Enterprise operation supervision system based on data analysis

Abstract

The invention discloses an enterprise operation supervision system based on data analysis, relates to the technical field of data analysis, and overcomes the defects of dependence on static data and incapability of capturing a long-term trend by collecting time sequence data D and calculating an inertial behavior vector Hinert and establishing a long-term stable behavior mode of each service node. By calculating the current behavior vector Hcurr and acquiring the behavior deviation vector Dd, whether the service node deviates from long-term inertia can be accurately identified, and the problem that a traditional system lags in abnormal response to enterprise services is avoided. And a comprehensive risk vector Rrisk is calculated in combination with an inertial behavior vector Hinert, so that the defect that an existing enterprise supervision system can only perform static analysis based on current data and cannot pre-judge a business change trend in advance is overcome. By optimizing the scheme Oopt, service adjustment is performed, an operation strategy is accurately optimized, and the problem that a traditional system lacks dynamic adjustment capability and is difficult to intervene in the early stage is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and specifically provides an enterprise operation supervision system based on data analysis. Background Art

[0002] In today's digital age, enterprise operations are increasingly relying on data analysis technology, which belongs to the large field of enterprise management and intelligent data analysis. With the rapid development of information technology, data analysis has become one of the important means for enterprises to enhance competitiveness and optimize operations. In this large field, data-driven enterprise operation supervision is a direction that has received increasing attention, which involves the management and optimization of multiple aspects such as enterprise production, supply chain, sales, finance, and human resources.

[0003] In the invention with the Chinese patent application number CN202411582488.3, an enterprise digital intelligent operation supervision system is disclosed. An enterprise digital intelligent operation supervision system includes: an enterprise operation data acquisition module, an enterprise operation graph structure data construction module, and an operation data anomaly detection module. By constructing enterprise operation data into graph structure data, the present invention can strengthen the connection between different entities in the operation data, and with the assistance of the connection between different entities, an operation data anomaly detection model is used to detect anomalies in the operation data, thereby realizing the automatic supervision of enterprise operations; and in the process of detecting anomalies in the operation data through the operation data anomaly detection model, virtual edges are used to strengthen the connection between nodes of the same type, and the order nodes are also strengthened, thereby further strengthening the connection between different entities in the operation data and improving the accuracy of feature extraction for enterprise operation data.

[0004] It can be seen that currently, the core functions of most enterprise operation supervision systems still rely on static data analysis, and its main disadvantage is the lack of the ability to judge long-term trends and behavioral inertia. These systems often collect data using fixed time windows, such as daily, weekly, or monthly data aggregation, rather than observing the changing trends of business data through continuous analysis. Although this method can provide clear historical data, it is difficult to capture the gradual trends and potential anomalies of business changes, resulting in the lack of the ability to model long-term behavioral inertia. This makes it impossible for enterprises to formulate effective customer retention strategies and may ultimately lead to customer loss. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the present invention provides an enterprise operation supervision system based on data analysis, which solves the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: An enterprise operation supervision system based on data analysis, comprising a data acquisition and processing module, a behavioral inertia analysis module, an abnormal deviation detection module, a trend prediction and evaluation module, a decision execution module, and an evaluation and iterative optimization module; The data acquisition and processing module collects enterprise operation data, performs data cleaning and data standardization preprocessing, constructs a time series for the preprocessed enterprise operation data, and forms a time series data set D; The behavioral inertia analysis module calculates the inertial behavior vector Hinert of each business node according to the obtained time series data set D; The abnormal deviation detection module extracts the current business characteristics of the business node to form the current behavior vector Hcurr, calculates based on the inertial behavior vector Hinert to obtain the behavior deviation vector Dd, compares it with the deviation threshold Deu to obtain the deviation status result, and then triggers trend analysis according to the deviation status result; When the trend prediction and evaluation module triggers trend analysis, it uses a time series prediction model for prediction to obtain the future behavior vector Hfuture, and comprehensively evaluates it with the inertial behavior vector Hinert to obtain the comprehensive risk vector Rrisk; The decision execution module compares the obtained comprehensive risk vector Rrisk with the preset optimization threshold Ymax to obtain the optimization plan Oopt; The evaluation and iterative optimization module compares the inertial behavior vector Hinert after executing the optimization plan Oopt to judge the adjustment status of the optimization plan Oopt.

