Intelligent analysis method for enterprise digital transformation based on big data analysis

Abstract

The present application relates to the technical field of big data analysis, and discloses an intelligent analysis method for enterprise digital transformation based on big data analysis.The present application solves the technical problems of traditional analysis methods lacking dynamic feature portraits, insufficient mining of implicit behavior characteristics, and low accuracy of positioning the short board in the transformation stage.The present application first collects enterprise multi-source heterogeneous data and pre-processes, extracts explicit state features and implicit behavior features containing data flow barrier coefficients and decision-making intelligence indexes, and fuses to build a two-dimensional feature portrait;Then, through weighted Euclidean distance combined with random forest weighting, similarity matching is completed, and the reference sample set is screened;Finally, the enterprise stage is identified according to the sample set, the core short board is positioned by benchmarking, and a diagnosis report is generated;The method realizes the full-dimensional quantification and accurate diagnosis of enterprise transformation state, provides scientific and feasible decision-making basis for enterprise digital transformation, and greatly improves the objectivity and pertinence of analysis.

Description

Technical Field

[0001] This invention relates to the field of big data analytics, and in particular to an intelligent analytics method for enterprise digital transformation based on big data analytics. Background Technology

[0002] With the deep integration of the digital economy and the real economy, digital transformation has become an inevitable choice for enterprises to enhance their core competitiveness. The industry urgently needs precise analysis methods based on big data to comprehensively assess, identify stages, and pinpoint shortcomings in the status of enterprise digital transformation.

[0003] Traditional enterprise digital transformation analysis methods often employ a unified indicator system for horizontal comparison or rely on expert experience for qualitative assessment. These methods suffer from the following significant technical problems: When the analysis platform possesses massive amounts of enterprise data, it cannot dynamically construct a digital profile of the analyzed enterprise, making it impossible to accurately match reference samples with similar characteristics from massive historical data. Furthermore, because internal digital activities are scattered across multiple heterogeneous systems, the log data generated by these systems lacks a unified framework for extracting implicit features. This results in a large amount of behavioral data that could reflect the enterprise's true level of digitalization being unusable, ultimately making it difficult to quantify the enterprise's true transformation stage and pinpoint its shortcomings.

[0004] Therefore, there is an urgent need for an intelligent analysis method for enterprise digital transformation based on big data analytics to address the many shortcomings of existing technologies. Summary of the Invention

[0005] The technical problem this invention aims to solve is that existing enterprise digital transformation analysis methods cannot construct a digital profile of an enterprise or accurately determine its true level of digitalization. Therefore, we propose an intelligent analysis method for enterprise digital transformation based on big data analysis.

[0006] To achieve the above objectives, this application adopts the following technical solution: an intelligent analysis method for enterprise digital transformation based on big data analysis. This method constructs a dual-dimensional feature profile that integrates explicit states and implicit behaviors, and performs comparative analysis based on a reference sample set with dynamic similarity matching to accurately identify the enterprise's transformation stage and shortcomings. The method includes the following steps:

[0007] Step 1: Collect heterogeneous data from multiple sources within the enterprise and preprocess it to generate standardized behavioral sequences;

[0008] Step 2: Extract explicit state features and implicit behavioral features from the preprocessed data. Implicit behavioral features include at least the data flow barrier coefficient, which is used to quantify the degree of data flow obstruction, and the decision intelligence index, which is used to assess the level of intelligence in the decision-making process.

[0009] Step 3: Integrate explicit state characteristics and implicit behavioral characteristics to construct a two-dimensional feature profile that represents the state of enterprise digital transformation;

[0010] Step 4: Perform similarity matching between the two-dimensional feature profile and historical data in the company's own database to select a reference sample set;

[0011] Step 5: Identify the current stage of digital transformation of the enterprise based on the reference sample set, and locate the shortcomings of transformation by comparing with the characteristics of benchmark enterprises, and generate a diagnostic report.

[0012] Preferably, the multi-source heterogeneous data in step one includes structured operational data, semi-structured process data, and unstructured behavioral data from enterprise resource planning systems, customer relationship management systems, office automation systems, and production execution systems; structured operational data includes financial data, sales data, inventory data, and equipment operation data; semi-structured process data includes approval flow logs, work order transfer records, and project collaboration records; and unstructured behavioral data includes system login logs, function module click flows, API call records, and data query logs.

[0013] Preferably, the preprocessing in step one includes: cleaning the collected data to remove duplicate data and correct format errors; filling missing values ​​using mean interpolation or nearest neighbor interpolation; removing outliers based on the 3σ principle or box plot method; aligning the timestamps of different systems to Coordinated Universal Time; and mapping the original operation records to preset digital behavior labels such as "data query", "process approval", "report generation" and "system configuration" according to business semantics.

