Operation cost prediction method and system based on artificial intelligence

Abstract

The invention relates to the technical field of cost prediction, in particular to an operation cost prediction method and system based on artificial intelligence, and the method comprises the following steps: collecting and sorting human resource cost, material cost and energy use cost, creating an independent data set for each cost category, representing nodes through different cost categories, and predicting the cost of each cost category; edges are represented by interaction among cost categories, node weights are determined through cost influence analysis, and a cost factor relation graph is constructed. According to the method, the adaptability and accuracy of cost prediction are effectively improved by constructing the cost factor relation graph and the dependency relation graph and introducing dynamic data processing and time sequence analysis, node data are allowed to be updated in real time to reflect changes of market and operation conditions, and the accuracy of cost prediction is improved through quantitative cost deviation analysis. The difference between prediction and reality can be accurately determined, cost control is finer through screening and trend analysis of key nodes, and resource allocation optimization and budget making in a complex market environment are helped.

Description

Technical Field

[0001] The present invention relates to the technical field of cost forecasting, and in particular to an operation cost forecasting method and system based on artificial intelligence. Background Art

[0002] The core content of the field of cost forecasting technology involves the use of historical data and models to estimate cost changes in future periods. It is widely used in many fields such as financial management, project management, and resource allocation. It aims to provide companies with more accurate budget formulation and cost control methods, including but not limited to cost analysis, cost allocation, and cost optimization, all of which are achieved through quantitative analysis and related forecasting models.

[0003] Among them, the AI-based operating cost forecasting method refers to the use of AI technology to predict the costs in the process of enterprise operations, and the technical matters targeted include data collection, data analysis, and the construction of cost forecasting models. The specific method uses AI algorithms to analyze historical cost data to identify cost trends and driving factors, thereby building a model to predict future costs.

[0004] Traditional cost forecasting methods rely on historical data and use static models, which often lags behind when facing rapid market changes. The lack of rapid response to real-time market changes means that forecast results often fail to accurately reflect current economic conditions. For example, when resource prices fluctuate rapidly, traditional models fail to adjust in a timely manner, often resulting in large deviations between actual budgets and forecast values. In addition, existing technologies fail to deeply analyze the dependencies between cost factors and lack the identification of key cost drivers, which limits the potential for cost optimization and leads to inaccurate cost control, increasing operational financial risks. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an operation cost prediction method based on artificial intelligence.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: an operation cost prediction method based on artificial intelligence, comprising the following steps:

[0007] S1: Collect and organize human resource costs, material costs, and energy usage costs, create independent data sets for each cost category, represent nodes with differential cost categories, represent edges with interactions between cost categories, determine node weights through cost influence analysis, and construct a cost factor relationship diagram;

[0008] S2: Based on the cost factor relationship graph, the dependencies between nodes are analyzed, the node data is updated to map the changes in market and operating conditions, and the node relationship strength is adjusted to reflect the actual cost dynamics, thereby obtaining an updated dependency relationship graph;

[0009] S3: using the updated dependency graph, calculating the influence score of each node, selecting nodes with scores higher than the threshold as key nodes according to the set threshold, and performing time series analysis on the key nodes. Based on the time series analysis results and in combination with historical cost data, linear trend analysis is applied to predict future cost fluctuations, capturing future change trends of key cost factors, and comprehensively generating trend prediction analysis results;

[0010] S4: Based on the trend forecast analysis results, combined with actual cost data, cost deviation analysis is performed to determine the core difference area between the predicted value and the actual value, and the results are determined based on the core difference area. The evaluation criteria for cost impact are adjusted, and the forecast error is optimized through iterative data comparison to generate verified cost forecast results.

[0011] As a further solution of the present invention, the steps of constructing the cost factor relationship diagram are:

[0012] S111: Collect human resource costs, material costs and energy usage costs, classify and sort the data according to cost types, and obtain cost classification records;

[0013] S112: Based on the cost classification records, the difference cost category data set is used as a node, and the edge relationship is established by utilizing the interaction between the cost categories, using the formula:

[0014]

[0015] Calculate the correlation weight W between cost categories ij , generate the preliminary structure of the relationship graph, where C ik and C jk denote the costs of cost categories i and j on item k respectively;

[0016] S113: Perform influence analysis based on the preliminary structure of the relationship graph, determine node weights, reflect the influence and relevance of the differential cost categories, and generate a cost factor relationship graph.

[0017] As a further solution of the present invention, the step of obtaining the updated dependency graph is:

[0018] S211: According to the cost factor relationship diagram, combined with the sensitivity of node cost changes, by analyzing the response of the difference nodes to the cost changes, the dependency between each node is calculated item by item to obtain the node dependency analysis result;

[0019] S212: using the inter-node dependency analysis results, combined with the latest change data of market and operating conditions, updating the data mapping node by node, revealing the changes of current market and operating conditions, and generating adjusted node data mapping results;

[0020] S213: Reconfigure the node connection relationship strength in the cost factor relationship graph according to the adjusted node data mapping result, and modify the connection weight in a digital manner, using the formula:

[0021]

[0022] Calculate the new connection strength C between nodes i and j ij , generate an updated dependency graph, where N ij is the dependency analysis value between nodes i and j, ΔO i represents the change in the operating condition of node i, V i is the market change at node i.