[0007] Preferably, the data acquisition and processing module includes a collection unit and a processing unit; The collection unit collects the original enterprise operation business data from multiple enterprise management systems, and after calculating the original enterprise operation business data, forms the original data set Draw; Among them, the enterprise management systems include the ERP system, CRM system, SCM system, and OMS system, The original data set Draw includes customer transaction data Cfreq, supply chain data Sstb, order fulfillment data Ofolw, and inventory data Icons; Among them, the customer transaction data Cfreq is obtained by the OMS system and the CRM system by counting the number of customer transactions in the past N days and calculating the average value; The customer transaction data Cfreq is specifically obtained through the following calculation formula: ; In the formula, U represents the number of customer orders, and U(i) represents the number of customer orders on the i-th day; The supply chain data Sstb is obtained by calculating the ratio of the standard deviation to the mean of the supplier delivery time through the SCM system; The supply chain data Sstb is specifically obtained through the following calculation formula: ; In the formula, σdel represents the standard deviation of the supplier delivery time, and μdel represents the mean of the supplier delivery time; The order fulfillment data Ofolw is obtained by calculating the average time from order creation to delivery through the OMS system; The order fulfillment data Ofolw is specifically obtained through the following calculation formula: ; In the formula, TcreateD represents the time from order creation to delivery, and TcreateP represents the time from order creation to payment; The inventory data Icons is obtained by calculating the ratio of the inventory consumption rate to the sales rate through the ERP system; The inventory data Icons is specifically obtained through the following calculation formula: ; In the formula, Iout represents the inventory consumption rate, specifically representing the daily inventory reduction, and Iin represents the inventory replenishment rate, specifically representing the daily storage increase.

[0008] Preferably, the processing unit performs data cleaning and data standardization preprocessing on the collected original data set Draw, then arranges the preprocessed original data set Draw in chronological order to form a time series data set D, and synchronously stores the time series data set D as a historical data set Dhist. The time series data set D includes standard customer transaction data Cnorm, standard supply chain data Snorm, standard order fulfillment data Onorm, and standard inventory data Inorm; Among them, data cleaning includes using the interpolation method at the front and back time points to remove missing values and using Z-score to filter data outside 3 times the standard deviation; Data standardization preprocessing includes using the Z-score standardization method to normalize the data in the original data set Draw.

[0009] Preferably, the behavior inertia analysis module includes a time aggregation unit and a vector generation unit; The time aggregation unit calculates the historical behavior trends of different business nodes based on the obtained time series data set D, including performing weighted aggregation processing on the time series data set D at different time points using an exponentially decaying weighted calculation method to obtain a feature vector Xw after weighted aggregation processing. The feature vector Xw includes feature customer transaction data Cwe, feature supply chain data Swe, feature order fulfillment data Owe, and feature inventory data Iwe; The feature vector Xw is obtained through the following calculation method: ; In the formula, M represents the length of the historical time window, w(i) represents the time decay weight, specifically representing the decay weight at the i-th time point, and D(i, j) represents the j-th standard data in the time series data set at the i-th time point; Among them, the time decay weight w(i) is specifically , where λ represents the time decay factor, specifically used to control the influence degree of past data.

[0010] Preferably, the vector generation unit calculates the inertial behavior vector Hinert of each business node based on the obtained feature vector Xw; The inertial behavior vector Hinert includes vector customer transaction data Cin, vector supply chain data Sin, vector order fulfillment data Oin, and vector inventory data Iin; The vector customer transaction data Cin is obtained through the following calculation formula: ; In the formula, Cwe(i) represents the feature customer transaction data at the i-th time point; The vector supply chain data Sin is obtained through the following calculation formula: ; In the formula, Swe(i) represents the feature supply chain data at the i-th time point; The vector order fulfillment data Oin is obtained through the following calculation formula: ; In the formula, Owe(i) represents the feature order fulfillment data at the i-th time point; The vector inventory data Iin is obtained through the following calculation formula: ; In the formula, Iwe(i) represents the feature inventory data at the i-th time point.

