate the digital capability of enterprise, and improve the overall value evaluation efficiency of vertical capability of enterprise.

Description

Technical Field

[0001] This invention relates to the field of digital capability assessment, and in particular to a method and system for assessing enterprise digital capabilities based on multi-source data fusion. Background Technology

[0002] With the rapid development of information technology and data analytics, enterprise digital transformation is deepening, and digital capabilities have become a crucial indicator for measuring core competitiveness. Especially driven by big data, cloud computing, the Internet of Things, and artificial intelligence, the massive amounts of heterogeneous data generated by enterprises in production, supply chain, R&D, and customer service provide rich information for digital capability assessment. However, enterprise business activities are highly complex and involve frequent cross-system collaborations. Traditional methods relying on single data sources or manual experience often fail to comprehensively and accurately quantify an enterprise's digital capabilities. Existing digital capability assessment methods primarily depend on questionnaires, key indicator statistics, or single business process analysis. While these methods can provide some quantitative references, they often overlook the correlations between multi-source heterogeneous data, the global dependency structure of business activities, and the transmission patterns of digital value across different business stages. The assessment results lack precision and operability, failing to truly reflect an enterprise's digital resilience and potential capabilities in the face of market changes and business disruptions. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention proposes a method and system for assessing enterprise digital capabilities based on multi-source data fusion, thereby resolving at least one of the aforementioned technical issues.

[0004] To achieve the above objectives, this invention provides a method for assessing enterprise digital capabilities based on multi-source data fusion, comprising the following steps: Step S1: Obtain multi-source heterogeneous data of the target enterprise, analyze digital behavior trajectories, and construct a digital behavior matrix; Step S2: Model business topology dependencies based on the digital behavior matrix and construct an inter-business dependency network; Step S3: Collect multi-dimensional business results of enterprises, conduct digital value correlation analysis on the inter-business dependency network, and construct a value conversion efficiency curve; Step S4: Perform perturbation injection simulation based on the value conversion efficiency curve to extract the digital value evolution trajectory; Step S5: Calculate the resilience response indicators for different business areas of the enterprise based on the digital value evolution trajectory, conduct a comprehensive digital capability assessment, and output a capability assessment report.

[0005] In this invention, step S1 specifically includes the following steps: Acquire multi-source heterogeneous data from the target company; Cross-modal semantic alignment and timestamp synchronization are performed on the multi-source heterogeneous data to obtain a multi-source fusion dataset; Enterprise digital behavior trajectory analysis is performed based on multi-source fusion dataset to obtain behavioral feature tensors; A digital behavior matrix is ​​constructed by performing a low-dimensional embedding mapping on the behavioral feature tensor.

[0006] In this invention, step S2 specifically involves the following steps: Enterprise business activities are identified based on digital behavior matrices, and each business activity node is marked. Calculate the data dependency strength and redundant path ratio between each business activity node to obtain the relationship between nodes; Based on the relationship between nodes, a graph structure analysis [3.1] is performed to obtain circular dependency and abnormal coupling paths; Business topology dependency modeling is performed based on circular dependencies and abnormal coupling paths to construct a dependency network between businesses.

[0007] In this invention, step S3 specifically involves the following steps: Collect multi-dimensional business results from enterprises, including order conversion rate, inventory turnover cycle, R&D delivery cycle, and customer response time. Based on the inter-business dependency network, causal relationship modeling is performed on the multi-dimensional business results of the enterprise to generate business transmission paths; Calculate the time delay attenuation coefficient and information loss rate of the business transmission path to obtain the data value conversion function; Based on the data value transformation function, we can model the data value of enterprises in multiple fields and construct a value transformation efficiency curve.

[0008] In this invention, step S4 specifically involves the following steps: Define market disturbance factors; The value conversion efficiency curve is perturbed based on the aforementioned market disturbance factors to generate a virtual shock sequence; Monte Carlo simulation was performed on the virtual impact sequence to obtain the simulation results; The simulation results are decomposed into future time windows to extract the evolution trajectory of digital value.

[0009] In this invention, step S5 specifically involves the following steps: The elasticity response indicators for different business areas of an enterprise are calculated based on the digital value evolution trajectory. These elasticity response indicators include efficiency recovery speed, resource dynamic reallocation cost, and path reconstruction complexity. Based on the elastic response index, multidimensional response coupling is performed to construct the elastic response surface; Digital value zoning and comprehensive digital capability assessment are performed based on elastic response surfaces, and a capability assessment report is output.

[0010] In this invention, the specific steps for performing digital value partitioning and comprehensive digital capability assessment based on elastic response surfaces, and outputting a capability assessment report, are as follows: Digital value partitioning is performed based on the elastic response surface to obtain digital value partitions; the digital value partitions include stable operating ranges, performance transition boundaries, and capability failure critical regions. A digital capability resilience assessment is conducted on digital value zones to obtain a resilience assessment index. Based on the elastic response surface, reverse reasoning is performed to identify potential capability zones with high elasticity but low efficiency. Calculate the asset utilization rate and algorithm invocation potential of the potential capability zone to obtain implicit digital capabilities; A comprehensive digital capability assessment is conducted based on the resilience assessment index and implicit digital capabilities, and a capability assessment report is output.

[0011] This invention provides an enterprise digital capability assessment system based on multi-source data fusion, used to execute the enterprise digital capability assessment method based on multi-source data fusion as described above, including: The data acquisition module is used to acquire multi-source heterogeneous data from the target enterprise, analyze digital behavior trajectories, and construct a digital behavior matrix. The topology dependency module is used to model business topology dependencies based on a digital behavior matrix and build a dependency network between businesses. The value conversion module is used to collect multi-dimensional business results, perform digital value correlation analysis on the inter-business dependency network, and construct a value conversion efficiency curve. The perturbation simulation module is used to perform perturbation injection simulation based on the value conversion efficiency curve and extract the digital value evolution trajectory. The assessment module is used to calculate the resilience response indicators of different business areas of an enterprise based on the evolution trajectory of digital value, and to conduct a comprehensive digital capability assessment, outputting a capability assessment report.