[0023] As a further solution of the present invention, the analysis steps of the key node time series are:

[0024] S311: Based on the updated dependency graph, use the formula:

[0025]

[0026] Calculate the influence score of each node i i , generating a node influence score, where N(i) represents the set of nodes directly connected to node i, and w ij is the weight connecting i and j, d ij is the distance between nodes;

[0027] S312: According to the node influence score, combined with the screening threshold T, each node influence score is screened and a score Score is selected. i For nodes higher than T, a list of key nodes is generated;

[0028] S313: Apply time series analysis to the key node list to reveal the behavior of the nodes in the time dimension, analyze the change trend of the nodes in the time dimension, and obtain the key node time series analysis results.

[0029] As a further solution of the present invention, the steps of obtaining the trend prediction analysis results are:

[0030] S321: Collect and integrate the time series analysis results and historical cost data of the key nodes, record the operation cost, maintenance cost, and traffic and connection number behavior data of the nodes, verify the integrity and consistency of the data, and obtain integrated data;

[0031] S322: Perform trend analysis on the integrated data, check the relationship between cost and time, determine the fluctuation pattern of historical data, identify key driving factors and potential trend lines of cost changes, and output preliminary trend identification results;

[0032] S323: Based on the preliminary trend identification results, organize and display trend data and perform visualization to predict potential fluctuations in future costs and analyze future change trends of key cost factors to obtain trend forecast analysis results.

[0033] As a further solution of the present invention, the step of determining the core difference area is:

[0034] S411: extracting actual cost data of the current period from a known cost database according to the trend forecast analysis result, performing preliminary cost comparison, and generating preliminary comparison results;

[0035] S412: Analyze each cost data in the preliminary comparison result, using the formula:

[0036]

[0037] Calculate the percentage deviation Δp between the predicted value and the actual value i , get the deviation analysis report, where P i represents the prediction cost of the ith item, A i represents the actual cost of item i;

[0038] S413: Based on the deviation analysis report, cost items whose deviations exceed a set threshold are screened and marked to generate core difference analysis results.

[0039] As a further solution of the present invention, the steps for obtaining the verified cost prediction result are:

[0040] S421: Based on the core difference area data, all cost items exceeding a predetermined threshold are recorded, along with the amount, time, and relevant responsible departments, to form a difference data report;

[0041] S422: Based on the difference data report, re-examine and adjust the cost evaluation standard to reflect the fluctuation and change trend of the actual cost data, and update the evaluation standard in combination with the deviation between the cost data and the forecast to obtain the adjusted evaluation standard;

[0042] S423: Re-analyze the historical data using the adjusted evaluation criteria, compare the cost forecasts and actual data before and after the adjustment, identify and verify the improvement effect of the evaluation criteria, and generate verified cost forecast results through iterative comparison.

[0043] An operation cost prediction system based on artificial intelligence, comprising:

[0044] The data classification module collects human resource costs, material costs, and energy usage costs, classifies and organizes them, uses differential cost category data sets as nodes, uses the interactions between cost categories to establish edge relationships, calculates the correlation weights between cost categories, determines node weights, reflects the influence and relevance of differential cost categories, and generates a cost factor relationship diagram;

[0045] The dependency updating module calculates the dependency between each node item by item according to the cost factor relationship graph by analyzing the response of the difference node to the cost change, updates the data mapping of each node node by node in combination with the latest change data, reconfigures the node connection relationship strength in the cost factor relationship graph, and generates an updated dependency relationship graph;

[0046] The trend forecasting and analysis module calculates the influence score of each node based on the updated dependency graph, and screens the influence score of each node to determine the key nodes in combination with the screening threshold, and performs time series analysis, and checks the relationship between cost and time in combination with historical cost data, predicts potential fluctuations in future costs, and analyzes future change trends of key cost factors to obtain trend forecasting and analysis results;

[0047] Based on the trend forecast analysis results, the cost assessment and adjustment module extracts the actual cost data of the current period from the known cost database and performs a preliminary cost comparison, calculates the deviation percentage between the predicted value and the actual value, screens out cost items whose deviations exceed the set threshold, records all cost items that exceed the predetermined threshold, reviews and adjusts the cost assessment standards, compares the cost forecasts and actual data before and after the adjustment, identifies and verifies the improvement effects of the assessment standards, and generates verified cost forecast results through iterative comparison.