[0011] Preferably, the abnormal deviation detection module includes a construction unit and a deviation judgment unit; The building unit forms the current behavior vector Hcurr by extracting the time series data set D of the current business characteristics of the business node, marking the real-time time point tnow, specifically Hcurr = {tnow, Cnorm, Snorm, Onorm, Inorm}; The deviation judgment unit calculates the deviation degree from the inertial behavior vector Hinert according to the obtained current behavior vector Hcurr, obtains the behavior deviation vector Dd, compares the obtained behavior deviation vector Dd with the deviation threshold Deu, obtains the deviation state result, and then triggers the trend analysis according to the deviation state result.

[0012] Preferably, the behavior deviation vector Dd is obtained through the following calculation formula: ; In the formula, n represents the total number of vectors, Hinert(k) represents the k-th vector in the inertial behavior vector, and Hcurr(k) represents the k-th vector in the current behavior vector; The deviation state result is obtained through the following comparison method: When the behavior deviation vector Dd ≥ the deviation threshold Deu, the obtained deviation state result is the trigger result, triggering the trend prediction; When the behavior deviation vector Dd < the deviation threshold Deu, the obtained deviation state result is the non-trigger result, not triggering the trend prediction.

[0013] Preferably, the trend prediction evaluation module includes a risk prediction unit; When the risk prediction unit triggers the trend analysis, it extracts the historical data set Dhist within a fixed period as the input data, uses the time series prediction model to predict the future behavior vector Hfuture of the input data, and comprehensively evaluates it with the inertial behavior vector Hinert to obtain the comprehensive risk vector Rrisk; The predicted future behavior vector Hfuture is obtained through the following calculation formula: ; In the formula, Hfuture(k) represents the predicted future behavior vector of the k-th vector, Dhist(t - p, k) represents the k-th vector of the historical data set Dhist at the past time t - p, represents the regression coefficient, specifically representing the regression coefficient at the past time p, represents the error term; The comprehensive risk vector Rrisk is obtained through the following calculation formula: .

[0014] Preferably, the decision execution module includes a solution generation unit; The solution generation unit compares the obtained comprehensive risk vector Rrisk with the preset optimization threshold Ymax to obtain the comprehensive risk optimization evaluation result, and obtains the optimization solution Oopt according to the comprehensive risk optimization evaluation result; When the comprehensive risk vector Rrisk < the optimization threshold Ymax, the obtained comprehensive risk optimization evaluation result is the non-adjustment result, and the optimization solution Oopt is not generated; When the comprehensive risk vector Rrisk ≥ the optimization threshold Ymax, the obtained comprehensive risk optimization evaluation result is the adjustment result, and the optimization solution Oopt is generated, including calculating the comprehensive risk vector Rrisk to obtain the proportion adjustment amount to adjust the original enterprise operation business data, and sending the optimization solution Oopt to the relevant enterprise management system for notification.

[0015] Preferably, the evaluation and iterative optimization module includes an evaluation unit; The evaluation unit marks the inertia behavior vector Hinert after executing the optimization solution Oopt as the solution execution inertia behavior vector F Hinert, and then compares it with the inertia behavior vector Hinert before executing the optimization solution Oopt. After obtaining the difference, it is marked as the vector deviation factor Hh. By comparing the vector deviation factor Hh with the preset optimization trigger threshold Hopt, the adjustment status of the optimization solution Oopt is judged. The adjustment status includes continuous adjustment and non-adjustment; The adjustment status of the optimization solution Oopt is obtained through the following comparison method: When the vector deviation factor Hh < the optimization trigger threshold Hopt, the obtained adjustment status of the optimization solution Oopt is non-adjustment, and the optimization solution Oopt is continuously executed; When the vector deviation factor Hh ≥ the optimization trigger threshold Hopt, the obtained adjustment status of the optimization solution Oopt is continuous adjustment, and the optimization solution Oopt is regenerated twice.