[0012] The specific beneficial effects of this invention are as follows: By acquiring multi-source heterogeneous data from enterprises and analyzing digital behavior trajectories, a digital behavior matrix can be constructed. This allows for the quantification of complex, cross-system business activities into unified feature representations, comprehensively capturing operational patterns in enterprise production, supply chain, R&D, and customer service, improving data integration and information usability. Based on the digital behavior matrix, a business topology dependency network can be constructed, clearly depicting the dependencies and potential risk nodes between business activities, identifying key business nodes and weak links, and providing structured support for analyzing the overall structure of business processes, optimizing resource allocation, and modeling subsequent value transformation. By collecting multi-dimensional operational results and conducting digital value correlation analysis, a value transformation efficiency curve can be constructed, quantifying the contribution of each business activity to operational indicators, revealing the efficiency relationship between data input and business value, and helping to identify high-value and inefficient business links. Based on the value transformation efficiency curve, perturbation injection simulation can simulate market, supply chain, and external environmental shocks, extracting the digital value evolution trajectory, revealing the value fluctuations and recovery trends of enterprises under different perturbation conditions, and improving the quantitative understanding of business resilience, risk propagation, and value stability. By calculating resilience response indicators based on the evolution trajectory of digital value and conducting a comprehensive evaluation, the recovery speed, resource allocation capabilities, and path reconstruction complexity of different business areas of an enterprise can be quantified, potential digital capabilities and optimization space can be identified, and a quantifiable digital capability assessment report can be generated. Attached Figure Description

[0013] Figure 1 This is a flowchart illustrating the steps of an enterprise digital capability assessment method based on multi-source data fusion according to the present invention. Figure 2 This is a detailed flowchart illustrating the implementation steps of step S1. Figure 3 This is a detailed flowchart illustrating the implementation steps of step S2; Figure 4 This is a schematic diagram of the value conversion efficiency curve. Detailed Implementation

[0014] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0015] This application provides a method and system for assessing enterprise digital capabilities based on multi-source data fusion. The implementing entities of the method and system include, but are not limited to, the following: mechanical equipment, data processing platforms, cloud server nodes, network upload devices, etc., which can be considered general computing nodes of this application. The data processing platform includes, but is not limited to, at least one of the following: an audio-visual management system, an information management system, and a cloud-based data management system.

[0016] Please see Figures 1 to 4This invention provides a method for assessing enterprise digital capabilities based on multi-source data fusion, comprising the following steps: Step S1: Obtain multi-source heterogeneous data of the target enterprise, analyze digital behavior trajectories, and construct a digital behavior matrix; Step S2: Model business topology dependencies based on the digital behavior matrix and construct an inter-business dependency network; Step S3: Collect multi-dimensional business results of enterprises, conduct digital value correlation analysis on the inter-business dependency network, and construct a value conversion efficiency curve; Step S4: Perform perturbation injection simulation based on the value conversion efficiency curve to extract the digital value evolution trajectory; Step S5: Calculate the resilience response indicators for different business areas of the enterprise based on the digital value evolution trajectory, conduct a comprehensive digital capability assessment, and output a capability assessment report.

[0017] In the embodiments of the present invention, see Figure 1 This is a flowchart illustrating the steps of an enterprise digital capability assessment method based on multi-source data fusion according to the present invention. In this example, the steps of the enterprise digital capability assessment method based on multi-source data fusion include: Step S1: Obtain multi-source heterogeneous data of the target enterprise, analyze digital behavior trajectories, and construct a digital behavior matrix; In this embodiment, multi-source heterogeneous data from the target enterprise is acquired, including production execution system operation logs, supply chain collaboration platform interaction records, industrial IoT device sensing data, enterprise application software call chain tracing data, and interaction data from external e-commerce and government platforms. For different data types, appropriate feature extraction methods are employed: text-based operation logs are encoded into vector representations using natural language processing techniques, and semantic features are generated using a pre-trained language model; time-series sensor data are extracted using a sliding window statistical method to extract mean, variance, frequency domain features, and trend features; and call chain data are transformed into vector representations using graph embedding methods to convert structural information. The extracted multi-source features are processed synchronously in the time dimension, with data resampling and interpolation performed at a unified time granularity, including 5 seconds or 1 minute, and missing data is imputed using KNN or mean imputation. Subsequently, feature vectors from different modalities are semantically aligned across modalities, and each modal feature is uniformly mapped to the same semantic space using a fully connected mapping or alignment loss function, forming a multi-source fusion dataset. The fused behavioral data is divided according to business processes or time series, and behavioral feature tensors are constructed. These tensors are then compressed into digital behavioral matrices using low-dimensional embedding mapping methods, such as PCA or autoencoders. The rows of the matrix correspond to time periods or business units, and the columns correspond to the embedded feature dimensions. Step S2: Model business topology dependencies based on the digital behavior matrix and construct an inter-business dependency network; In this embodiment, the behavior matrix sample is divided into business activity nodes according to business processes, such as order processing, inventory management, R&D delivery, and customer response. An initial dependency strength matrix between nodes is obtained by calculating the event co-occurrence frequency and sequence dependency relationships between nodes, and combined with a redundant path ratio index, the flexibility and redundancy between nodes are quantified. Subsequently, the nodes and weighted edges are transformed into a directed graph, where nodes represent business activities and edge weights represent dependency strength and redundant path ratio. Structural analysis is performed on this graph, identifying cyclic dependencies using a strongly connected component algorithm, and anomaly detection methods are used to identify anomalously coupled paths, such as paths with dependency strengths significantly higher than the average.

[0018] Step S3: Collect multi-dimensional business results of enterprises, conduct digital value correlation analysis on the inter-business dependency network, and construct a value conversion efficiency curve; In this embodiment, based on the construction of a business dependency network, multi-dimensional business results of the enterprise are collected, including indicators such as order conversion rate, inventory turnover cycle, R&D delivery cycle, and customer response timeliness, and mapped to business activity nodes. Causal modeling methods, including structural equation modeling or Bayesian networks, are employed to analyze the impact strength of business activities on each indicator and generate business transmission paths. For each path, a time lag attenuation coefficient and information loss rate are calculated to quantify the response delay and information attenuation degree of business activities on business results, thereby establishing a data value conversion function. By weighted and accumulated conversion functions for each path, a value conversion efficiency curve is formed, with the horizontal axis representing the amount of business activity or data input and the vertical axis representing the improvement in corresponding business results, reflecting the efficiency of digital value output of the enterprise in different business areas and time nodes.