[0014] Preferably, the explicit status characteristics in step two include the number of IT systems deployed by the enterprise, the online business process rate, and the maturity index of the data platform construction; the number of IT systems is the number of various business systems that the enterprise has launched and is running; the online business process rate is calculated by the ratio of the number of business processes processed online to the total number of business processes; the maturity index of the data platform construction is obtained by comprehensively scoring the data collection coverage, data model completeness, and number of data service interfaces.

[0015] Preferably, when extracting the data flow barrier coefficient in step two, the following parameters need to be statistically analyzed from the data access log:

[0016] The total number of departments within the enterprise, the number of times each employee in each department accesses their department's data, the total number of times all employees in each department access data, and the organizational level code for each department. The organizational level code is assigned a value based on the department's position in the enterprise's management structure, with the highest management level code being 1, and the code for lower-level departments increasing as the level decreases.

[0017] Preferably, when extracting the decision intelligence index in step two, the following parameters need to be statistically analyzed from the business decision log:

[0018] The total number of enterprise business decision-making scenarios, the number of decisions made automatically by the system without human intervention in each scenario, the number of decisions made manually in each scenario, the standard deviation of the time taken for all automated decisions in each scenario, and the arithmetic mean of the confidence scores of all automated decision results. Decision time refers to the time interval from triggering a decision to generating a result. Confidence score is obtained by normalizing the probability value or score output by the decision model to the range of 0 to 1.

[0019] Preferably, in step four, the similarity matching uses weighted Euclidean distance to calculate the similarity between the current enterprise feature profile and the historical enterprise feature vectors. The weights of each feature dimension are determined by the information gain obtained by training a random forest model on the company's own enterprise database. The top-ranked similarity profiles are then selected. These companies serve as a reference sample set. Take an integer between 5 and 20.

[0020] Preferably, in step five, the digital transformation stage is divided into five levels: initial level, standardization level, integration level, optimization level, and leading level. The weakness identification is achieved by calculating the difference between the current enterprise's characteristic dimensions and the corresponding characteristic dimensions of benchmark enterprises in the next transformation stage in the reference sample set, selecting the top three dimensions with the largest differences as the main weaknesses, and generating improvement suggestions based on the transformation path data of similar enterprises in the historical database.

[0021] Preferably, the data flow barrier coefficient Calculated using the following formula:

[0022] ;

[0023] in This indicates the total number of departments within the company;

[0024] Indicates the first The number of times each department's internal employees access their department's data;

[0025] Indicates the first Total number of data accesses by all employees in each department;

[0026] Indicates the first Organizational hierarchy coding for each department;

[0027] Indicates the first Organizational hierarchy coding for each department;

[0028] This is a hierarchical normalization constant, and its value represents the maximum span of the enterprise's management levels.

[0029] It is a natural constant;

[0030] This indicates the range from the 1st to the 2nd. Each department seeks a summation;

[0031] Indicates that except for the first Summation is performed on all departments except for the one mentioned department;

[0032] Indicates the first The department and the first The absolute value of the difference in departmental level codes.

[0033] Preferred decision-making intelligence index Calculated using the following formula:

[0034] ;

[0035] in, This represents the total number of types of business decision-making scenarios for an enterprise;

[0036] Indicates the first The number of decisions made automatically by the system in decision-making scenarios without human intervention;

[0037] Indicates the first The number of decisions that require human intervention in decision-making scenarios;

[0038] Indicates the first The standard deviation of the time spent on all automated decisions in similar decision-making scenarios;

[0039] This is a time normalization constant, and its value is the average of the time consumed by automated decision-making in all decision-making scenarios.

[0040] This represents the arithmetic mean of the confidence levels of all automated decision outcomes.

[0041] This represents the maximum confidence level, with a value of 1.

[0042] It is a natural constant;

[0043] Indicates the range from category 1 to category 2. Summation is performed in decision-making scenarios.