[0016] The present invention provides an enterprise operation supervision system based on data analysis, which has the following beneficial effects: (1) By collecting time series data D and calculating the inertial behavior vector Hinert, we can effectively establish the long-term stable behavior pattern of each business node, making up for the defect that the traditional enterprise operation supervision system relies on static data and cannot capture long-term trends. By calculating the current behavior vector Hcurr and obtaining the behavior deviation vector Dd, we can accurately identify whether the business node deviates from the long-term inertia, avoiding the problem that the traditional system has a delayed response to abnormal business operations and lacks real-time monitoring capabilities. In addition, the inertial behavior vector Hinert is combined to calculate the comprehensive risk vector Rrisk, thereby overcoming the shortcomings of the existing enterprise supervision system that can only perform static analysis based on current data and cannot predict business change trends in advance. Through the optimization solution Oopt, business adjustments are made and operation strategies are accurately optimized, which solves the problem that the traditional system lacks dynamic adjustment capabilities and is difficult to intervene in the early stage, and breaks through the limitations of the existing system in static analysis, delayed response, unpredictability and difficulty in optimization and adjustment.

[0017] (2) Through cross-system integration of customer transaction data Cfreq, supply chain data Sstb, order fulfillment data Oflow, and inventory data Icons in the ERP system, CRM system, SCM system, and OMS system, data fusion of multiple business modules is achieved, effectively solving the problem of scattered data sources and lack of data linkage among business modules. Normalizing the original data set Draw not only enhances the comparability of the data, but also improves the calculation accuracy of subsequent behavioral inertia analysis, anomaly detection, and trend prediction. Finally, the system stores the time series data set D as a historical data set Dhist to ensure long-term data accumulation, providing enterprises with data-driven decision-making capabilities based on long-term spans, breaking through the limitations of traditional systems that can only rely on short-term data and cannot perform long-term trend prediction and optimization.

[0018] (3) By calculating the comprehensive risk vector Rrisk, the trend changes of the enterprise's business can be perceived in advance, making up for the defect that the traditional system cannot give early warnings of operation risks, resulting in lagged decision-making. Secondly, the solution generation unit intelligently evaluates whether optimization and adjustment are needed by comparing the comprehensive risk vector Rrisk with the preset optimization threshold Ymax, and accurately controls the adjustment range of the original enterprise operation business data by calculating the proportional adjustment amount, making the optimization solution more flexible and targeted, and effectively solving the problem that the traditional system has a single adjustment method and it is difficult to accurately match the optimization range with the risk degree. In addition, by calculating the vector deviation factor Hh of the solution execution and comparing it with the optimization trigger threshold Hopt, it is automatically judged whether the optimization solution needs further adjustment, ensuring that the system does not solidify the strategy due to one-time optimization, but forms an adaptive adjustment mechanism, overcoming the disadvantages that the traditional optimization solution cannot verify the effect after adjustment and relies on manual judgment to determine whether secondary optimization is needed. Through the above mechanism, not only can the business change trend be predicted in advance, the anomalies be quickly responded to and the operation strategy be accurately optimized, but also dynamic feedback adjustment can be carried out after the optimization execution, enabling the enterprise to achieve continuous optimization and maintain the optimal operation state in the complex and changeable market environment. Description of the Drawings

[0019] Figure 1 Schematic diagram of the steps of an enterprise operation supervision system based on data analysis according to the present invention; Figure 2 Schematic diagram of the data transmission and task distribution process; Figure 3 Schematic diagram of the threshold trigger judgment process block diagram. Detailed Embodiments

