Digital intelligent management accounting system and method based on big data

Abstract

The invention relates to the technical field of accounting management, in particular to a digital intelligent management accounting system and method based on big data, and the system comprises a data perception layer, an intelligent analysis layer, a decision optimization layer and a dynamic feedback layer which are connected in sequence, and all layers are interacted through a distributed service bus to form a closed loop; the data sensing layer collects and preprocesses multi-source heterogeneous data; the intelligent analysis layer analyzes and models by means of a dynamic parameter adjustment algorithm, a multi-dimensional cost correlation model, an LSTM and XGBoost hybrid model and the like; the decision optimization layer outputs a strategy and realizes resource matching by using an improved genetic algorithm; and the dynamic feedback layer monitors deviation and feeds back optimization. According to the method, data islands are broken, the precision of budget, cost and risk analysis is improved, management accounting is transformed from post-event accounting to pre-event prediction and in-event control, and enterprises are assisted to improve decision efficiency and core competitiveness.

Description

Technical Field

[0001] This invention relates to the field of accounting management technology, specifically to a digital and intelligent management accounting system and method based on big data. Background Technology

[0002] With the deepening development of the digital economy era, the business management environment is undergoing profound changes. Especially in traditional industries such as garment manufacturing, market competition is becoming increasingly fierce, product life cycles are shortening, and consumer demands are becoming increasingly personalized. This places higher demands on the management accounting system of enterprises. Traditional management accounting mainly focuses on ex-post accounting and static analysis, which is difficult to adapt to the rapidly changing market environment. Against this backdrop, leveraging big data, artificial intelligence, the Internet of Things, and other digital technologies to empower management accounting and achieve the transformation from accounting-oriented to decision support-oriented has become an inevitable choice for enterprises to enhance their core competitiveness.

[0003] Currently, some information systems are already being applied to financial management in enterprises. For example, some companies use ERP systems for financial data management, independent budgeting software for financial budgeting, and cost accounting modules for cost calculation; others use business intelligence tools for financial data analysis or independent risk assessment systems for financial risk monitoring. These existing technological solutions have improved the efficiency of enterprise financial management to some extent, but they still have significant limitations.

[0004] However, existing management accounting solutions suffer from the following major drawbacks: First, the systems are independent of each other, forming data silos, and there is a lack of effective coordination in budget management, cost control, risk warning, and performance evaluation, making it difficult to form a unified management view. Second, existing solutions rely heavily on static historical data and simple rules, lacking the ability to collect and integrate multi-source heterogeneous data in real time, resulting in insufficient forecast accuracy and lagging decision support. Third, traditional methods struggle to establish effective closed-loop feedback mechanisms, making it impossible to dynamically adjust budget strategies and optimize resource allocation based on implementation, thus failing to fully realize the proactive value creation function of management accounting. These shortcomings severely restrict enterprises' adaptability and decision-making capabilities in complex and ever-changing market environments. Summary of the Invention

[0005] In view of the above-mentioned shortcomings of the existing technology, the present invention provides a big data-based intelligent management accounting system and method, which can effectively solve the problems mentioned in the existing technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] This invention provides a big data-based intelligent management accounting system, comprising a data perception layer, an intelligent analysis layer, a decision optimization layer, and a dynamic feedback layer connected in sequence. Each layer interacts with data through a distributed service bus to form a closed-loop control system.

[0008] The data sensing layer is used to collect and preprocess multi-source heterogeneous data, including:

[0009] The multi-source data acquisition module collects enterprise business data, financial data, market data, and industry dynamic data in real time through API interface unit, IoT device access unit, and unstructured data parsing unit;

[0010] The data preprocessing module uses a data cleaning engine to standardize the collected data and stores the processed data in a distributed database.

[0011] The intelligent analysis layer, connected to the data perception layer, is used to analyze and model the preprocessed data, including:

[0012] The budget analysis module is used to generate rolling budget plans through a dynamic parameter adjustment algorithm;

[0013] The cost analysis module is used to calculate the cost elasticity coefficient based on a multidimensional cost correlation model and to construct a cost transmission matrix through a graph neural network.

[0014] The risk analysis module is used to perform risk warning using a hybrid model that combines LSTM time series prediction and XGBoost classification, and to calculate the conditional risk probability through a Bayesian network.