[0044] The technical effects and advantages of this invention are as follows:

[0045] This invention constructs a dual-dimensional feature profile that integrates explicit states and implicit behaviors. It uses data flow barrier coefficients and decision-making intelligence indices to quantitatively analyze an enterprise's implicit digital capabilities, fully leveraging the value of multi-source heterogeneous log data to comprehensively depict the true state of enterprise transformation. This breaks through the limitations of traditional methods that only focus on explicit construction indicators. Simultaneously, it accurately selects reference samples from massive amounts of enterprise data, avoiding biases from horizontal comparisons of unified indicators. By benchmarking against leading enterprises to identify core weaknesses and generating targeted improvement suggestions based on historical transformation cases, it significantly improves the objectivity and accuracy of transformation analysis, providing a scientific and implementable decision-making basis for the strategic formulation and execution of enterprise digital transformation. Attached Figure Description

[0046] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts:

[0047] Figure 1 This is a schematic diagram of the process structure of the present invention;

[0048] Figure 2 This is a schematic diagram of the logical structure of the present invention. Detailed Implementation

[0049] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0050] Reference Figures 1-2 As shown, the present invention provides a technical solution: The present invention provides an intelligent analysis method for enterprise digital transformation based on big data analysis. It constructs a two-dimensional feature profile that integrates explicit state features and implicit behavioral features, uses dynamic similarity matching technology to select a reference sample set that is most similar to the target enterprise's features from a historical enterprise database, and then accurately identifies the target enterprise's current digital transformation stage. By comparing the features with those of benchmark enterprises, it accurately locates the transformation shortcomings and finally generates a targeted diagnostic report.

[0051] The method consists of five main steps: Step 1 is the data acquisition and preprocessing stage, which is responsible for collecting raw data from the enterprise's multi-source heterogeneous information systems and performing standardization processing; Step 2 is the feature extraction stage, which is responsible for extracting explicit state features and implicit behavioral features from the preprocessed data; Step 3 is the feature fusion stage, which is responsible for fusing explicit state features and implicit behavioral features to construct a two-dimensional feature profile; Step 4 is the similarity matching stage, which is responsible for calculating the similarity between the feature profile of the target enterprise and the historical enterprise database to select a reference sample set; Step 5 is the diagnostic output stage, which is responsible for identifying the transformation stage and locating the transformation shortcomings based on the reference sample set, and generating a diagnostic report.

[0052] Compared with existing qualitative assessment methods that rely solely on financial indicators or single dimensions, this invention introduces quantitative analysis of implicit behavioral characteristics, particularly the two key indicators of data flow barrier coefficient and decision intelligence index. This allows for a deeper understanding of the true state and potential problems of enterprises during their digital transformation, thereby achieving a more accurate and objective transformation diagnosis.

[0053] In step one, it is first necessary to collect multi-source heterogeneous data from the enterprise. The multi-source heterogeneous data involved in this invention comes from multiple core business systems within the enterprise, specifically including the following categories:

[0054] Structured operational data refers to business data with a fixed data format and clearly defined fields, mainly sourced from enterprise resource planning systems, production execution systems, etc.; specifically including: financial data, such as financial account data such as balance sheet, income statement, cash flow statement, etc.; sales data, including sales revenue, sales quantity, customer orders, sales channels, etc.; inventory data, covering raw material inventory, work-in-process inventory, finished goods inventory, etc.; equipment operation data, including equipment integration rate, failure rate, maintenance records, etc.

[0055] Semi-structured process data refers to business process data that has certain format specifications but relatively flexible field structures. It mainly comes from office automation systems and enterprise workflow engines. Specifically, it includes: approval flow logs, which record the initiation, transfer, approval and archiving information of various approval processes; work order flow records, including the creation, allocation, processing and closure information of IT work orders, maintenance work orders, service work orders, etc.; and project collaboration records, which involve the entire life cycle data of projects, including project initiation, task allocation, progress tracking and deliverables.

[0056] Unstructured behavioral data refers to procedural data that records user actions, primarily collected through system tracking. Specifically, it includes: system login logs, recording user login time, login IP address, online duration, session identifier, etc.; functional module clickstreams, recording user click paths, dwell times, access sequences, and other interactive behaviors on the front-end interface; application programming interface call records, including technical logs such as interface caller, call parameters, return results, and response time; and data query logs, recording user-initiated data query conditions, query scope, and export operations.

[0057] The collection of the aforementioned multi-source heterogeneous data is achieved through system integration, data push, or log collection. For scenarios with high real-time requirements, message queues such as Kafka are used for real-time data collection. For supplementary collection of historical data, batch import is performed through database export or ETL tools.

[0058] The collected raw data often suffers from inconsistent formats, missing values, duplicate data, outliers, and other issues, making it unsuitable for direct use in subsequent feature analysis. Therefore, in step one, the collected data needs to undergo systematic preprocessing, which includes the following steps:

[0059] Data cleaning is the first step in preprocessing, mainly addressing the following issues: removing duplicate data by comparing the primary key or key field identifiers of data records to delete duplicate records; correcting format errors by standardizing date formats, numeric formats, encoding formats, etc., exported from different systems, such as converting different formats like "2024 / 01 / 15", "2024-01-15", and "20240115" into a standard format; and processing invalid data by removing invalid records that are empty, have severely missing fields, or clearly do not conform to business logic.