[0020] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0021] Embodiment 1 The present invention provides an enterprise operation supervision system based on data analysis. Please refer to Figure 1 、 Figure 2 and Figure 3 , including a data acquisition and processing module, a behavioral inertia analysis module, an abnormal deviation detection module, a trend prediction and evaluation module, a decision execution module, and an evaluation and iterative optimization module; The data acquisition and processing module collects enterprise operation data, performs data cleaning and data standardization preprocessing, and constructs a time series for the preprocessed enterprise operation data to form a time series data set D; The behavioral inertia analysis module calculates the inertial behavior vector Hinert of each business node based on the obtained time series data set D as the benchmark behavior model of the current business node; The abnormal deviation detection module extracts the current business features of the business node to form the current behavior vector Hcurr, calculates based on the inertial behavior vector Hinert to obtain the behavior deviation vector Dd, compares it with the deviation threshold Deu to obtain the deviation status result, and then triggers trend analysis according to the deviation status result; When the trend prediction and evaluation module triggers trend analysis, it uses a time series prediction model for prediction to obtain the future behavior vector Hfuture, and comprehensively evaluates it with the inertial behavior vector Hinert to obtain the comprehensive risk vector Rrisk; The decision execution module compares the obtained comprehensive risk vector Rrisk with the preset optimization threshold Ymax to obtain the optimization plan Oopt; The evaluation and iterative optimization module compares the inertial behavior vector Hinert after executing the optimization plan Oopt to judge the adjustment status of the optimization plan Oopt.

[0022] In this embodiment, by collecting the time series data D and calculating the inertial behavior vector Hinert, a long-term stable behavior pattern of each business node can be effectively established, making up for the defects of traditional enterprise operation supervision systems that rely on static data and cannot capture long-term trends. By calculating the current behavior vector Hcurr and obtaining the behavior deviation vector Dd, it can accurately identify whether the business node deviates from the long-term inertia, avoiding the problems of lagging abnormal response and lack of real-time monitoring ability of traditional systems for enterprise operations. By combining the inertial behavior vector Hinert to calculate the comprehensive risk vector Rrisk, it overcomes the deficiency of existing enterprise supervision systems that can only perform static analysis based on current data and cannot predict business change trends in advance. Through the optimization plan Oopt, business adjustments are made to precisely optimize the operation strategy, solving the problems of lack of dynamic adjustment ability and difficulty in intervening at an early stage in traditional systems, breaking through the limitations of existing systems in aspects such as static analysis, lagging response, inability to predict, and difficulty in optimizing and adjusting, providing more intelligent, precise, and dynamic operation supervision capabilities for enterprises, improving business stability, reducing operation risks, and enhancing the market competitiveness of enterprises.

[0023] Embodiment 2 This embodiment is an explanatory description based on Embodiment 1. Please refer to Figure 1 , specifically: The data collection and processing module includes a collection unit and a processing unit; The collection unit collects the original enterprise operation business data from multiple enterprise management systems, and after calculating the original enterprise operation business data, forms the original data set Draw; Among them, the enterprise management system includes an ERP system, a CRM system, an SCM system, and an OMS system. The original dataset Draw includes customer transaction data Cfreq, supply chain data Sstb, order fulfillment data Ofolw, and inventory data Icons. Among them, the customer transaction data Cfreq is obtained by counting the number of customer transactions in the past N days through the OMS system and the CRM system and calculating the average value. The customer transaction data Cfreq is specifically obtained through the following calculation formula: ; In the formula, U represents the number of customer orders, and U(i) represents the number of customer orders on the i-th day; the customer transaction data Cfreq is used to reflect the recent transaction frequency of customers. If it drops significantly, there may be customer loss. The supply chain data Sstb is obtained by calculating the ratio of the standard deviation to the mean of the supplier delivery time through the SCM system. The supply chain data Sstb is specifically obtained through the following calculation formula: ; In the formula, σdel represents the standard deviation of the supplier delivery time, and μdel represents the mean of the supplier delivery time; the supply chain data Sstb is used to measure the stability of the supplier delivery time. If there is large fluctuation, there may be performance risk. The order fulfillment data Ofolw is obtained by calculating the average time from order creation to delivery through the OMS system. The order fulfillment data Ofolw is specifically obtained through the following calculation formula: ; In the formula, TcreateD represents the time from order creation to delivery, and TcreateP represents the time from order creation to payment; the order fulfillment data Ofolw is used to measure the order turnover efficiency. If the value is too high, there may be performance delay risk. The inventory data Icons is obtained by calculating the ratio of the inventory consumption rate to the sales rate through the ERP system. The inventory data Icons is specifically obtained through the following calculation formula: ; In the formula, Iout represents the inventory consumption rate, specifically representing the daily inventory reduction, and Iin represents the inventory replenishment rate, specifically representing the daily storage increase.