[0048] Compared with the prior art, the advantages and positive effects of the present invention are:

[0049] In the present invention, by constructing a cost factor relationship diagram and a dependency diagram, dynamic data processing and time series analysis are introduced to effectively improve the adaptability and accuracy of cost forecasting, allowing real-time updating of node data to reflect changes in market and operating conditions. Through quantitative cost deviation analysis, the difference between prediction and reality can be accurately determined. The screening and trend analysis of key nodes make cost control more refined, helping to optimize resource allocation and budget formulation in a complex market environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a main step flow chart of the present invention;

[0051] Figure 2 A flow chart for constructing a cost factor relationship diagram of the present invention;

[0052] Figure 3 A flowchart for obtaining the dependency graph updated by the present invention;

[0053] Figure 4 It is the analysis flow chart of the key node time series of the present invention;

[0054] Figure 5 A flowchart for obtaining trend prediction analysis results of the present invention;

[0055] Figure 6 A flowchart for determining the core difference area of ​​the present invention;

[0056] Figure 7 This is a flow chart for obtaining the cost prediction results after verification of the present invention. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0058] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0059] See also Figure 1 , an operation cost prediction method based on artificial intelligence, comprising the following steps:

[0060] S1: Collect and organize human resource costs, material costs, and energy usage costs, create independent data sets for each cost category, represent nodes with differential cost categories, represent edges with interactions between cost categories, determine node weights through cost influence analysis, and construct a cost factor relationship diagram;

[0061] S2: Based on the cost factor relationship graph, the dependencies between nodes are analyzed, the node data is updated to map the changes in market and operating conditions, and the node relationship strength is adjusted to reflect the actual cost dynamics to obtain an updated dependency graph;

[0062] S3: Using the updated dependency graph, calculate the influence score of each node, select nodes with scores higher than the threshold as key nodes according to the set threshold, and perform time series analysis on the key nodes. Based on the time series analysis results and combined with historical cost data, apply linear trend analysis to predict future cost fluctuations, capture future change trends of key cost factors, and comprehensively generate trend forecast analysis results;

[0063] S4: Based on the trend forecast analysis results and combined with the actual cost data, cost deviation analysis is performed to determine the core difference area between the predicted value and the actual value. The results are determined based on the core difference area, and the evaluation criteria for cost impact are adjusted. The forecast error is optimized through iterative data comparison to generate verified cost forecast results.

[0064] The cost factor relationship diagram includes node identification results, relationship types and impact weights; the updated dependency diagram includes adjusted connection strength, real-time market data and cost dynamic representation records; the trend forecast analysis results include key cost forecast records, time dependency analysis results and cost fluctuation range; the verified cost forecast results include quantitative records of difference areas and adjustment feedback records.

[0065] See also Figure 2 , the steps to construct the cost factor relationship diagram are:

[0066] S111: Collect human resource costs, material costs and energy usage costs, classify and sort the data according to cost types, and obtain cost classification records;

[0067] First, it is necessary to collect cost data on human resources, materials, and energy usage from various departments of the company, and communicate with the financial managers of each department to ensure the completeness and accuracy of the data. By reviewing the financial reports and related cost records of the past year, the data is preliminarily classified, and then Excel tables are used for data entry and preliminary analysis to ensure that each cost is accurately recorded in the correct classification. The data includes but is not limited to employee wages, welfare expenditures, departmental material purchase costs, energy expenditures, etc. By using VBA scripts to automate data processing, further data cleaning and structuring are carried out, such as deleting duplicate records, filling missing values, and adjusting data formats to meet subsequent data analysis needs. After completion, the data set will be imported into data analysis software, such as PowerBI, to create a dynamic data dashboard to facilitate monitoring and analysis of different types of cost data, forming a cost classification record that can be intuitively displayed.

[0068] S112: Based on the cost classification records, the difference cost category data set is used as the node, and the edge relationship is established by using the interaction between cost categories, using the formula:

[0069]

[0070] Calculate the correlation weight W between cost categories ij , generate the preliminary structure of the relationship graph, where C ik and C jk denote the costs of cost categories i and j on item k respectively;

[0071] Assume there are three projects, of which the costs of the two cost categories A and B on these three projects are

[0072] C A =[100, 200, 300]

[0073] C B =[150, 250, 350]

[0074] To calculate the correlation weight between two cost categories, first calculate the two parts of the numerator and denominator.

[0075] Calculate the numerator:

[0076]

[0077] Compute the two parts of the denominator:

[0078]

[0079] Substituting the values ​​into the original formula to calculate W AB :

[0080]

[0081] The results show that the correlation between cost categories A and B is very high (0.997), close to 1, which means that the spending trends of these two cost categories on the project are very similar, which can reveal potential cost savings points, such as the possibility of joint procurement or shared resources to reduce overall costs.