[0060] Missing value handling is a crucial step in ensuring data integrity. This invention employs the following strategies for missing value imputation: For numeric fields, mean interpolation is used, with the historical mean of the field serving as the imputation value for missing values; or nearest neighbor interpolation is used, with linear interpolation based on the values ​​of adjacent records before and after the missing value; for categorical fields, mode imputation or inference imputation based on business rules is used; in practical applications, an appropriate imputation method is selected based on data distribution characteristics and business scenarios.

[0061] Outlier removal is a crucial step in ensuring data quality. This invention employs two complementary outlier detection methods: the three-standard-deviation principle, which identifies outliers for numerical data that follows or approximately follows a normal distribution, values ​​exceeding the mean ± three standard deviations; and the box plot method, which uses the interquartile range principle of box plots to identify outliers for non-normally distributed data, values ​​less than Q1 - 1.5 × IQR or greater than Q3 + 1.5 × IQR. Detected outliers are not directly deleted but are marked as pending confirmation, allowing business personnel to decide whether to correct or remove them based on the actual situation.

[0062] Time alignment is a key step in solving the problem of inconsistent time in multi-source data. Since timestamps from different systems may use different time zones or different precisions, this invention converts the timestamps of all systems into Coordinated Universal Time for storage and calculation. For the problem of inconsistent time precision, milliseconds are used as the smallest unit of time. For scenarios where business needs to be displayed in a specific time zone, time zone conversion is performed again in the final output.

[0063] Behavioral tag mapping is the core step in transforming raw operation records into standardized digital behavioral tags. This invention maps the complex raw operation records into four pre-defined categories of digital behavioral tags based on business semantics: data query tags, including data retrieval, report viewing, and indicator browsing; process approval tags, including process initiation, approval operations, and process tracking; report generation tags, including report creation, report export, and data analysis; and system configuration tags, including parameter settings, permission configuration, and function on / off. This tagging process makes subsequent feature extraction and analysis more standardized and efficient.

[0064] After the above preprocessing, the original multi-source heterogeneous data is transformed into a standardized behavior sequence. Each record contains standardized fields such as timestamp, behavior subject (employee or system), behavior type (digital behavior tag), behavior object (business entity), and behavior result.

[0065] Explicit state characteristics are objective indicators that can be directly observed and statistically analyzed from enterprise information systems and reflect the current state of enterprise digital transformation. In step two, this invention extracts the following three types of explicit state characteristics:

[0066] The number of information technology systems refers to the total number of various business information systems that an enterprise has launched and is running. When making statistics, it is necessary to distinguish different types of systems, including but not limited to: ERP systems, CRM systems, OA systems, MES systems, supply chain management systems, product lifecycle management systems, and business intelligence systems. The statistical method is to enumerate all information systems in use within the enterprise, remove historical systems that have been taken offline or discontinued, and finally use the number of valid systems as the value of this feature.

[0067] The online business process penetration rate is calculated using the following formula:

[0068] Online business process rate = Number of business processes processed online / Total number of business processes

[0069] Among them, online business processes refer to business processes that are circulated electronically through information systems and completed with full-process approval or execution; the total number of business processes refers to the total number of all business processes that the enterprise needs to process within the evaluation period, including online and offline processes; this feature reflects the breadth of the enterprise's digital transformation of processes from the perspective of business execution.

[0070] The maturity index for data platform construction is a comprehensive scoring indicator, calculated by weighting the following three sub-dimensions: Data collection coverage rate, which refers to the proportion of core business data of an enterprise that is incorporated into a unified data platform; the calculation formula is the number of business items covered by collected data divided by the total number of business items; Data model completeness, which refers to the quantity and quality of data models already built in the enterprise's data warehouse or data lake, and can be quantified through indicators such as model document completeness and model reuse rate; Number of data service interfaces, which refers to the total number of data service interfaces provided by the enterprise, including data query interfaces, data push interfaces, and data synchronization interfaces. The calculation formula for the maturity index of data platform construction is as follows:

[0071] Data platform construction maturity index = w1 × data collection coverage + w2 × data model completeness + w3 × number of data service interfaces;

[0072] w1, w2, and w3 are preset weighting coefficients that can be adjusted according to industry characteristics and the actual situation of the enterprise.