[0015] The decision optimization layer, connected to the intelligent analysis layer, is used to generate management decisions based on the analysis results, including:

[0016] The strategy generation module is used to output cost optimization strategies and resource allocation schemes;

[0017] The resource allocation module is used to achieve dynamic matching of resources through an improved genetic algorithm;

[0018] The dynamic feedback layer is connected to the data perception layer, the intelligent analysis layer, and the decision optimization layer, respectively, and includes:

[0019] The monitoring module is used to collect decision execution data in real time and calculate execution deviations;

[0020] The adjustment module is used to feed back execution deviations to the corresponding level for parameter adjustment and model optimization;

[0021] The budget analysis module generates a rolling budget plan through a dynamic parameter adjustment algorithm, the formula of which is:

[0022] ;

[0023] in, This is the budget value for the current period. This is the budget value from the previous period. For income fluctuations, Based on the benchmark income, Cost fluctuation amount Based on the baseline cost, , These are the weighting coefficients.

[0024] Furthermore, the multi-source data acquisition module includes:

[0025] The API interface unit connects to enterprise ERP systems and external market data platforms via the RESTful protocol.

[0026] The IoT device access unit collects real-time operating status data of production and logistics equipment.

[0027] The unstructured data parsing unit uses natural language processing technology to parse text-based industry data and public opinion data.

[0028] Furthermore, the formula for calculating the cost elasticity coefficient by the cost analysis module is as follows:

[0029] ;

[0030] in, Based on the baseline cost, For business parameters, The sensitivity of costs to business parameters;

[0031] The cost transmission matrix The update formula for the graph neural network is:

[0032] ;

[0033] in, For adjacency weight, For bias, This is the activation function.

[0034] Furthermore, in the risk analysis module, the fusion formula for the hybrid model is:

[0035] ;

[0036] Where pi is the predicted probability of the i-th sub-model. These are the weighting coefficients;

[0037] The formula for calculating the conditional risk probability using the Bayesian network is as follows:

[0038] ;

[0039] Where Y represents the risk event and X represents the observed data.

[0040] Furthermore, in the resource allocation module, the fitness function of the improved genetic algorithm is:

[0041] ;

[0042] in, As a result of resource allocation, For business target value, These are the weighting coefficients.

[0043] Furthermore, in the dynamic feedback layer, the formula for calculating the execution deviation is:

[0044] ;

[0045] in, This is the budget deviation. For the j-th influencing factor, Let be the change in the j-th influencing factor.

[0046] Furthermore, the standard processing of the data preprocessing module includes outlier removal, missing value imputation, data type conversion, and data anonymization.

[0047] The data-driven intelligent management accounting method includes the following steps:

[0048] S1. Data Acquisition and Preprocessing Steps: Real-time acquisition of business, financial, market and industry dynamic data through the data perception layer, standardized processing using the data cleaning engine and then stored in the distributed database;

[0049] S2, Intelligent Analysis Steps:

[0050] S21. Budget Analysis: Generate a rolling budget plan through a dynamic parameter adjustment algorithm, and assess the budget execution risk through Monte Carlo simulation;

[0051] S22. Cost Analysis: Calculate the cost elasticity coefficient based on a multidimensional cost correlation model, and construct and update the cost transmission matrix through a graph neural network;

[0052] S23. Risk Analysis: A hybrid model of LSTM and XGBoost is used for risk warning, and the probability of conditional risk is calculated through a Bayesian network.

[0053] S3, Decision Optimization Steps:

[0054] S31. Output cost optimization strategies and resource allocation schemes through the strategy generation engine;

[0055] S32. Use an improved genetic algorithm for dynamic resource matching and scheme optimization;

[0056] S4. Dynamic Feedback Steps:

[0057] S41. Collect execution data and calculate execution deviation through the real-time monitoring module;

[0058] S42. Locate the driving factors of the deviation through causal inference technology and generate an optimization suggestion report;

[0059] S43. Feed the analysis results back to the data perception layer to form a closed-loop control.

[0060] The technical solution provided by this invention has the following advantages compared with the known prior art:

[0061] This invention constructs a closed-loop control system comprising a data perception layer, an intelligent analysis layer, a decision optimization layer, and a dynamic feedback layer. It organically integrates core functions such as budget management, cost control, risk warning, and resource allocation, achieving real-time connectivity of business and financial data and collaborative linkage of intelligent subsystems. This overcomes the drawbacks of fragmented systems and data in existing technologies, providing a unified and comprehensive data view for enterprise management decisions.