[0082] S113: Based on the preliminary structure of the relationship graph, influence analysis is performed to determine the node weights, reflect the influence and relevance of the differential cost categories, and generate a cost factor relationship graph;

[0083] After obtaining the preliminary structure of the relationship graph, the influence analysis method is used to determine the weight of each cost category node. First, define an influence evaluation standard, such as calculating the centrality index of each node, which can show the importance of a certain cost category in the overall cost structure. Using the NetworkX library in the Python programming language to build and analyze network graphs, it can provide a wide range of algorithms to process the relationship between nodes and edges, including the calculation of various network centrality indicators such as degree centrality and closeness centrality. For each node, the direct connection strength with other nodes will be calculated, and the value will be integrated into a weight coefficient, which is obtained by iteratively calculating the standardized value of each node's influence. Each iteration will adjust the weight according to the previous result until a balance is reached, at which time the weight of each node will reflect its actual role and importance in the cost impact network. Finally, the calculation results will be converted into a visual chart, and Gephi software will be used for highly customized visual presentation, so that decision makers can see at a glance which cost categories have key influence, which is of great value for optimizing resource allocation and making strategic decisions.

[0084] See also Figure 3 , the steps to obtain the updated dependency graph are:

[0085] S211: According to the cost factor relationship diagram, combined with the sensitivity of node cost changes, by analyzing the response of the difference nodes to the cost changes, the dependency between each node is calculated item by item to obtain the node dependency analysis result;

[0086] Starting from the cost factor relationship diagram, through on-site cost data collection, including direct costs such as raw material costs, labor costs and indirect costs such as equipment depreciation and management fees, the collected data includes historical cost data and current real-time data. Next, using the moving average and exponential smoothing methods, the cost data of each node is analyzed in time series to determine the trend and cyclical changes of cost changes. Subsequently, the cost sensitivity analysis of each node will be carried out by calculating the impact of cost changes on node operations. This calculation involves the correlation analysis of the cost change rate and node output. The Pearson or Spearman correlation coefficient is used to evaluate the impact of cost changes on node efficiency. The sensitivity of each node to cost changes can be accurately identified. Sensitivity will be used as a key indicator to evaluate the dependency between nodes. The final results of the dependency analysis between nodes will be presented in the form of tables and graphs.

[0087] S212: using the inter-node dependency analysis results, combined with the latest change data of market and operating conditions, updating the data mapping node by node, revealing the changes of current market and operating conditions, and generating adjusted node data mapping results;

[0088] Inheriting the inter-node dependency analysis results obtained above, the dependency data is combined with the latest changes in market and operating conditions to update the node data. The specific operation includes collecting the latest market data and operating condition change information, such as market demand changes, raw material price fluctuations, policy and regulatory adjustments, etc. In the data update process of each node, the external changes are first quantified into computable indicators, such as converting price fluctuations into percentage changes, and converting policy adjustments into increases or decreases in operating costs. Subsequently, the quantified market and operating data will be combined with the node dependency results, and the adjusted node data will be calculated by the weighted average method. In the specific calculation, the dependency of each node is used as the weight, and the market and operating change data are used as the influencing factors for weighted synthesis, so as to obtain the data mapping of each node under the latest market and operating conditions. The data mapping results are stored in the form of database tables to support efficient query and subsequent processing, ensuring that all node data are updated in real time to reflect the latest business environment.

[0089] S213: Based on the adjusted node data mapping result, the node connection relationship strength in the cost factor relationship graph is reconfigured, and the connection weight is modified digitally, using the formula:

[0090]

[0091] Calculate the new connection strength C between nodes i and j ij , generate an updated dependency graph, where N ij is the dependency analysis value between nodes i and j, ΔO i represents the change in the operating condition of node i, V i is the market change of node i;

[0092] Assume that the dependency analysis result N between two nodes i and j is ij is 100, indicating that node i is highly dependent on node j in a certain business process. i 20 indicates that the recent operating environment has changed significantly. Market change V i It is set to 50, indicating that the market’s influence on node i increases significantly.

[0093] Calculate the absolute value of the dependency analysis result and take the square root:

[0094]

[0095] Calculate the reciprocal of the change in operating conditions plus 1:

[0096]

[0097] Multiply these two results together:

[0098] 10×0.0476=0.476

[0099] Calculate the logarithm of the market change volume:

[0100] log(50)≈1.6990

[0101] Adding the product and the logarithm gives the final connection strength:

[0102] 0.476+1.6990=2.175

[0103] The final value of 2.175 indicates that the connection strength between nodes i and j is adjusted to 2.175 under the current market and operating conditions. The results show that even when operating conditions change significantly and market influence increases, the dependency between nodes i and j can still be maintained at a high level, ensuring business continuity. In this way, the connection strength in the network diagram can be dynamically adjusted according to actual conditions, ensuring the stability and efficient operation of the entire network.

[0104] See also Figure 4 , the analysis steps of the key node time series are:

[0105] S311: Based on the updated dependency graph, use the formula:

[0106]

[0107] Calculate the influence score of each node i i , generating a node influence score, where N(i) represents the set of nodes directly connected to node i, and w ij is the weight connecting i and j, d ij is the distance between nodes;

[0108] Assume that node i is connected to three nodes j, and the connection weights between the nodes are w ij are 2, 2 and 1 respectively, and the distance d ij They are 1, 2 and 1 respectively.