[0073] Implicit behavioral characteristics are soft indicators deeply mined from a company's daily operational data, reflecting its digital operational capabilities and data-driven culture. Unlike explicit behavioral characteristics, implicit behavioral characteristics are difficult to observe directly and require quantification through specific algorithms and models. This invention focuses on extracting the following two types of implicit behavioral characteristics:

[0074] The data flow barrier coefficient is one of the core implicit behavioral features proposed in this invention, used to quantify the degree of obstruction to data flow within an enterprise. The higher the data flow barrier, the lower the degree of data sharing between different departments within the enterprise, the more serious the data silo phenomenon, and the greater the deep-seated obstacles to digital transformation.

[0075] Before calculating the data flow barrier coefficient, the following statistical parameters need to be extracted from the data access logs:

[0076] (1) Total number of departments The number of departments within the company that manage independent data access permissions;

[0077] (2) Number of internal department visits : No. The total number of times employees within a department accessed their department's data;

[0078] (3) Total number of visits to the department : No. The total number of data accesses by all employees in each department, including the total number of accesses to data within their own department and the total number of accesses to data from other departments;

[0079] (4) Departmental organizational hierarchy coding The numerical code is assigned based on the department's position in the enterprise management structure, with the highest management level coded as 1, and the code for lower-level departments increasing as the level decreases; for example, the General Manager's Office is level 1, each business unit / functional department is level 2, each business unit is level 3, and the grassroots team or project team is level 4.

[0080] Based on the above parameters, the data flow barrier coefficient Calculated using the following formula:

[0081] ;

[0082] The formula is designed to take into full account the two opposing factors of data flow: "internal aggregation" and "cross-departmental barriers".

[0083] The numerator represents the average percentage of data accessed within a department; the higher this ratio, the more likely departments are to access data within their own departments, and the higher the internal aggregation degree of data flow; among which, Indicates the first Internal access percentage of each department The bigger, The smaller the value, the higher the degree of data secrecy in that department.

[0084] The denominator represents a correction term for cross-departmental flow resistance; this term takes into account the data flow resistance between different levels of departments. This indicates the hierarchical distance between two departments. The smaller the hierarchical distance, the higher the hierarchical distance. The closer the index is to 1, the easier it is for data to flow across levels; the greater the hierarchical distance, the closer the index is to 0, indicating that data flow across levels is more difficult. This is a hierarchical normalization constant, taking the value of the maximum span of management levels in the enterprise; for example, if the enterprise has 4 management levels, then... .

[0085] The data flow barrier coefficient P ranges from (0, +∞). The higher the value, the higher the data flow barriers within the enterprise, and the more serious the data silo phenomenon. The smaller the value, the smoother the data flow within the enterprise and the higher the degree of data sharing.

[0086] The decision intelligence index is another core implicit behavioral feature proposed in this invention, used to assess the level of intelligence in an enterprise's business decision-making process. The higher the decision intelligence index, the more the enterprise relies on data-driven and algorithmic models for business decisions, rather than on human experience and judgment.

[0087] Before calculating the decision intelligence index, the following statistical parameters need to be extracted from the business decision logs:

[0088] (1) Total number of decision-making scenario types The number of business decision-making scenarios involved in the enterprise during the evaluation period, such as: inventory replenishment decisions, customer credit assessments, marketing campaign recommendations, production scheduling decisions, etc.

[0089] (2) Number of automated decision-making processes : No. The number of decisions made automatically by the system in decision-making scenarios without human intervention;

[0090] (3) Number of times human intervention in decision-making : No. The number of decisions that require human intervention in decision-making scenarios, including decisions that are completely dominated by humans and decisions made in collaboration with machines;

[0091] (4) Standard deviation of time spent on automated decision-making : No. The standard deviation of the time taken for all automated decisions in a decision-making scenario. Decision time is defined as the time interval from when a decision is triggered to when the decision result is generated;

[0092] (5) Average confidence level of automated decision-making : The arithmetic mean of the confidence scores of all automated decision results; the confidence score is obtained by normalizing the probability value or score output by the decision model to the [0,1] interval.

[0093] Based on the above parameters, the decision-making intelligence index Calculated using the following formula:

[0094] ;

[0095] This formula takes into account three dimensions: the degree of automation in decision-making, the stability of decision-making, and the reliability of decision-making.

[0096] This represents the automation rate of the k-th type of decision-making scenario, i.e., the proportion of automated decisions to total decisions; the higher this proportion, the more the decision-making scenario relies on the system to complete it automatically.