[0062] This invention introduces advanced intelligent analysis methods such as dynamic parameter adjustment algorithms, LSTM and XGBoost hybrid models, and graph neural networks. It can dynamically predict and simulate optimization of budgets, costs, and risks based on multi-source data from internal and external sources. By leveraging a dynamic feedback layer, it forms a closed-loop control of "decision-execution-feedback-optimization," enabling management accounting to shift from ex-post accounting to ex-ante prediction and in-process control. This significantly improves the company's response speed to market changes and the accuracy of management strategies. Attached Figure Description

[0063] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0064] Figure 1 This is a schematic diagram of the system structure of the present invention;

[0065] Figure 2 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0067] The present invention will be further described below with reference to embodiments.

[0068] Example:

[0069] Reference Figure 1 The big data-based intelligent management accounting system includes a data perception layer, an intelligent analysis layer, a decision optimization layer, and a dynamic feedback layer connected in sequence. Each layer interacts with data through a distributed service bus to form a closed-loop control system.

[0070] The data awareness layer is used to collect and preprocess multi-source heterogeneous data, including:

[0071] The multi-source data acquisition module collects enterprise business data, financial data, market data, and industry dynamic data in real time through API interface unit, IoT device access unit, and unstructured data parsing unit;

[0072] The data preprocessing module uses a data cleaning engine to standardize the collected data and stores the processed data in a distributed database.

[0073] In a specific embodiment, taking a garment manufacturing enterprise as an example, the multi-source data acquisition module connects to the enterprise's ERP system, external e-commerce platforms, and industry data analysis platforms through an API interface unit to acquire order data, financial voucher data, and market sales data in real time. The IoT device access unit is deployed on sewing machines in the production workshop and AGV equipment in logistics warehousing to collect data such as equipment runtime, energy consumption, and production capacity. The unstructured data parsing unit uses natural language processing technology to parse text data such as industry reports and social media sentiment. The data preprocessing module uses a cleaning engine to remove abnormal order data, fills in missing equipment energy consumption values, converts financial data in different formats into a unified JSON format, and anonymizes customer privacy information before finally storing it in a distributed database.

[0074] It enables real-time aggregation and standardized processing of heterogeneous data from multiple sources both inside and outside the enterprise, breaking down information barriers between the original ERP system, equipment data, and external market data. Enterprises can leverage this layer to achieve real-time integration of order, production, and financial data, reducing data processing latency from 24 hours to within one hour, ensuring that subsequent analysis and decision-making processes are based on timely and accurate data.

[0075] The system employs a multi-module collaborative data acquisition approach, covering all dimensions of data, including business, finance, and marketing. Compared to traditional single-data acquisition models, it offers broader data coverage and faster timeliness. The data preprocessing workflow is complete, including key steps such as outlier removal and missing value imputation, while also incorporating data anonymization to ensure both data quality and information security. The application of a distributed database provides stable support for massive data storage and possesses excellent scalability, adapting to the growing data volume needs of enterprises.

[0076] The intelligent analysis layer, connected to the data perception layer, is used to analyze and model the preprocessed data, including:

[0077] The budget analysis module is used to generate rolling budget plans through a dynamic parameter adjustment algorithm;

[0078] The cost analysis module is used to calculate the cost elasticity coefficient based on a multidimensional cost correlation model and to construct a cost transmission matrix through a graph neural network.

[0079] The risk analysis module is used to perform risk warning using a hybrid model that combines LSTM time series prediction and XGBoost classification, and to calculate the conditional risk probability through a Bayesian network.

[0080] In one specific embodiment, the budget analysis module displays the previous period's monthly budget value for Company A. For 1 million yuan, the monthly income fluctuation amount The benchmark income is 50,000 yuan. The cost fluctuation is 1 million yuan. The base cost is 30,000 yuan. It is 800,000 yuan, with a weighting coefficient. Set to 0.6. Set it to 0.4, substitute it into the dynamic parameter adjustment algorithm formula, and calculate the current period budget. The budget is set at 1.045 million yuan, enabling dynamic budget adjustments. The budget analysis module employs a dynamic parameter adjustment algorithm, abandoning the traditional static budget preparation method. It can adjust the budget in real time based on changes in business data, offering greater flexibility.