[0109] Calculate the contribution of node i and its first neighbor node:

[0110]

[0111] Calculate the contribution of node i and the second neighbor node:

[0112]

[0113] Calculate the contribution of node i and the third neighbor node:

[0114]

[0115] Add up all contributions to get the total influence score of node i:

[0116] 1.0+0.67+0.5=2.17

[0117] The results show that the total influence score of node i is 2.17. The score reflects the overall influence of node i in its network after taking into account the distance between nodes. A higher score indicates that node i has a stronger influence on its direct neighboring nodes, and the assessment of this influence is adjusted by taking into account the actual physical or logical distance of each neighbor node from i.

[0118] S312: Based on the node influence score and the screening threshold T, each node influence score is screened and the score Score is selected. i For nodes higher than T, a list of key nodes is generated;

[0119] When determining the screening threshold T, it is first necessary to collect the influence scores of all nodes in the past certain period of time through network monitoring. The specific operation is: summarize the influence scores of all nodes and calculate their average and standard deviation. The average value is calculated by adding all the scores and dividing them by the total number of nodes, while the standard deviation is calculated based on the sum of the squares of the difference between each node's score and the average score, divided by the total number of nodes. In the process of setting T, the average score provides a baseline, and the addition of two times the standard deviation ensures that only those nodes that are significantly above the average level are selected. This process involves not only statistical calculations, but also needs to match the actual operating conditions of the network, such as the activity cycle and main functions of the nodes. Ultimately, the T value determined by this method can effectively screen out the truly critical nodes from a large number of nodes, providing support for subsequent network management and optimization decisions.

[0120] S313: applying time series analysis to the key node list to reveal the behavior of the nodes in the time dimension, analyzing the change trend of the nodes in the time dimension, and obtaining the key node time series analysis results;

[0121] During the execution of time series analysis, data is captured regularly from key nodes in the network and then stored in the enterprise data warehouse. The collected data is first preprocessed, including cleaning erroneous data and filling missing values ​​to ensure data quality. The data is processed using the Pandas and Statsmodels libraries in R or Python. The first step is to determine the time series model that is suitable for the data. The commonly used model is the autoregressive moving average (ARMA) model. The parameter selection of the model, such as the order of the autoregressive term and the moving average term, is selected automatically by the program based on the Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC) to ensure the predictive performance of the model. After completing the parameter setting of the model, the model is fitted. Finally, the future node behavior is predicted based on the fitted model. The prediction results will help network administrators understand possible network behavior trends, so that they can prepare in advance in actual operation and maintenance and optimize the performance and stability of the network.

[0122] See also Figure 5 , the steps to obtain the trend forecast analysis results are:

[0123] S321: Collect and integrate the time series analysis results and historical cost data of key nodes, record the operation cost, maintenance cost, and traffic and connection behavior data of the nodes, verify the integrity and consistency of the data, and obtain integrated data;

[0124] Perform data collection and integration, collect data from multiple departments within the enterprise, including cost data provided by the finance department and operational data provided by the IT department. The specific operations are as follows: Automatically export detailed cost records from the financial system for the past few years, including but not limited to fixed asset depreciation, operating costs, and employee salaries. At the same time, obtain performance indicators of key nodes, such as traffic data, number of connections, and fault response time. In order to ensure the accuracy and completeness of the data, format and clean the exported data, including eliminating duplicate records, filling missing values, and correcting abnormal data. In addition, regular communication between departments ensures that all relevant data has been updated and reflects the latest operating conditions, which not only improves the accuracy of data analysis, but also provides a solid foundation for subsequent cost analysis and forecasting.

[0125] S322: Perform trend analysis on the integrated data, check the relationship between cost and time, determine the fluctuation pattern of historical data, identify the key drivers of cost changes and potential trend lines, and output preliminary trend identification results;

[0126] Based on the above results, data analysis is performed to identify key factors and potential trends in cost changes. All collected data is imported into the data analysis tool, and the same processing is performed on the performance indicators of the nodes. Through comparative analysis, the specific impact of specific events or changes on cost and performance can be observed, for example, how a new technology deployment affects operating costs or how a market activity affects data traffic. In addition, the team will conduct correlation analysis to explore the relationship between different data points, for example, whether cost increases are related to increased fault response time. Detailed data analysis not only helps identify historical trends, but also predicts possible future cost fluctuations and provides decision support for management.