[0097] This is the stability correction term for decision-making time; where The standard deviation of the time spent on automated decision-making reflects the volatility of the decision-making process; This is a time normalization constant, taking the average value of the automated decision-making time across all decision-making scenarios; the purpose of this correction term is to address situations where decision-making time fluctuates significantly. Larger It will approach 0, which will have a punitive effect on the overall index, indicating that the decision-making process is not stable and reliable enough.

[0098] For the decision confidence normalization term, where The value is 1, so this item is actually the arithmetic mean of the confidence level of automated decision-making; this factor reflects the reliability of the decision results.

[0099] Overall, the decision-making intelligence index The value range is [0,1]; The closer the value is to 1, the higher the level of intelligent decision-making of the enterprise, and the stronger its culture and ability to drive data-driven decision-making. The closer the value is to 0, the more the enterprise still relies on manual decision-making, and the greater the potential for deeper improvement in digital transformation.

[0100] In step three, the explicit state features and implicit behavioral features extracted in step two need to be integrated to construct a two-dimensional feature profile that represents the state of the enterprise's digital transformation.

[0101] The data structure of a two-dimensional feature profile can be represented as a multi-dimensional feature vector, where the first five dimensions are the core feature dimensions: number of information technology systems, online business process rate, maturity of data platform construction, data flow barrier coefficient, and decision-making intelligence index; subsequent dimensions can be expanded according to actual needs, such as adding supplementary features like digital investment intensity, average digital training time per person, and data quality score.

[0102] The feature fusion strategy adopts a normalized splicing method: First, the feature dimensions are normalized to eliminate the difference in units; then, the normalized features are spliced ​​in a preset order to form a unified two-dimensional feature profile vector.

[0103] The dual-dimensional feature profile comprehensively depicts the digital transformation status of enterprises from two dimensions: explicit infrastructure and implicit operational capabilities, providing a complete data foundation for subsequent similarity matching and diagnostic analysis.

[0104] In step four, it is necessary to perform similarity matching between the target company's two-dimensional feature profile and the historical company data in the company's own database; this invention uses weighted Euclidean distance as the similarity measurement method.

[0105] The formula for calculating the weighted Euclidean distance is as follows:

[0106] ;

[0107] in, For the feature vector of the target company, For the feature vector of historical enterprises, The total number of feature dimensions. For the first The weights of each feature dimension.

[0108] Weights of each feature dimension The random forest model was trained on our own enterprise database, and information gain was used as the basis for determining the weights.

[0109] The specific steps are as follows:

[0110] (1) Prepare training dataset: Select historical enterprise samples labeled with transformation stage from our own enterprise database. Each sample contains a two-dimensional feature vector and the corresponding transformation stage label;

[0111] (2) Training the random forest classification model: Using feature vectors as input features and transition stage labels as target variables, train the random forest classification model;

[0112] (3) Calculate information gain: Using the trained random forest model, calculate the contribution of each feature dimension to the classification task, i.e., the information gain value; the larger the information gain, the more significant the role of the feature in distinguishing different transition stages.

[0113] (4) Normalized weights: The information gain values ​​of each feature dimension are normalized to obtain the weights of each dimension. .

[0114] The feature weights determined by the above method can objectively reflect the degree of contribution of different features to the identification of the transformation stage, avoiding the bias that may be caused by subjective weighting.

[0115] After calculating the weighted Euclidean distance between the target company and all historical companies, the companies are sorted by distance from smallest to largest, and the companies with the highest similarity are selected. One enterprise was used as a reference sample set; The value ranges from 5 to 20, and can be adjusted according to the size of the historical database and the sample distribution density.

[0116] The companies in the reference sample set should have the following characteristics: they should have a high degree of similarity to the target companies in terms of digital transformation features; and they should cover different stages of transformation in order to identify the stages and pinpoint the shortcomings later.

[0117] This invention categorizes enterprise digital transformation into the following five levels:

[0118] Level 1: Initial Level; The characteristics of enterprise digitalization at this stage are: a small number of information technology systems, mainly single business systems; a low rate of online business processes, with a large number of business processes still being handled manually offline; a data platform has not yet been established or is in its infancy; high barriers to data flow, with data from different departments being fragmented; and decision-making mainly relies on human experience, with very few applications of automated decision-making.

[0119] Level 2: Standardization Level; At this stage, enterprises have established multiple core business information systems and achieved online processing of major business processes; they have begun to plan and build data platforms; data sharing between departments has begun to be standardized, but certain barriers still exist; some scenarios have begun to try automated decision-making, but the application scope is limited.