[0024] The processing unit performs data cleaning and data standardization preprocessing on the collected original dataset Draw, and then arranges the preprocessed original dataset Draw in chronological order to form a time series dataset D. At the same time, the time series dataset D is stored as a historical dataset Dhist. The time series dataset D includes standard customer transaction data Cnorm, standard supply chain data Snorm, standard order fulfillment data Onorm, and standard inventory data Inorm; Among them, data cleaning includes using the interpolation method at the front and back time points to remove missing values and using Z-score to filter data outside 3 times the standard deviation; Data standardization preprocessing includes using the Z-score standardization method to normalize the data in the original dataset Draw.

[0025] In this embodiment, by integrating customer transaction data Cfreq, supply chain data Sstb, order fulfillment data Oflow, and inventory data Icons in the ERP system, CRM system, SCM system, and OMS system across systems, data fusion of multiple business modules is achieved, effectively solving the problems of scattered data sources and lack of data linkage among business modules. Secondly, by calculating the customer transaction data Cfreq, the system can accurately depict the customer transaction activity, making up for the deficiency of the traditional system in quantitatively analyzing the customer churn trend; by calculating the standard deviation ratio of the supply chain data Sstb, the fluctuation of supplier performance can be monitored in real time, overcoming the defect that the traditional system relies on subjective judgment for supplier evaluation and it is difficult to identify performance risks in advance; by calculating the average time from order creation to delivery using the order fulfillment data Oflow, the order turnover efficiency can be accurately measured, making up for the problem that the traditional system only relies on static indicators for order fulfillment and cannot dynamically monitor the supply chain stability; by calculating the ratio of the inventory consumption rate to the replenishment rate using the inventory data Icons, the system can dynamically evaluate the inventory matching degree, solving the problem that the traditional system lacks quantitative evaluation of the consumption trend in inventory management and is prone to inventory backlog or shortage. On the other hand, the processing unit preprocesses the data through data cleaning and data standardization to ensure the integrity and comparability of the data, overcoming the drawbacks of the traditional system where analysis results are deviated due to data noise and data scale inconsistency affects the model accuracy. The missing values are filled by the interpolation method at the front and back time points, and the data outside 3 times the standard deviation is filtered using the Z-score method, making the data more stable and suitable for machine learning analysis. In addition, through Z-score standardization, the system can normalize the original data set Draw into the standard customer transaction data Cnorm, standard supply chain data Snorm, standard order fulfillment data Onorm, and standard inventory data Inorm, not only enhancing the comparability of the data, but also improving the calculation accuracy of subsequent behavior inertia analysis, anomaly detection, and trend prediction. Finally, the system stores the time series data set D as the historical data set Dhist to ensure long-term data accumulation, providing the enterprise with the ability of data-driven decision-making based on a long time span and breaking through the limitation of the traditional system that can only rely on short-term data and cannot perform long-term trend prediction and optimization.

[0026] Embodiment 3 This embodiment is an explanatory description based on Embodiment 2. Please refer to Figure 1 and Figure 2 , specifically: The behavior inertia analysis module includes a time aggregation unit and a vector generation unit; The time aggregation unit calculates the historical behavior trends of different business nodes based on the obtained time series dataset D, including performing weighted aggregation processing on the time series dataset D at different time points using the exponentially decaying weighted calculation method to obtain the feature vector Xw after weighted aggregation processing. The feature vector Xw includes feature customer transaction data Cwe, feature supply chain data Swe, feature order fulfillment data Owe, and feature inventory data Iwe; The feature vector Xw is obtained through the following calculation method: ; In the formula, M represents the length of the historical time window, w(i) represents the time decay weight, specifically representing the decay weight at the i-th time point, and D(i, j) represents the j-th standard data in the time series dataset at the i-th time point; Among them, the time decay weight w(i) is specifically , where λ represents the time decay factor, specifically used to control the influence degree of past data.