[0081] Cost analysis module: uses production man-hours as business parameters Benchmark cost The cost elasticity coefficient is calculated based on a multidimensional cost correlation model, with a value of 800,000 yuan. A value of 0.8 indicates that for every 10% change in production hours, the cost changes by 8%. A 3×3 cost transmission matrix M is constructed using a graph neural network, with adjacency weights... Based on the correlation strength settings of fabric procurement, production and processing, logistics and transportation, the ReLU activation function and bias are applied. After adjustment, the updated matrix accurately depicts the cost transmission path. The cost analysis module integrates a multi-dimensional cost correlation model and a graph neural network, which can not only calculate the sensitivity of costs to business parameters, but also clearly present the cost transmission relationship. Compared with traditional cost analysis methods, it has a more comprehensive analysis dimension and greater depth.

[0082] Risk Analysis Module: Employs a hybrid LSTM and XGBoost model for early warning of risks such as raw material price fluctuations and inventory backlog. The two sub-models predict probabilities of [missing information]. , Weighting coefficient , The final risk probability after fusion The system calculates the conditional risk probability of "rising raw material prices" using a Bayesian network. Given observed data such as rising international cotton futures prices, the conditional risk probability is calculated to be 0.82, achieving accurate risk assessment. The risk analysis module employs a hybrid LSTM and XGBoost model, combining the advantages of both algorithms to improve prediction accuracy. The introduction of a Bayesian network to calculate the conditional risk probability makes risk assessment more targeted, providing more precise decision support for enterprise risk management.

[0083] The decision optimization layer, connected to the intelligent analysis layer, is used to generate management decisions based on the analysis results, including:

[0084] The strategy generation module is used to output cost optimization strategies and resource allocation schemes;

[0085] The resource allocation module is used to achieve dynamic matching of resources through an improved genetic algorithm;

[0086] In one specific embodiment, the strategy generation module, based on intelligent analysis results, outputs cost optimization strategies for company A, such as changing the fabric procurement channel from a single supplier to multi-supplier bidding, aiming to reduce procurement costs by 10%. Simultaneously, it outputs resource allocation plans, such as allocating production capacity to the production line of popular apparel items. The resource allocation module achieves dynamic matching through an improved genetic algorithm, setting a fitness function F, and using the output of each production line as the resource allocation result. Business target value Forecast sales volume, weighting coefficient Based on product profit margins, and after algorithmic optimization, 80% of production resources are allocated to the top three profit-margin products. This drives a shift in corporate decision-making from experience-driven to data-driven, optimizing resource allocation efficiency and cost control. After implementing the output cost optimization strategy, Company A experienced a decrease in procurement costs. With resources shifted towards high-profit products, resource utilization increased by 20%, overall product profitability significantly improved, core product market supply capacity was enhanced, and market share was further consolidated.

[0087] The strategy generation module can output specific and implementable cost optimization and resource allocation solutions based on the accurate data results of the intelligent analysis layer. Compared with traditional fuzzy decision suggestions, it is more practical. The resource allocation module adopts an improved genetic algorithm, which accurately measures the matching degree between resource allocation and business objectives through a fitness function. Through iterative optimization, it can find the optimal resource allocation solution, avoiding the resource waste problem caused by relying on subjective judgment in traditional resource allocation, and making resource allocation more reasonable and efficient.

[0088] The dynamic feedback layer, connected to the data perception layer, intelligent analysis layer, and decision optimization layer, includes:

[0089] The monitoring module is used to collect decision execution data in real time and calculate execution deviations;

[0090] The adjustment module is used to feed back execution deviations to the corresponding level for parameter adjustment and model optimization;

[0091] In one specific embodiment, after Company A implements a cost optimization strategy, the monitoring module collects real-time data and finds that the actual procurement cost has decreased by 8%, which deviates from the target value of 10%. Substituting this into the execution deviation formula, the influencing factors are... The budget deviation is calculated by taking into account factors such as fluctuations in supplier quotes and changes in purchase volume. The value is 20,000 yuan; the adjustment module uses causal inference technology to locate the deviation driving factor as a core supplier not participating in the bidding, generates an optimization suggestion report to add two high-quality fabric suppliers, and feeds the results back to the data perception layer and intelligent analysis layer to adjust the procurement data collection dimensions and budget algorithm weights to form a closed-loop optimization.