[0120] Level 3: Integration Level; At this stage, enterprises have completed the integration of core business systems and achieved cross-system data interoperability; the data platform has been basically built and can provide unified data services; data flow barriers have been significantly reduced and cross-departmental data access is relatively smooth; automated decision-making has been applied in multiple business scenarios.

[0121] Level 4: Optimization Level; At this stage, the enterprise's data platform is highly mature and has a sound data governance system; data-driven decision-making has become the norm, and the decision-making intelligence index is relatively high; intelligent applications are being explored, such as intelligent recommendation and intelligent prediction; digital transformation begins to shift from efficiency improvement to value creation.

[0122] Level 5: Leading Level; Enterprises at this stage are at the forefront of the industry in terms of digitalization, becoming a benchmark for digital transformation in the industry; they operate data as assets and have outstanding data value monetization capabilities; they have a high degree of intelligent decision-making, enabling intelligent decision-making in complex scenarios; and they have strong digital innovation capabilities, continuously driving business model innovation.

[0123] Based on the reference sample set selected in step four, the following method is used to identify the transformation stage of the target company:

[0124] (1) The number of enterprises in each stage of transformation in the statistical reference sample set;

[0125] (2) Calculate the weighted Euclidean distance between the target firm and the feature centers of each transformation stage;

[0126] (3) Select the transformation stage with the shortest distance as the current transformation stage of the target enterprise;

[0127] (4) At the same time, output the credibility assessment of this stage. The credibility calculation formula is: credibility = 1 - (minimum distance / second smallest distance). The closer this value is to 1, the higher the credibility of the stage identification.

[0128] The specific methods for identifying weaknesses are as follows:

[0129] (1) Determine the next transformation stage for the target company;

[0130] (2) Select benchmark enterprises in the next stage of transformation from the reference sample set;

[0131] (3) Calculate the difference between each feature dimension of the target company and the corresponding feature dimension of the benchmark company, and use the normalized difference;

[0132] (4) Select the three dimensions with the largest differences as the main weaknesses of the target company;

[0133] (5) Based on the transformation path data of similar enterprises in the historical database, generate targeted improvement suggestions for each shortcoming dimension.

[0134] For example, if the target company has the largest difference between itself and the benchmark company in the data flow barrier coefficient dimension, then that dimension is identified as the main weakness. The system will search the historical database for similar company cases that have successfully reduced data flow barriers, summarize the measures they have taken, such as establishing unified data standards, deploying data exchange platforms, and implementing data governance systems, and generate targeted improvement suggestions.

[0135] Ultimately, the diagnostic report includes the following: the target company's current transformation stage and credibility assessment; current values ​​of each characteristic dimension and gaps compared to benchmark companies; identified major weaknesses and their priority ranking; improvement suggestions and priority ranking for each weakness; and referenced benchmark company case information.

[0136] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. An intelligent analysis method for enterprise digital transformation based on big data analytics, characterized in that: By constructing a two-dimensional feature profile that integrates explicit states and implicit behaviors, and conducting comparative analysis based on a reference sample set of dynamic similarity matching, the system can accurately identify the transformation stage and shortcomings of enterprises. This includes the following steps: Step 1: Collect heterogeneous data from multiple sources within the enterprise and preprocess it to generate standardized behavioral sequences; Step 2: Extract explicit state features and implicit behavioral features from the preprocessed data. The implicit behavioral features include at least a data flow barrier coefficient for quantifying the degree of data flow obstruction and a decision intelligence index for assessing the level of intelligence in the decision-making process. Step 3: Integrate the explicit state features and implicit behavioral features to construct a two-dimensional feature profile that represents the state of enterprise digital transformation; Step 4: Perform similarity matching between the dual-dimensional feature profile and historical data in the enterprise's own database to select a reference sample set; Step 5: Identify the current stage of digital transformation of the enterprise based on the reference sample set, and locate the shortcomings of transformation by comparing with the characteristics of benchmark enterprises, and generate a diagnostic report.

2. The intelligent analysis method for enterprise digital transformation based on big data analysis according to claim 1, characterized in that, The multi-source heterogeneous data mentioned in Step 1 includes structured operational data, semi-structured process data, and unstructured behavioral data from enterprise resource planning systems, customer relationship management systems, office automation systems, and production execution systems. The structured operational data includes financial data, sales data, inventory data, and equipment operation data. The semi-structured process data includes approval flow logs, work order transfer records, and project collaboration records. The unstructured behavioral data includes system login logs, function module click flows, API call records, and data query logs.