[0019] Step S4: Perform perturbation injection simulation based on the value conversion efficiency curve to extract the digital value evolution trajectory; In this embodiment, market disturbance factors are introduced based on the value conversion efficiency curve, including sudden increases and decreases in order volume, raw material supply disruptions, platform overload, and changes in the external environment. These disturbance factors are normalized and then injected into the curve to form a virtual impact sequence. The disturbance sequence is subjected to large-scale random sampling and iterative simulation using the Monte Carlo method to generate a large number of possible value conversion paths. During the simulation, the value conversion rate change of each trajectory in the future time window is recorded, and the digital value evolution trajectory is extracted using time series decomposition methods. The fluctuation trend and recovery process of value under different disturbance conditions are analyzed to obtain the dynamic response characteristics of the enterprise's digital value to internal and external disturbances.

[0020] The disturbance injection dimensions are specifically as follows: Order dimension: including sudden increases or decreases in order volume, changes in order type, and fluctuations in customer groups. Supply chain dimension: raw material supply disruptions, transportation delays, and fluctuations in supply costs. Platform / system dimension: platform overload, computing resource limitations, or service node failures. External environment dimension: policy changes, market demand fluctuations, competitor behavior, and macroeconomic or natural environmental shocks. Disturbances in each dimension can be injected independently or combined to form composite disturbances, facilitating the analysis of the value conversion system's sensitivity and resilience to multi-factor impacts.

[0021] The specific timing rules for disturbances are as follows: Time-triggered: Disturbances are injected at preset time points, such as quarterly, monthly, or specific event windows. Random-triggered: Disturbance time points are generated through probability distributions to simulate irregular shocks. Continuous / periodic: The disturbance continues to act within a certain time window or exhibits a periodic fluctuation pattern, such as the periodic fluctuations in raw material supply. Progressive intensity: The intensity of the disturbance can accumulate or decay over time, used to simulate the response process to long-term market pressure or rapid shocks.

[0022] Step S5: Calculate the resilience response indicators for different business areas of the enterprise based on the digital value evolution trajectory, conduct a comprehensive digital capability assessment, and output a capability assessment report.

[0023] In this embodiment, elasticity response indicators for different business areas of an enterprise are calculated based on the digital value evolution trajectory, including efficiency recovery speed, resource dynamic reconfiguration cost, and path reconstruction complexity. Efficiency recovery speed is quantified by the period required for the trajectory to return to the initial value level; resource dynamic reconfiguration cost is quantified by analyzing the resource adjustment consumption of business activity nodes in the trajectory; path reconstruction complexity is measured by the number of path adjustments and loop reconstructions of disturbed nodes in the business topology-dependent network. The elasticity indicators for each business area are weighted and combined to generate an elasticity response surface, reflecting the enterprise's comprehensive elasticity performance under different disturbance scenarios. Through surface analysis and partitioning, business areas are divided into high-elasticity, high-efficiency, low-efficiency, and potential capability zones, and implicit digital capabilities are calculated by combining asset utilization and algorithm call potential. The elasticity assessment index and implicit capability score are weighted and summarized to form a comprehensive digital capability assessment, and a capability assessment report is output, including elasticity partitioning for each business area, value evolution trajectory, potential capability analysis, and the enterprise's digital capability score.

[0024] In this embodiment, see Figure 2 The diagram below illustrates the detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include: Acquire multi-source heterogeneous data from the target company; Cross-modal semantic alignment and timestamp synchronization are performed on the multi-source heterogeneous data to obtain a multi-source fusion dataset; Enterprise digital behavior trajectory analysis is performed based on multi-source fusion dataset to obtain behavioral feature tensors; A digital behavior matrix is ​​constructed by performing a low-dimensional embedding mapping on the behavioral feature tensor.

[0025] In this embodiment, a multi-source data acquisition system is constructed for the target enterprise. Through interface integration and data access modules, unified data is collected from the production execution system (MES), including operation logs, supply chain collaboration platform interaction records, industrial IoT (IIoT), including equipment sensing data, enterprise application software call chain tracing data, and interaction data from external e-commerce and government platforms. During implementation, data access can be achieved using message queues and data acquisition agents. The MES and supply chain system retrieve log data via API interfaces at minute-level intervals (e.g., 1-minute sampling periods). IIoT devices upload sensor data (including temperature, pressure, and vibration) via the MQTT protocol at second-level intervals (e.g., 1-5 seconds). Application call chain data is collected through a distributed tracing system, including call chain information. External platform data is obtained through HTTP interfaces or data subscription mechanisms. To ensure data integrity and consistency, a data verification mechanism is introduced during the acquisition phase, including CRC checks and packet loss retransmission strategies. A unified data access standard is set for different data sources, such as JSON or Parquet format. A data tagging system is used to identify data from different sources, including source_id and system_type, and the acquisition timestamp is recorded.

[0026] Different feature encoding methods are employed for different data types, including log text, time-series sensor data, and call chain graph structure data: For text-based data, such as operation logs and interaction records, semantic vectors can be extracted based on pre-trained language models, such as BERT-like models, with 768 dimensions; for time-series sensor data, a sliding window is used, such as a window length of 60 seconds and a step size of 10 seconds, for statistical feature extraction, including mean, variance, and frequency domain features; for call chain data, graph embedding methods, such as Node2Vec, are used to transform the link structure into a vector representation, with 128 dimensions. Feature mapping functions, such as fully connected neural networks or alignment loss functions, are used to uniformly map different modal features to the same semantic space, such as a 256-dimensional unified vector space. For time alignment, a timestamp-based synchronization strategy is adopted, using a fixed time granularity, such as 5 seconds or 10 seconds, as a baseline, to resample and interpolate data at different frequencies, including linear interpolation or spline interpolation, and missing data is repaired using mean imputation or KNN completion methods.