[0127] S323: Based on the preliminary trend identification results, organize and display the trend data, and perform visualization to predict the potential fluctuations of future costs, and analyze the future change trends of key cost factors to obtain trend prediction analysis results;

[0128] Based on the analysis results, predict future cost changes and generate trend forecast reports. During the process, the data analysis team will use the previous analysis results to build a comprehensive report that details the possible causes and future trends of cost fluctuations. The report will contain a variety of charts, such as bar charts, line charts, and scatter charts, which show how key cost factors change over time. In addition, the report also includes a series of recommended measures, such as potential cost savings opportunities and preventive maintenance tips, designed to help companies reduce unforeseen expenses and optimize financial planning. Through this stage of work, companies can gain a comprehensive perspective, understand cost dynamics, and adjust strategies accordingly to adapt to market changes and internal operational needs.

[0129] See also Figure 6 , the steps to determine the core difference area are:

[0130] S411: extracting actual cost data of the current period from a known cost database based on the trend forecast analysis results, performing preliminary cost comparison, and generating preliminary comparison results;

[0131] In the process of extracting cost data, we first connect to the company's financial management to store all financial transaction data from the past few years, including but not limited to daily operating costs, project costs and capital expenditures of various departments. Using the SQL query language, we perform a detailed query operation for the data in the current cycle, which covers the number, date of occurrence, amount and relevant departments of each cost. In addition, the collection of each cost data must meet the audit quality standards to ensure the integrity and accuracy of the data. The data is then screened and classified to eliminate any outliers or irrelevant data. The key is to accurately extract the actual cost data in the current cycle from a large amount of financial data and match it with the budget data. Next, the data is aggregated into a preliminary comparative data set, which not only includes the comparison of actual costs and budget costs, but also analyzes the deviation of each cost item relative to the budget, providing basic data for subsequent cost control analysis.

[0132] S412: Analyze each cost data in the preliminary comparison results, using the formula:

[0133]

[0134] Calculate the percentage deviation Δp between the predicted value and the actual value i , get the deviation analysis report, where P i represents the prediction cost of the ith item, A i represents the actual cost of item i;

[0135] Assume that the actual value of a cost item is 80,000 yuan and the forecast value is 75,000 yuan, then:

[0136] P i =75000

[0137] A i =80000

[0138] Enter the formula to calculate:

[0139]

[0140] Δp i =6.25%

[0141] The calculated result of 6.25% shows that the budget control deviation of this cost item is 6.25%, indicating that there is a significant deviation between the budget setting and the actual costs incurred.

[0142] S413: Based on the deviation analysis report, cost items whose deviations exceed the set threshold are screened and marked to generate core difference analysis results;

[0143] In the process of analyzing cost deviations, the key is to identify the cost items that deviate most significantly from the budget. First, import data from the detailed deviation analysis report provided above, which includes the budget, actual expenditure and deviation percentage of each cost item. Use Excel or Python pandas library to process and sort the data to filter out cost items whose deviation percentage exceeds the preset threshold (such as 10%). This screening process involves complex conditional judgments and data operations to ensure that only cost items with truly large deviations are marked. Subsequently, in-depth analysis is conducted on the cost items that exceed the threshold to explore the reasons for the deviation, such as market price fluctuations, operational errors, or unreasonable budget settings. Each analysis needs to be recorded in detail so that management can understand the specific background and environment in which the deviation occurs. Finally, the analysis results are summarized to form a report on the core difference area, describing the main problem areas in budget control and recommended adjustment measures, providing a basis for formulating more effective cost management strategies.

[0144] See also Figure 7 , the steps to obtain the verified cost prediction results are:

[0145] S421: Based on the core difference area data, all cost items exceeding the predetermined threshold are recorded, along with the amount, time and relevant responsible departments, to form a difference data report;

[0146] An in-depth analysis of each cost item in the core variance area, focusing on items that exceed management's expected threshold. The specific process includes detailed records of the amount, time of occurrence, and responsible department for each cost item. First, the accuracy of the data is verified by manually checking paper financial records and electronic ledgers, and then the information is manually entered into an Excel worksheet for further analysis. In this step, each cost record is annotated for easy tracking, and special attention is paid to recording those unusually high or low cost data. In addition, interviews are conducted with department heads to collect additional information on cost fluctuations, such as unexpected events or unplanned expenses. The aggregation of all data is intended to form a comprehensive and detailed variance report, which will be used to support cost control decisions and strategy adjustments.

[0147] S422: Based on the difference data report, review and adjust the cost assessment criteria to reflect the fluctuation and change trend of the actual cost data, and update the assessment criteria in combination with the deviation between the cost data and the forecast to obtain the adjusted assessment criteria;

[0148] After receiving the detailed variance report, we turn to a detailed review and adjustment of the existing cost assessment criteria. The first step in this process is to analyze the data in the variance report item by item to identify possible factors that cause cost deviations. For example, by manually comparing the cost differences between different departments or projects, we try to find patterns or anomalies. Then, for each problem point found, we organize a group discussion with financial analysts, project managers, and department heads to jointly evaluate the root cause of each deviation and propose feasible solutions. The results of the discussion and analysis will be used to adjust the methods and parameters of cost forecasting, such as revising the budget allocation method or resetting the cost control threshold, to ensure that the evaluation criteria can more truly reflect the actual operation.