[0092] The monitoring module collects execution data and calculates deviations in real time, enabling more timely detection of problems compared to the traditional periodic review model. The adjustment module introduces causal inference technology, which not only detects deviations but also accurately locates their root causes, avoiding the problem of only knowing the deviations in traditional feedback but not being able to effectively trace the causes. At the same time, the analysis results are fed back to the preceding levels to achieve full-process optimization, forming a complete closed loop, allowing the management accounting system to continuously improve itself and fully exert its proactive value creation function.

[0093] The budget analysis module generates a rolling budget plan through a dynamic parameter adjustment algorithm. The algorithm formula is as follows:

[0094] ;

[0095] in, This is the budget value for the current period. This is the budget value from the previous period. For income fluctuations, Based on the benchmark income, Cost fluctuation amount Based on the baseline cost, , These are the weighting coefficients.

[0096] Furthermore, the multi-source data acquisition module includes:

[0097] The API interface unit connects to enterprise ERP systems and external market data platforms via the RESTful protocol.

[0098] The IoT device access unit collects real-time operating status data of production and logistics equipment.

[0099] The unstructured data parsing unit uses natural language processing technology to parse text-based industry data and public opinion data.

[0100] Furthermore, the formula for calculating the cost elasticity coefficient in the cost analysis module is as follows:

[0101] ;

[0102] in, Based on the baseline cost, For business parameters, The sensitivity of costs to business parameters;

[0103] Cost transmission matrix The update formula for the graph neural network is:

[0104] ;

[0105] in, For adjacency weight, For bias, This is the activation function.

[0106] Furthermore, in the risk analysis module, the fusion formula for the hybrid model is:

[0107] ;

[0108] Where pi is the predicted probability of the i-th sub-model. These are the weighting coefficients;

[0109] The formula for calculating the conditional risk probability using a Bayesian network is:

[0110] ;

[0111] Where Y represents the risk event and X represents the observed data.

[0112] Furthermore, in the resource allocation module, the fitness function of the improved genetic algorithm is:

[0113] ;

[0114] in, As a result of resource allocation, For business target value, These are the weighting coefficients.

[0115] Furthermore, in the dynamic feedback layer, the formula for calculating the execution deviation is:

[0116] ;

[0117] in, This is the budget deviation. For the j-th influencing factor, Let be the change in the j-th influencing factor.

[0118] Furthermore, the standard processing steps in the data preprocessing module include outlier removal, missing value imputation, data type conversion, and data anonymization.

[0119] Reference Figure 2 The data-driven intelligent management accounting method includes the following steps:

[0120] S1. Data Acquisition and Preprocessing Steps: Real-time acquisition of business, financial, market and industry dynamic data through the data perception layer, standardized processing using the data cleaning engine and then stored in the distributed database;

[0121] S2, Intelligent Analysis Steps:

[0122] S21. Budget Analysis: Generate a rolling budget plan through a dynamic parameter adjustment algorithm, and assess the budget execution risk through Monte Carlo simulation;

[0123] S22. Cost Analysis: Calculate the cost elasticity coefficient based on a multidimensional cost correlation model, and construct and update the cost transmission matrix through a graph neural network;

[0124] S23. Risk Analysis: A hybrid model of LSTM and XGBoost is used for risk warning, and the probability of conditional risk is calculated through a Bayesian network.

[0125] S3, Decision Optimization Steps:

[0126] S31. Output cost optimization strategies and resource allocation schemes through the strategy generation engine;

[0127] S32. Use an improved genetic algorithm for dynamic resource matching and scheme optimization;

[0128] S4. Dynamic Feedback Steps:

[0129] S41. Collect execution data and calculate execution deviation through the real-time monitoring module;

[0130] S42. Locate the driving factors of the deviation through causal inference technology and generate an optimization suggestion report;

[0131] S43. Feed the analysis results back to the data perception layer to form a closed-loop control.

[0132] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the protection scope of the technical solutions of the embodiments of the present invention.