[0027] Data is segmented according to business processes or time series, such as by order cycle, production batch, or fixed time window, to construct behavioral sequence samples. For each data segment, behavioral events are defined, including equipment start / stop, order flow, and system call anomalies. Key behavioral nodes are identified using a rule engine or machine learning-based event detection model, such as a sequence model based on LSTM or Transformer. In the specific implementation, the sequence length can be set to 100-200 steps, and a sliding window is used to generate training samples. Behavioral pattern features are extracted through supervised or self-supervised learning. Multidimensional features are introduced to construct a behavioral feature tensor, for example, generating a tensor under a three-dimensional structure of "time × behavior type × data source," such as a dimension of T × N × M, where T≈120, N≈50, and M≈5. In addition, anomaly detection algorithms can be combined, such as IsolationForest or an autoencoder based on reconstruction error, to identify abnormal behavioral trajectories, and the degree of anomaly is added as an additional feature to the tensor, including normalization to the 0-1 range.

[0028] For the constructed high-dimensional behavioral feature tensor, a low-dimensional embedding mapping method is used to achieve feature compression and structural representation. The tensor is standardized, including Z-score standardization, to eliminate differences in the dimensions of different features; then, an appropriate dimensionality reduction method is selected for embedding mapping. In the specific implementation, linear and nonlinear methods can be combined: for example, principal component analysis, including PCA, is used to reduce the dimension from high dimensions (e.g., 6000 dimensions) to lower dimensions (e.g., 128 dimensions), retaining more than 95% of the variance information; then, nonlinear dimensionality reduction methods, including t-scores, SNE, or UMAP, are used to further compress it to 32 or 64 dimensions to enhance the clustering structure representation capability of the features. In deep learning scenarios, autoencoders, including Autoencoders, can also be used for end-to-end training, with the encoding layer set to a layer-by-layer compression structure (e.g., 1024→256→64). During training, the mean squared error loss function is used, the learning rate is set to 0.001, and the number of training epochs is 50–100. The low-dimensional embedding results are arranged according to the sample dimensions to construct a digital behavior matrix, including, for example, the number of samples K × the feature dimension D, typically 1000×64. This matrix can compactly express the digital behavior characteristics of enterprises.

[0029] In this embodiment, see Figure 3 The diagram below illustrates the detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include: Identify enterprise business activities based on digital behavior matrices and mark each business activity node; Calculate the data dependency strength and redundant path ratio between each business activity node to obtain the relationship between nodes; Graph structure analysis is performed based on the relationships between nodes to obtain circular dependencies and abnormal coupling paths; Business topology dependency modeling is performed based on circular dependencies and abnormal coupling paths to construct a dependency network between businesses.

[0030] In this embodiment, cluster analysis is performed on the matrix samples. Similarity-based clustering methods, such as K-means, spectral clustering, or density clustering, divide the behavioral samples into several classes, each corresponding to a potential business activity. The number of clusters can be determined based on the silhouette coefficient or elbow rule, generally covering the main business types of the enterprise. The clustering results are combined with behavioral event mapping rules to label each class of samples as a specific business activity node. For example, the behavioral sequence of order creation, approval, and production instruction issuance is labeled as order processing, and "equipment start / stop, fault handling, and maintenance records" are labeled as "equipment operation and maintenance." During the labeling process, time series pattern matching, including dynamic time warping, is used to determine the start and end times of business activities to ensure that nodes not only reflect business logic but also capture actual operational time characteristics. For rare or abnormal behavioral sequences, graph-based classification methods or rule engines are used to further complete node labeling, ensuring the comprehensiveness and accuracy of business activity identification and forming a clear set of business activity nodes.

[0031] After the business activity nodes are labeled, the data dependency strength and redundant path ratio between nodes are calculated to quantify the relationships between business activities. Data dependency strength is measured by analyzing the proportion of behavioral events from the source node to the target node relative to the total number of events at the source node, and weighted by cross-node interaction weights to form a dependency strength matrix. The redundant path ratio is calculated by statistically analyzing the ratio of the number of multiple feasible paths from the source node to the target node to the total number of possible paths, used to assess the flexibility and redundancy level of the business process. The number of paths is counted using graph traversal methods, including breadth-first or depth-first traversal, and the path length is normalized, typically considering paths with a length of 3 to 5 hops, forming a node relationship matrix.

[0032] Based on the node association matrix, the graph structure is analyzed to identify circular dependencies and anomalous coupling paths. Nodes and edges form a directed weighted graph, with edge weights representing dependency strength or redundancy percentage. Circular dependencies are identified using a strongly connected component algorithm, and the dependency weights of each node in the loop are recorded to assess the potential impact of the loop on the business process. Anomalous coupling paths are determined through threshold screening or statistical methods; for example, paths with significantly high or low dependency strength are marked as anomalous coupling, indicating potential risks. Indicators such as loop length, average dependency strength, and loop frequency are calculated to quantify the impact of circular dependencies on the overall business process.

[0033] After identifying circular dependencies and anomalous coupling paths, business topology dependency modeling is performed to construct an inter-business dependency network. Key cycles and anomalous paths in the directed weighted graph are prioritized, and network analysis methods are used to generate the business dependency network structure. Edges with dependency strength below a certain threshold are filtered as redundant or optimizable paths, while edges with high dependency strength are retained as critical dependencies. The topology network includes not only direct dependencies but also quantifies indirect influence relationships through propagation methods, achieving global visualization of business dependencies. Network structure metrics include node degree, loop density, connectivity, and critical node centrality. Key business nodes are analyzed through degree centrality and betweenness centrality.

[0034] In one specific embodiment, the enterprise's core business system includes five business activity nodes: order processing (A), inventory management (B), procurement management (C), shipment scheduling (D), and financial settlement (E). A digital behavior matrix M is constructed using system logs and business execution records, where each element... This represents the amount of data interaction between node i and node j, including, for example, the number of calls or the number of data entries.