[0027] The vector generation unit calculates the inertial behavior vector Hinert of each business node based on the obtained feature vector Xw as the benchmark behavior model of the current business node; The inertial behavior vector Hinert includes vector customer transaction data Cin, vector supply chain data Sin, vector order fulfillment data Oin, and vector inventory data Iin; The vector customer transaction data Cin is obtained through the following calculation formula: ; In the formula, Cwe(i) represents the feature customer transaction data at the i-th time point; The vector supply chain data Sin is obtained through the following calculation formula: ; In the formula, Swe(i) represents the feature supply chain data at the i-th time point; The vector order fulfillment data Oin is obtained through the following calculation formula: ; In the formula, Owe(i) represents the feature order fulfillment data at the i-th time point; The vector inventory data Iin is obtained through the following calculation formula: ; In the formula, Iwe(i) represents the feature inventory data at the i-th time point.

[0028] The abnormal deviation detection module includes a construction unit and a deviation judgment unit; The construction unit forms the current behavior vector Hcurr by extracting the time series data set D of the current business characteristics of the business node and marking the real-time time point tnow, specifically Hcurr = {tnow, Cnorm, Snorm, Onorm, Inorm}; The deviation judgment unit calculates the deviation degree between the obtained current behavior vector Hcurr and the inertial behavior vector Hinert, obtains the behavior deviation vector Dd, compares the obtained behavior deviation vector Dd with the deviation threshold Deu, obtains the deviation state result, and then triggers trend analysis according to the deviation state result.

[0029] The behavior deviation vector Dd is obtained through the following calculation formula: ; In the formula, n represents the total number of vectors, Hinert(k) represents the k-th vector in the inertial behavior vector, and Hcurr(k) represents the k-th vector in the current behavior vector; The deviation state result is obtained through the following comparison method: When the behavior deviation vector Dd ≥ the deviation threshold Deu, the obtained deviation state result is the trigger result, triggering trend prediction; When the behavior deviation vector Dd < the deviation threshold Deu, the obtained deviation state result is the non-trigger result, not triggering trend prediction.

[0030] In this embodiment, the weighted aggregated feature vector Xw is calculated through the time series data set D, and the historical behavior trend of the business node is constructed, enabling the system to automatically adapt to the characteristics of higher weights for recent business changes and reduced influence of long-term data, and overcoming the problem of the traditional system treating all historical data equally, resulting in an overly large influence of long-term data. Further, the vector generation unit calculates the inertial behavior vector Hinert through the feature vector Xw to form the benchmark behavior model of enterprise operation, ensuring that the enterprise operation monitoring not only depends on the current data but also can comprehensively consider the long-term business model, making the system more advantageous in dealing with seasonal fluctuations, periodic changes, and long-term behavior pattern recognition. Secondly, the abnormal deviation detection module obtains the current behavior vector Hcurr through the construction unit, marks the operation state of the real-time business node, and calculates the behavior deviation vector Dd using the deviation judgment unit, realizing the real-time comparison between the current operation mode and the long-term inertial mode. This process enables the system to not only identify single abnormal events but also detect trend changes of gradual deviations, overcoming the defect that the traditional system can only respond to short-term anomalies and is difficult to discover hidden operation problems. Finally, based on the comparison between Dd and the deviation threshold Deu, the system accurately determines the business state and automatically triggers trend prediction analysis when the behavior of the business node deviates significantly, effectively improving the enterprise's abnormal response speed and reducing operation losses caused by decision-making lags. In summary, through the combination of inertial behavior analysis and deviation monitoring, the problems of the traditional supervision system being difficult to automatically adapt to long-term business changes, unable to accurately distinguish short-term fluctuations from long-term deviations, and having limited abnormal identification ability are solved, enabling the enterprise to have more efficient operation monitoring ability, more accurate business risk assessment ability, and more agile response decision-making ability, thereby improving the overall management efficiency and enhancing the enterprise's competitiveness and risk resistance ability.