[0149] S423: Reanalyze the historical data using the adjusted evaluation criteria, compare the cost forecasts and actual data before and after the adjustment, identify and verify the improvement effect of the evaluation criteria, and generate verified cost forecast results through iterative comparison;

[0150] Apply the adjusted evaluation criteria to historical data to verify the actual effect of the changes. In this process, Excel is first used to reorganize and reanalyze the data, manually import the adjusted parameters, and then use basic data analysis functions, such as sorting and filtering, to compare the differences in data before and after the adjustment. After each analysis, the results are manually recorded, with special attention paid to recording data points that show significant improvement and data points that still have deviations. In addition, multiple rounds of data verification will be carried out, including backtracking historical data to ensure the applicability and accuracy of the new standards, and to ensure that each data iteration can provide direct evidence of improvements in cost control and forecast accuracy. The final report will detail the performance of the adjusted standards in actual applications and recommended strategies for future cost management.

[0151] An operation cost prediction system based on artificial intelligence, comprising:

[0152] The data classification module collects human resource costs, material costs, and energy usage costs, classifies and organizes them, uses differential cost category data sets as nodes, uses the interactions between cost categories to establish edge relationships, calculates the correlation weights between cost categories, determines node weights, reflects the influence and relevance of differential cost categories, and generates a cost factor relationship diagram;

[0153] The dependency update module calculates the dependencies between each node item by item based on the cost factor relationship graph by analyzing the response of the difference node to the cost change, updates the data mapping of each node node by node based on the latest change data, reconfigures the node connection strength in the cost factor relationship graph, and generates an updated dependency graph;

[0154] The trend forecasting and analysis module calculates the influence score of each node based on the updated dependency graph, and screens the influence score of each node in combination with the screening threshold to determine the key nodes, and then conducts time series analysis. Combined with historical cost data, it checks the relationship between cost and time, predicts potential fluctuations in future costs, and analyzes future trends of key cost factors to obtain trend forecasting and analysis results;

[0155] Based on the trend forecast analysis results, the cost assessment and adjustment module extracts the actual cost data of the current period from the known cost database and conducts a preliminary cost comparison, calculates the deviation percentage between the forecast value and the actual value, screens out cost items with deviations exceeding the set threshold, records all cost items that exceed the predetermined threshold, reviews and adjusts the cost assessment standards, compares the cost forecast and actual data before and after the adjustment, identifies and verifies the improvement effect of the assessment standards, and generates verified cost forecast results through iterative comparison.

[0156] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. An operation cost prediction method based on artificial intelligence, characterized in that: The following steps are involved: S1: Collect and organize human resource costs, material costs, and energy usage costs, create independent data sets for each cost category, represent nodes with differential cost categories, represent edges with interactions between cost categories, determine node weights through cost influence analysis, and construct a cost factor relationship diagram; S2: Based on the cost factor relationship graph, the dependencies between nodes are analyzed, the node data is updated to map the changes in market and operating conditions, and the node relationship strength is adjusted to reflect the actual cost dynamics, thereby obtaining an updated dependency relationship graph; S3: using the updated dependency graph, calculating the influence score of each node, selecting nodes with scores higher than the threshold as key nodes according to the set threshold, and performing time series analysis on the key nodes. Based on the time series analysis results and in combination with historical cost data, linear trend analysis is applied to predict future cost fluctuations, capturing future change trends of key cost factors, and comprehensively generating trend prediction analysis results; S4: Based on the trend forecast analysis results, combined with actual cost data, cost deviation analysis is performed to determine the core difference area between the predicted value and the actual value, and the results are determined based on the core difference area. The evaluation criteria for cost impact are adjusted, and the forecast error is optimized through iterative data comparison to generate verified cost forecast results.

2. The method for predicting operating costs based on artificial intelligence according to claim 1, characterized in that: The steps for constructing the cost factor relationship diagram are as follows: S111: Collect human resource costs, material costs and energy usage costs, classify and sort the data according to cost types, and obtain cost classification records; S112: Based on the cost classification records, the difference cost category data set is used as a node, and the edge relationship is established by utilizing the interaction between the cost categories, using the formula: Calculate the correlation weight W between cost categories ij , generate the preliminary structure of the relationship graph, where C ik and C jk denote the costs of cost categories i and j on item k respectively; S113: Perform influence analysis based on the preliminary structure of the relationship graph, determine node weights, reflect the influence and relevance of the differential cost categories, and generate a cost factor relationship graph.