[0035] Calculate the data dependency strength between nodes based on the digital behavior matrix. : ; Here, the denominator is the total output of node i, representing the dependency ratio. Then, the number of different paths from node i to node j is counted. Calculate the proportion of redundant paths : ; based on and A directed graph is constructed to represent business dependencies, and circular dependency paths and abnormal coupling paths are identified through graph analysis. To quantify the importance of nodes, node dependency centrality can be calculated. : ; This value reflects the criticality of a node within the overall business network. Ultimately, by combining circular dependencies, anomalous coupling, and node centrality, a dependency network between enterprise businesses is constructed, allowing visualization of critical dependencies, bottleneck nodes, and redundant paths.

[0036] In this embodiment, step S3 includes the following steps: Collect multi-dimensional business results from enterprises, including order conversion rate, inventory turnover cycle, R&D delivery cycle, and customer response time.

[0037] Based on the inter-business dependency network, causal relationship modeling is performed on the multi-dimensional business results of the enterprise to generate business transmission paths; Calculate the time delay attenuation coefficient and information loss rate of the business transmission path to obtain the data value conversion function; Based on the data value transformation function, we can model the data value of enterprises in multiple fields and construct a value transformation efficiency curve.

[0038] In this embodiment, the collected data typically includes order conversion rate, inventory turnover cycle, R&D delivery cycle, and customer response time. Order conversion rate measures the efficiency from potential customers to actual transactions. The calculation formula is the number of actual orders divided by the number of potential orders. In experiments, daily or weekly time granularities can be used for statistical analysis to capture short-term fluctuations and trend changes. Inventory turnover cycle reflects the efficiency of enterprise inventory management. It requires extracting inventory levels and inbound / outbound times from inventory logs and calculating the average turnover days. The data granularity is typically at the daily level. R&D delivery cycle focuses on the length of time from project initiation to delivery of a new product or project. It requires integrating data from project management systems and R&D task management platforms, accumulating the completion times of tasks at each stage. In experiments, time windows can be set to weeks or months to ensure sufficient stability of the sample data. Customer response time reflects the speed at which an enterprise responds to customer requests or issues. It can be obtained through customer relationship management systems, including CRM or service ticket systems, recording information such as request receipt time, initial response time, and processing completion time. During the data collection process, the data is cleaned, missing values ​​are filled, and anomalies are detected. Anomaly handling methods can include box plot outlier determination or the 3σ principle, and unreasonable data can be removed or interpolated.

[0039] A business dependency network is constructed, treating each operational indicator as a network node and the potential influence relationships between indicators as network edges. The initial edge establishment can rely on expert rules or historical data correlation analysis, such as screening candidate associations using Pearson or Spearman correlation coefficients, with a threshold set between 0.5 and 0.7 to ensure the reliability of potential causal relationships. Subsequently, causal discovery algorithms, such as Granger causality tests and structural equation modeling (including SEM or Bayesian networks), are used to quantitatively test the causal relationships of the candidate edges. Taking the Granger test as an example, assuming that order conversion rates may affect R&D delivery cycles, time series modeling, including ARIMA or VAR models, is used to test the predictive contribution of historical order changes to R&D delivery cycles, with a significance level typically set at 0.05. In experiments, sliding time windows, such as 12-week windows, are applied to multi-indicator data. The causal strength of each window is calculated and smoothed and averaged, ultimately forming a causal network of business transmission paths.

[0040] The causal network of the business transmission path, based on the constructed inter-business dependency network, models the causal relationships of multi-dimensional business results, generating business transmission paths from upstream business nodes to downstream result nodes. Within the generated business transmission paths, hierarchical divisions can be made based on path length, key node weights, and the scope of business impact. Paths are categorized into first-level paths that directly affect core business results, second-level paths that indirectly affect auxiliary processes, and third-level paths that affect low-frequency or peripheral businesses. The importance of each path is determined by calculating its cumulative dependency weight, node centrality, and data flow contribution. Paths that contribute the most to the overall enterprise value and have the highest information transmission efficiency are defined as core paths, while paths with limited overall value contribution, longer time delays, or higher information loss rates are defined as secondary paths. Based on the business transmission paths, the time delay attenuation coefficient and information loss rate of each path are calculated to obtain a data value transformation function. This function is then used to model the data value of the enterprise across multiple domains, generating value transformation efficiency curves.

[0041] Based on causal networks, it is necessary to quantitatively characterize the timeliness and information fidelity of the business transmission process. The time lag decay coefficient represents the degree of delay and attenuation in the transmission of a change in a business indicator to downstream indicators. For example, the impact of changes in order conversion rate on inventory turnover cycle may have a 2-3 week delay; the decay coefficient can be calculated using an exponential decay model. ,in The attenuation coefficient, obtained in experiments through least-squares fitting of historical sequence data, typically ranges from 0.1 to 0.5. The information loss rate measures the loss of service signals or the increase in noise during transmission. It can be calculated using mutual information reduction or the signal-to-noise ratio, including SNR changes. In experiments, the original time series is compared with the predicted series to calculate the information retention ratio for each service path. Combining the time-delay attenuation coefficient and the information loss rate forms a data value transformation function. The function value represents the effectiveness of converting upstream indicator data into downstream operational value. For example, for a company's order conversion rate, inventory turnover path, attenuation coefficient of 0.25, and information loss rate of 0.15, the data value conversion function value is approximately 0.85, indicating that the information transmission efficiency of this path is relatively high. This function can be calculated for all business paths, quantifying the value flow of multi-dimensional operational data within the company.

[0042] The value functions of different business paths are weighted and aggregated. Weights can be set based on business criticality or contribution, such as 0.4 for order conversion rate, 0.3 for R&D delivery cycle, 0.2 for inventory turnover cycle, and 0.1 for customer response time. This aggregated value function represents the overall data value of the enterprise, reflecting the value of various data across multiple business domains. A value conversion efficiency curve is then plotted, with data input volume (e.g., the number of data entries collected or data dimension coverage) on the horizontal axis and the output of the comprehensive data value function on the vertical axis. In experiments, the data input volume can be set from 10,000 to 1 million entries, with the vertical axis function value ranging from 0 to 1. The curve shape reflects the efficiency of data value utilization: initially, it may show rapid growth, then tend to plateau, indicating the diminishing marginal value effect of data. Analysis of the curve's slope, peak value, and saturation value allows for the evaluation of the enterprise's data value conversion efficiency and potential for improvement across multiple domains.