3. The intelligent analysis method for enterprise digital transformation based on big data analysis according to claim 1, characterized in that, The preprocessing described in step one includes: cleaning the collected data to remove duplicate data and correct format errors; filling missing values ​​using mean interpolation or nearest neighbor interpolation; removing outliers based on the 3σ principle or box plot method; aligning the timestamps of different systems to Coordinated Universal Time; and mapping the original operation records to preset digital behavior labels such as "data query", "process approval", "report generation" and "system configuration" according to business semantics.

4. The intelligent analysis method for enterprise digital transformation based on big data analysis according to claim 1, characterized in that, The explicit status characteristics mentioned in step two include the number of IT systems deployed by the enterprise, the online business process rate, and the maturity index of the data platform construction; the number of IT systems is a statistical measure of the number of various business systems that the enterprise has launched and is running online; the online business process rate is calculated by the ratio of the number of business processes processed online to the total number of business processes; the maturity index of the data platform construction is obtained by comprehensively scoring the data collection coverage, data model completeness, and number of data service interfaces.

5. The intelligent analysis method for enterprise digital transformation based on big data analysis according to claim 1, characterized in that, When extracting the data flow barrier coefficient in step two, the following parameters need to be statistically analyzed from the data access log: The number of departments within the enterprise, the number of times each employee in each department accesses their department's data, the total number of times all employees in each department access data, and the organizational level code for each department. The organizational level code is assigned a value based on the department's position in the enterprise's management structure, with the highest management level code being 1 and the code for lower-level departments increasing as the level decreases.

6. The intelligent analysis method for enterprise digital transformation based on big data analysis according to claim 1, characterized in that, When extracting the decision intelligence index in step two, the following parameters need to be collected from the business decision log: The total number of enterprise business decision-making scenarios, the number of decisions made automatically by the system without human intervention in each scenario, the number of decisions made manually in each scenario, the standard deviation of the time taken for all automated decisions in each scenario, and the arithmetic mean of the confidence scores of all automated decision results. The decision time refers to the time interval from triggering a decision to generating a result, and the confidence score is obtained by normalizing the probability value or score output by the decision model to the range of 0 to 1.

7. The intelligent analysis method for enterprise digital transformation based on big data analysis according to claim 1, characterized in that, Step four, the similarity matching, uses weighted Euclidean distance to calculate the similarity between the current enterprise feature profile and the historical enterprise feature vectors. The weights of each feature dimension are determined by the information gain obtained from training a random forest model on the company's own enterprise database. The top-ranked similarity profiles are then selected. These companies serve as a reference sample set. Take an integer between 5 and 20.

8. The intelligent analysis method for enterprise digital transformation based on big data analysis according to claim 1, characterized in that, The digital transformation stages described in step five are divided into five levels: initial level, standardization level, integration level, optimization level, and leading level. The weakness identification is achieved by calculating the difference between the current enterprise's characteristic dimensions and the corresponding characteristic dimensions of benchmark enterprises in the next transformation stage in the reference sample set, selecting the top three dimensions with the largest differences as the main weaknesses, and generating improvement suggestions based on the transformation path data of similar enterprises in the historical database.

9. The intelligent analysis method for enterprise digital transformation based on big data analysis according to claim 5, characterized in that, Data flow barrier coefficient Calculated using the following formula: ； in This indicates the total number of departments within the company; Indicates the first The number of times each department's internal employees access their department's data; Indicates the first Total number of data accesses by all employees in each department; Indicates the first Organizational hierarchy coding for each department; Indicates the first Organizational hierarchy coding for each department; This is a hierarchical normalization constant, and its value represents the maximum span of the enterprise's management levels. It is a natural constant; This indicates the range from the 1st to the 2nd. Each department seeks a summation; Indicates that except for the first Summation is performed on all departments except for the one mentioned department; Indicates the first The department and the first The absolute value of the difference in departmental level codes.

10. The intelligent analysis method for enterprise digital transformation based on big data analysis according to claim 6, characterized in that, The decision-making intelligence index Calculated using the following formula: ； in, This represents the total number of types of business decision-making scenarios for an enterprise; Indicates the first The number of decisions made automatically by the system in decision-making scenarios without human intervention; Indicates the first The number of decisions that require human intervention in decision-making scenarios; Indicates the first The standard deviation of the time spent on all automated decisions in similar decision-making scenarios; This is a time normalization constant, and its value is the average of the time consumed by automated decision-making in all decision-making scenarios. This represents the arithmetic mean of the confidence levels of all automated decision outcomes. This represents the maximum confidence level, with a value of 1. It is a natural constant; Indicates the range from category 1 to category 2. Summation is performed in decision-making scenarios.

Images