3. The method for predicting operating costs based on artificial intelligence according to claim 2, characterized in that: The steps for obtaining the updated dependency graph are: S211: According to the cost factor relationship diagram, combined with the sensitivity of node cost changes, by analyzing the response of the difference nodes to the cost changes, the dependency between each node is calculated item by item to obtain the node dependency analysis result; S212: using the inter-node dependency analysis results, combined with the latest change data of market and operating conditions, updating the data mapping node by node, revealing the changes of current market and operating conditions, and generating adjusted node data mapping results; S213: Reconfigure the node connection relationship strength in the cost factor relationship graph according to the adjusted node data mapping result, and modify the connection weight in a digital manner, using the formula: Calculate the new connection strength C between nodes i and j ij , generate an updated dependency graph, where N ij is the dependency analysis value between nodes i and j, ΔO i represents the change in the operating condition of node i, V i is the market change at node i.

4. The method for predicting operating costs based on artificial intelligence according to claim 3, characterized in that: The analysis steps of the key node time series are: S311: Based on the updated dependency graph, use the formula: Calculate the influence score of each node i i , generating a node influence score, where N(i) represents the set of nodes directly connected to node i, and w ij is the weight connecting i and j, d ij is the distance between nodes; S312: According to the node influence score, combined with the screening threshold T, each node influence score is screened and a score Score is selected. i For nodes higher than T, a list of key nodes is generated; S313: Apply time series analysis to the key node list to reveal the behavior of the nodes in the time dimension, analyze the change trend of the nodes in the time dimension, and obtain the key node time series analysis results.

5. The method for predicting operating costs based on artificial intelligence according to claim 4, characterized in that: The steps for obtaining the trend forecast analysis results are as follows: S321: Collect and integrate the time series analysis results and historical cost data of the key nodes, record the operation cost, maintenance cost, and traffic and connection number behavior data of the nodes, verify the integrity and consistency of the data, and obtain integrated data; S322: Perform trend analysis on the integrated data, check the relationship between cost and time, determine the fluctuation pattern of historical data, identify key driving factors and potential trend lines of cost changes, and output preliminary trend identification results; S323: Based on the preliminary trend identification results, organize and display trend data and perform visualization to predict potential fluctuations in future costs and analyze future change trends of key cost factors to obtain trend forecast analysis results.

6. The method for predicting operating costs based on artificial intelligence according to claim 5, characterized in that: The steps for determining the core difference area are: S411: extracting actual cost data of the current period from a known cost database according to the trend forecast analysis result, performing preliminary cost comparison, and generating preliminary comparison results; S412: Analyze each cost data in the preliminary comparison result, using the formula: Calculate the percentage deviation Δp between the predicted value and the actual value i , get the deviation analysis report, where P i represents the prediction cost of the ith item, A i represents the actual cost of item i; S413: Based on the deviation analysis report, cost items whose deviations exceed a set threshold are screened and marked to generate core difference analysis results.

7. The method for predicting operating costs based on artificial intelligence according to claim 6, characterized in that: The steps for obtaining the verified cost prediction results are: S421: Based on the core difference area data, all cost items exceeding a predetermined threshold are recorded, along with the amount, time, and relevant responsible departments, to form a difference data report; S422: Based on the difference data report, re-examine and adjust the cost evaluation standard to reflect the fluctuation and change trend of the actual cost data, and update the evaluation standard in combination with the deviation between the cost data and the forecast to obtain the adjusted evaluation standard; S423: Re-analyze the historical data using the adjusted evaluation criteria, compare the cost forecasts and actual data before and after the adjustment, identify and verify the improvement effect of the evaluation criteria, and generate verified cost forecast results through iterative comparison.

8. An operation cost prediction system based on artificial intelligence, characterized in that: The system is used to execute the artificial intelligence-based operating cost prediction method according to any one of claims 1 to 7, comprising: The data classification module collects human resource costs, material costs, and energy usage costs, classifies and organizes them, uses differential cost category data sets as nodes, uses the interactions between cost categories to establish edge relationships, calculates the correlation weights between cost categories, determines node weights, reflects the influence and relevance of differential cost categories, and generates a cost factor relationship diagram; The dependency updating module calculates the dependency between each node item by item according to the cost factor relationship graph by analyzing the response of the difference node to the cost change, updates the data mapping of each node node by node in combination with the latest change data, reconfigures the node connection relationship strength in the cost factor relationship graph, and generates an updated dependency relationship graph; The trend forecasting and analysis module calculates the influence score of each node based on the updated dependency graph, and screens the influence score of each node to determine the key nodes in combination with the screening threshold, and performs time series analysis, and checks the relationship between cost and time in combination with historical cost data, predicts potential fluctuations in future costs, and analyzes future change trends of key cost factors to obtain trend forecasting and analysis results; Based on the trend forecast analysis results, the cost assessment and adjustment module extracts the actual cost data of the current period from the known cost database and performs a preliminary cost comparison, calculates the deviation percentage between the predicted value and the actual value, screens out cost items whose deviations exceed the set threshold, records all cost items that exceed the predetermined threshold, reviews and adjusts the cost assessment standards, compares the cost forecasts and actual data before and after the adjustment, identifies and verifies the improvement effects of the assessment standards, and generates verified cost forecast results through iterative comparison.

Images