[0043] In one specific implementation, multi-dimensional business results are collected, including order conversion rate, inventory turnover cycle, R&D delivery cycle, and customer response time. A causal relationship model is then established based on the inter-business dependency network to generate business transmission paths. For each path, a time delay attenuation coefficient α is calculated to reflect the delayed impact of information flow from upstream to downstream. ; in The processing time of each node on the path is given; the information loss rate β is calculated to reflect the degree of information attenuation during transmission. ; in Let α represent the information loss ratio at node i; combining α and β, we obtain the data value transformation function V, which reflects the actual contribution of data along the path to the enterprise's results. ； By calculating and summarizing all paths, a multi-domain data value modeling result is generated for the enterprise, and a value conversion efficiency curve is plotted. This intuitively presents the flow and conversion efficiency of enterprise data value, and identifies core and secondary paths. In this embodiment, step S4 includes the following steps: Define market disturbance factors; these factors include sudden increases and decreases in order volume on the demand side, interruptions in raw material supply on the supply side, platform overload on the system side, and changes in the external environment. The value conversion efficiency curve is perturbed based on the aforementioned market disturbance factors to generate a virtual shock sequence; Monte Carlo simulation was performed on the virtual impact sequence to obtain the simulation results; The simulation results are decomposed into future time windows to extract the evolution trajectory of digital value.

[0044] In this embodiment, market disturbance factors are defined and quantified, including sudden increases and decreases in order volume on the demand side, interruptions in raw material supply on the supply side, platform overload on the system side, and changes in the external environment. Each type of disturbance factor is parameterized based on historical data and empirical patterns. For example, a sudden increase in order volume can be quantified by the ratio of the maximum order volume to the average order volume in a historical period, with a range of 1.5 to 3 times; a sudden decrease in orders can be defined as an event with a decrease of more than 50%; raw material supply interruptions are represented by the number of days of stockouts or supply delays, typically set to a simulation range of 1 to 10 days; platform overload is quantified by the ratio of the number of simultaneously active tasks to the system's processing capacity, with a range of 0.8 to 1.5 times; changes in the external environment can be simulated through policies, seasonality, or sudden event indices. These disturbance factors are normalized, mapping each disturbance dimension to a unified scale to form a disturbance factor vector. The disturbance factors are then distributed to the input variables of the value conversion efficiency curve in chronological order, dynamically adjusting the curve to form a disturbance response function. For example, in the event of a sudden surge in order volume, the input data can be scaled up proportionally within a specific time window, such as the next four periods, to calculate the offset of the value conversion rate as the input changes. For a disruption in raw material supply, a downward trend in the value conversion rate can be simulated, with the magnitude set to 10% to 30%. Changes in external environmental indices are then weighted and superimposed onto the conversion efficiency curves of different business areas. By combining various disturbances with certain probabilities, including using uniform or normal distributions to simulate the frequency of different disturbances, a continuous time series is formed, resulting in a simulated virtual impact sequence. This sequence records the change in value conversion efficiency at each time point under the influence of the disturbance, providing dynamic input for Monte Carlo simulation.

[0045] Based on the generated virtual impact sequences, Monte Carlo simulations are conducted to analyze the possible evolution paths of enterprise digital value. Monte Carlo simulations generate a large number of value conversion efficiency curve simulation results by randomly sampling different combinations and magnitudes of perturbation factors. Each iteration records the mapping relationship between input perturbation and output efficiency. The number of iterations can be set from 1000 to 5000 to ensure the stability and reliability of the statistical results. During the simulation, for each impact sequence, the value conversion rate at each future time point is calculated, and the maximum deviation, average deviation, and fluctuation range caused by the perturbation are recorded. By summarizing all simulation results, the distribution of value conversion efficiency under different perturbation scenarios can be obtained, including indicators such as mean, variance, and confidence interval.

[0046] Based on Monte Carlo simulation results, future time windows are decomposed to extract the digital value evolution trajectory. The time window can be set to 6 to 12 future periods. Within each window, the value conversion rate of the simulation results is statistically analyzed, including the average level, upper limit, lower limit, and fluctuation range. Time series analysis methods, such as moving averages or exponential smoothing, are used to smooth the simulation curves within the window to more clearly display the value evolution trend. Cumulative value gain or loss indicators can be calculated for each trajectory, forming a set of value evolution paths under different disturbance conditions. By aggregating and comparing these trajectories, the patterns and sensitivities of digital value changes for enterprises in the face of different market disturbances can be revealed.

[0047] In this embodiment, step S5 includes the following steps: The elasticity response indicators for different business areas of an enterprise are calculated based on the digital value evolution trajectory. These elasticity response indicators include efficiency recovery speed, resource dynamic reallocation cost, and path reconstruction complexity. Based on the elastic response index, multidimensional response coupling is performed to construct the elastic response surface; Digital value zoning and comprehensive digital capability assessment are performed based on elastic response surfaces, and a capability assessment report is output.

[0048] In this embodiment, after obtaining the enterprise's digital value evolution trajectory, elasticity response indicators for different business areas are calculated, including efficiency recovery speed, resource dynamic reallocation cost, and path reconstruction complexity. Efficiency recovery speed measures how quickly each business area of ​​the enterprise recovers to its original value level after market disturbances or internal business shocks. The time period required for each digital value trajectory to return to 95% or 100% of its original level from its initial deviation is analyzed and used as the efficiency recovery speed indicator, with units of time periods or standardized ratios. Resource dynamic reallocation cost quantifies the enterprise's consumption of resource adjustments, scheduling, and investment during the recovery process. It can be obtained by simulating resource flows between business activity nodes, including order scheduling, manpower allocation, and raw material scheduling, and accumulating adjustment costs. Path reconstruction complexity describes the difficulty for the enterprise to reconstruct business processes or dependent paths during value recovery. It can be quantified by analyzing the number of reconstruction paths, node degree changes, and loop adjustments of disturbed nodes in the business topology dependency network. Each indicator is standardized to map the values ​​of different business areas to a unified scale between 0 and 1. Using efficiency recovery speed, resource dynamic reallocation cost, and path reconstruction complexity as three-dimensional coordinate axes, the indicator values ​​corresponding to different business areas or different disturbance scenarios are mapped onto a surface. Interpolation methods, such as 3D spline interpolation or Kriging interpolation, are used to generate smooth surfaces to reflect the overall resilience characteristics of different business areas under various indicator combinations. During the construction process, the surface can be weighted and adjusted according to the indicator range and weights. For example, the weight of efficiency recovery speed can be set to 0.5, resource reallocation cost to 0.3, and path reconstruction complexity to 0.2, thus reflecting the enterprise's focus on different dimensions of resilience response. Each point on the surface not only reflects the indicator performance of a single business area but also shows the interactive coupling relationship between different indicators. By observing the height, tilt, and fluctuation trends of the surface, the enterprise's comprehensive recovery capability and vulnerable points under shocks can be revealed.