Claims

1. A big data-based digital management accounting system, characterized in that, Comprise data perception layer, intelligent analysis layer, decision optimization layer and dynamic feedback layer connected in turn, each layer carries out data interaction through distributed service bus, forms closed loop control system; The data perception layer is used for collecting and preprocessing multi-source heterogeneous data, and comprises: A multi-source data collection module collects business data, financial data, market data and industry dynamic data of an enterprise in real time through an API interface unit, an IoT device access unit and an unstructured data analysis unit; A data preprocessing module performs standardization processing on the collected data through a data cleaning engine, and stores the processed data into a distributed database; The intelligent analysis layer is connected with the data perception layer, and is used for analyzing and modeling the preprocessed data, and comprises: A budget analysis module is used for generating a rolling budget scheme through a dynamic parameter adjustment algorithm; A cost analysis module is used for calculating a cost elasticity coefficient based on a multi-dimensional cost correlation model, and constructing a cost transmission matrix through a graph neural network; A risk analysis module is used for performing risk early warning through a hybrid model combining LSTM time series prediction and XGBoost classification, and calculating a conditional risk probability through a Bayesian network; The decision optimization layer is connected with the intelligent analysis layer, and is used for generating a management decision based on an analysis result, and comprises: A strategy generation module is used for outputting a cost optimization strategy and a resource allocation scheme; A resource allocation module is used for realizing dynamic matching of resources through an improved genetic algorithm; The dynamic feedback layer is connected with the data perception layer, the intelligent analysis layer and the decision optimization layer respectively, and comprises: A monitoring module is used for collecting decision execution data in real time and calculating execution deviation; An adjustment module is used for feeding back the execution deviation to a corresponding layer for parameter adjustment and model optimization; The budget analysis module generates a rolling budget scheme through a dynamic parameter adjustment algorithm, and the algorithm formula is: ； wherein, is a current period budget value, is a previous period budget value, is an income fluctuation amount, is a reference income, is a cost fluctuation amount, is a reference cost, , is a weight coefficient.

2. The big data-based digitalized management accounting system according to claim 1, characterized in that, The multi-source data collection module comprises: An API interface unit is connected with an enterprise ERP system and an external market data platform through a RESTful protocol; An IoT device access unit collects running state data of production equipment and logistics equipment in real time; An unstructured data analysis unit analyzes text type industry data and public opinion data through natural language processing technology.

3. The big data-based digitalized management accounting system according to claim 1, characterized in that, The cost analysis module calculates the cost elasticity coefficient according to the formula: ； wherein, is a reference cost, is a business parameter, is a sensitivity of the cost to the business parameter; The cost propagation matrix The graph neural network update formula is: ； wherein, is an adjacency weight, is a bias, is an activation function.

4. The big data-based digitalized management accounting system of claim 1, wherein, In the risk analysis module, the fusion formula of the hybrid model is: ； wherein pi is the prediction probability of the ith sub-model, is a weight coefficient; The formula for calculating the conditional risk probability of the Bayesian network is: ； Wherein, Y is a risk event, and X is observation data.

5. The big data-based digitalized management accounting system according to claim 1, characterized in that, In the resource allocation module, the fitness function of the improved genetic algorithm is: ； wherein, is a resource allocation result, is a service target value, is a weight coefficient.

6. The big data-based digitalized management accounting system according to claim 1, characterized in that, In the dynamic feedback layer, the formula for calculating the execution deviation is: ； wherein, is a budget deviation amount, is the jth influencing factor, is a change amount of the jth influencing factor.

7. The big data-based digitalized management accounting system according to claim 1, characterized in that, The standardization processing of the data preprocessing module includes outlier rejection, missing value filling, data type conversion and data desensitization.

8. A digital management accounting method based on big data, characterized in that, The method comprises the following steps: S1, data collection and preprocessing step: collecting business, financial, market and industry dynamic data in real time through the data perception layer, and storing the data into a distributed database after standardization processing by a data cleaning engine; S2, intelligent analysis step: S21, budget analysis: generate rolling budget plan through dynamic parameter adjustment algorithm, and evaluate budget execution risk through Monte Carlo simulation; S22, cost analysis: calculate cost elasticity coefficient based on multi-dimensional cost correlation model, and construct and update cost transmission matrix through graph neural network; S23, risk analysis: use LSTM and XGBoost hybrid model for risk early warning, and calculate conditional risk probability through Bayesian network; S3, decision optimization step: S31, output cost optimization strategy and resource allocation scheme through strategy generation engine; S32, use improved genetic algorithm for resource dynamic matching and scheme optimization; S4, dynamic feedback step: S41, collect execution data through real-time monitoring module and calculate execution deviation; S42, locate deviation driving factors through causal inference technology, and generate optimization suggestion report; S43, feedback the analysis results to the data perception layer to form a closed-loop control.

Images