[0049] The surface is divided according to its elasticity level. For example, areas with high surface height, fast recovery speed, and low cost are defined as high-elasticity areas; areas with slow recovery speed, high cost, and complex paths are defined as low-elasticity areas; and intermediate areas are defined as medium-elasticity areas. This division can be combined with clustering methods, including 3D K-means, density-based clustering algorithms, while also considering the local slope of the surface and indicator weights to accurately reflect the elasticity performance of business areas under different disturbance scenarios. Based on the partitioning results, a comprehensive assessment of the enterprise's overall digital capabilities is conducted. The elasticity scores of each business area are weighted and summed to generate a digital capability scoring matrix or index. For example, high-elasticity areas are assigned a value of 1, medium-elasticity areas 0.5, and low-elasticity areas 0. The weighted total score is calculated according to the proportion of each business area, thus forming a quantifiable and comparable digital capability indicator.

[0050] In this embodiment, the specific steps for performing digital value partitioning and comprehensive digital capability assessment based on elastic response surfaces, and outputting a capability assessment report, are as follows: Digital value partitioning is performed based on the elastic response surface to obtain digital value partitions; the digital value partitions include stable operating ranges, performance transition boundaries, and capability failure critical regions. A digital capability resilience assessment is conducted on digital value zones to obtain a resilience assessment index. Based on the elastic response surface, reverse reasoning is performed to identify potential capability zones with high elasticity but low efficiency. Calculate the asset utilization rate and algorithm invocation potential of the potential capability zone to obtain implicit digital capabilities; A comprehensive digital capability assessment is conducted based on the resilience assessment index and implicit digital capabilities, and a capability assessment report is output.

[0051] In this embodiment, based on the elasticity response index values ​​of different business domains on the surface, regions with high elasticity, fast recovery speed, and low cost are defined as stable operating intervals, characterized by highly flat areas on the surface. Regions on the surface where elasticity indices change rapidly or reach critical points are designated as performance transition boundaries, typically corresponding to regions where efficiency declines rapidly and resource reconfiguration costs rise sharply. Regions with low elasticity indices, high path reconstruction complexity, and significantly increased resource reconfiguration costs are designated as capability failure critical zones. In specific implementation, a smooth surface is generated through three-dimensional interpolation, and the local maximum change slope is identified using gradient calculation methods. The threshold can be set such that a local slope greater than 0.2 to 0.3 is marked as a transition boundary, and a slope greater than 0.5 is marked as a failure zone, ensuring that the partitioning results can accurately depict the changes in business capabilities with disturbances.

[0052] The recovery speed, resource dynamic reconfiguration cost, and path reconstruction complexity of business nodes within a partition are normalized and weighted to generate a single-point resilience value. Then, the resilience values ​​of all business nodes within each partition are averaged to obtain the partition resilience index. Weighting can be based on the company's strategic priorities; for example, a weight of 0.5 for efficiency recovery speed, 0.3 for resource reconfiguration cost, and 0.2 for path reconstruction complexity. The resilience index for stable operation is typically higher than 0.8, the performance transition boundary index is between 0.5 and 0.8, while the critical failure zone index is lower than 0.5. The partition resilience index allows for quantitative analysis of the overall resilience of the enterprise and its various business areas.

[0053] Regions on the surface with elasticity indices above thresholds (e.g., 0.7 or 0.8) but below-average value conversion efficiency are identified and marked as potential capability zones. These zones indicate room for efficiency improvement while maintaining high business resilience, allowing for further exploration of resource optimization and algorithmic potential. Backward reasoning, combining local gradient analysis and boundary detection methods, calculates the local slope and location of efficiency troughs on the surface, identifies potential capability nodes and corresponding business areas, and generates potential capability zone diagrams, forming a quantifiable potential map.

[0054] After identifying potential capability zones, their asset utilization rate and algorithm invocation potential are calculated to obtain implicit digital capabilities. Asset utilization rate is quantified by statistically analyzing the ratio of the actual utilization rate of equipment, platforms, and resources within the potential capability zone to their maximum potential capacity. For example, an average utilization rate of 60% to 80% for key equipment or software modules indicates that there is still potential for utilization. Algorithm invocation potential is measured by analyzing the ratio of the callable frequency of data-driven models or automated algorithms for each business activity within the potential capability zone to the actual number of calls. The value ranges from 0 to 1 and is used to reflect the potential level of intelligence.

[0055] By weighting and aggregating the resilience index and implicit capability score, an overall digital capability index is generated. The resilience index is assigned a weight of 0.6, and the implicit capability score a weight of 0.4, ensuring that it reflects both the enterprise's resilience to disruptions and its potential for improvement. The report includes resilience partitioning maps for each business area, potential capability zone analysis, asset utilization and algorithm potential statistics, a comprehensive digital capability score, and visualization curves.

[0056] In this embodiment, an enterprise digital capability assessment system based on multi-source data fusion is provided, used to execute the enterprise digital capability assessment method based on multi-source data fusion as described above, including: The data acquisition module is used to acquire multi-source heterogeneous data from the target enterprise, analyze digital behavior trajectories, and construct a digital behavior matrix. The topology dependency module is used to model business topology dependencies based on a digital behavior matrix and build a dependency network between businesses. The value conversion module is used to collect multi-dimensional business results, perform digital value correlation analysis on the inter-business dependency network, and construct a value conversion efficiency curve. The perturbation simulation module is used to perform perturbation injection simulation based on the value conversion efficiency curve and extract the digital value evolution trajectory. The assessment module is used to calculate the resilience response indicators of different business areas of an enterprise based on the evolution trajectory of digital value, and to conduct a comprehensive digital capability assessment, outputting a capability assessment report.

[0057] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0058] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein are implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for assessing enterprise digital capabilities based on multi-source data fusion, characterized in that, Includes the following steps: Step S1: Obtain multi-source heterogeneous data of the target enterprise, analyze digital behavior trajectories, and construct a digital behavior matrix; Step S2: Model business topology dependencies based on the digital behavior matrix and construct an inter-business dependency network; Step S3: Collect multi-dimensional business results of enterprises, conduct digital value correlation analysis on the inter-business dependency network, and construct a value conversion efficiency curve; Step S4: Perform perturbation injection simulation based on the value conversion efficiency curve to extract the digital value evolution trajectory; Step S5: Calculate the resilience response indicators for different business areas of the enterprise based on the digital value evolution trajectory, conduct a comprehensive digital capability assessment, and output a capability assessment report.

2. The enterprise digital capability assessment method based on multi-source data fusion according to claim 1, characterized in that, The specific steps of step S1 are as follows: Acquire multi-source heterogeneous data from the target company; Cross-modal semantic alignment and timestamp synchronization are performed on the multi-source heterogeneous data to obtain a multi-source fusion dataset; Enterprise digital behavior trajectory analysis is performed based on multi-source fusion dataset to obtain behavioral feature tensors; A digital behavior matrix is ​​constructed by performing a low-dimensional embedding mapping on the behavioral feature tensor.

3. The enterprise digital capability assessment method based on multi-source data fusion according to claim 2, characterized in that, The multi-source heterogeneous data includes production execution system operation logs, supply chain collaboration platform interaction records, industrial IoT device sensing data, enterprise application software call link tracing data, and external e-commerce and government platform interaction data.

4. The enterprise digital capability assessment method based on multi-source data fusion according to claim 3, characterized in that, The specific steps of step S2 are as follows: Enterprise business activities are identified based on digital behavior matrices, and each business activity node is marked. Calculate the data dependency strength and redundant path ratio between each business activity node to obtain the relationship between nodes; Graph structure analysis is performed based on the relationships between nodes to obtain circular dependencies and abnormal coupling paths; Business topology dependency modeling is performed based on circular dependencies and abnormal coupling paths to construct a dependency network between businesses.

5. The enterprise digital capability assessment method based on multi-source data fusion according to claim 4, characterized in that, Step S3 is as follows: Collect multi-dimensional business results from enterprises, including order conversion rate, inventory turnover cycle, R&D delivery cycle, and customer response time. Based on the inter-business dependency network, causal relationship modeling is performed on the multi-dimensional business results of the enterprise to generate business transmission paths; Calculate the time delay attenuation coefficient and information loss rate of the business transmission path to obtain the data value conversion function; Based on the data value transformation function, we can model the data value of enterprises in multiple fields and construct a value transformation efficiency curve.

6. The enterprise digital capability assessment method based on multi-source data fusion according to claim 5, characterized in that, The specific steps of step S4 are as follows: Define market disturbance factors; The value conversion efficiency curve is perturbed based on the aforementioned market disturbance factors to generate a virtual shock sequence; Monte Carlo simulation was performed on the virtual impact sequence to obtain the simulation results; The simulation results are decomposed into future time windows to extract the trajectory of digital value evolution.

7. The enterprise digital capability assessment method based on multi-source data fusion according to claim 6, characterized in that, The market disturbance factors include sudden increases and decreases in order volume on the demand side, interruptions in raw material supply on the supply side, platform overload on the system side, and changes in the external environment.

8. The enterprise digital capability assessment method based on multi-source data fusion according to claim 6, characterized in that, The specific steps of step S5 are as follows: The elasticity response indicators for different business areas of an enterprise are calculated based on the digital value evolution trajectory. These elasticity response indicators include efficiency recovery speed, resource dynamic reallocation cost, and path reconstruction complexity. Based on the elastic response index, multidimensional response coupling is performed to construct the elastic response surface; Digital value zoning and comprehensive digital capability assessment are performed based on elastic response surfaces, and a capability assessment report is output.

9. The enterprise digital capability assessment method based on multi-source data fusion according to claim 7, characterized in that, The specific steps for performing digital value partitioning and comprehensive digital capability assessment based on elastic response surfaces, and outputting a capability assessment report, are as follows: Digital value partitioning is performed based on the elastic response surface to obtain digital value partitions; the digital value partitions include stable operating ranges, performance transition boundaries, and capability failure critical regions. A digital capability resilience assessment is conducted on digital value zones to obtain a resilience assessment index. Based on the elastic response surface, reverse reasoning is performed to identify potential capability zones with high elasticity but low efficiency. Calculate the asset utilization rate and algorithm invocation potential of the potential capability zone to obtain implicit digital capabilities; A comprehensive digital capability assessment is conducted based on the resilience assessment index and implicit digital capabilities, and a capability assessment report is output.

10. A system for assessing enterprise digital capabilities based on multi-source data fusion, characterized in that, The method for performing the enterprise digital capability assessment based on multi-source data fusion as described in claim 1 includes: The data acquisition module is used to acquire multi-source heterogeneous data from the target enterprise, analyze digital behavior trajectories, and construct a digital behavior matrix. The topology dependency module is used to model business topology dependencies based on a digital behavior matrix and build a network of dependencies between businesses. The value conversion module is used to collect multi-dimensional business results, perform digital value correlation analysis on the inter-business dependency network, and construct a value conversion efficiency curve. The perturbation simulation module is used to perform perturbation injection simulation based on the value conversion efficiency curve and extract the digital value evolution trajectory. The assessment module is used to calculate the resilience response indicators of different business areas of an enterprise based on the evolution trajectory of digital value, and to conduct a comprehensive digital capability assessment, outputting a capability assessment report.

Images