[0031] Embodiment 4 This embodiment is an explanatory description carried out in Embodiment 3. Please refer to Figure 1 and Figure 2 , specifically: The trend prediction and evaluation module includes a risk prediction unit; When the risk prediction unit triggers trend analysis, it extracts the historical data set Dhist within a fixed period as input data, uses a time series prediction model to predict the future behavior vector Hfuture of the input data, and comprehensively evaluates it with the inertial behavior vector Hinert to obtain the comprehensive risk vector Rrisk; The future behavior vector Hfuture is obtained through the following calculation formula: ; In the formula, Hfuture(k) represents the predicted future behavior vector of the k-th vector, and Dhist(t - p, k) represents the k-th vector of the historical data set Dhist at the past time t - p. denotes the regression coefficient, specifically the regression coefficient at time p in the past. denotes the error term; Specific example illustration: Suppose we want to predict the standard customer transaction frequency Cnorm for the next month. At this time, we have data for the past three months as input data: Month = t - 3; Customer transaction frequency = 420; After being processed by the Z-score normalization method: Standard customer transaction frequency Cnorm = -0.6; Month = t - 2; Customer transaction frequency = 460; After being processed by the Z-score normalization method: Standard customer transaction frequency Cnorm = 0.2; Month = t - 1; Customer transaction frequency = 500; After being processed by the Z-score normalization method: Standard customer transaction frequency Cnorm = 1.1; Among them, the average value of the Z-score normalization method is: 450; The standard deviation is: 50: Based on the time series prediction model, using the data of the past one month and two months as input data, predict the future customer transaction frequency for the input data: Set the regression coefficient of the time series prediction model = 0.6, = 0.3; Error term = 0.1; It can be seen from this that the predicted future customer transaction frequency (normalized value): Hfuture(Cnorm) = (0.6 * 1.1) + (0.3 * 0.2) + (0.1) = 0.82; At this time, perform anti-normalization to restore the original customer transaction frequency to obtain: 0.82 * 50 + 450 = 491; Obtain the future customer transaction frequency = 491; The comprehensive risk vector Rrisk is obtained through the following calculation formula: .

[0032] The decision execution module includes a solution generation unit; The solution generation unit compares the obtained comprehensive risk vector Rrisk with the preset optimization threshold Ymax to obtain the comprehensive risk optimization evaluation result, and obtains the optimization solution Oopt according to the comprehensive risk optimization evaluation result; When the comprehensive risk vector Rrisk < optimization threshold Ymax, the obtained comprehensive risk optimization evaluation result is the non-adjustment result, and no optimization solution Oopt is generated; When the comprehensive risk vector Rrisk ≥ the optimization threshold Ymax, obtain the adjustment result of the comprehensive risk optimization evaluation result, and generate an optimization plan Oopt, including calculating the comprehensive risk vector Rrisk to obtain the proportion adjustment amount to adjust the original enterprise operation business data, and sending the optimization plan Oopt to the relevant enterprise management system for notification; The comprehensive risk vector Rrisk is the proportion adjustment amount. The proportion adjustment amount is obtained through the following method: ; In the formula, β represents the adjustment coefficient, which is specifically used to control the adjustment amplitude; TZL represents the adjustment amount, which specifically represents the proportion adjustment amount; The evaluation iteration optimization module includes an evaluation unit; The evaluation unit marks the inertial behavior vector Hinert after executing the optimization plan Oopt as the plan execution inertial behavior vector FHinert, and then compares it with the inertial behavior vector Hinert before executing the optimization plan Oopt. After obtaining the difference, it is marked as the vector deviation factor Hh. By comparing the vector deviation factor Hh with the preset optimization trigger threshold Hopt, judge the adjustment status of the optimization plan Oopt. The adjustment status includes continuing to adjust and not adjusting; The vector deviation factor Hh is obtained through the calculation formula; The adjustment status of the optimization plan Oopt is obtained through the following comparison method: When the vector deviation factor Hh < the optimization trigger threshold Hopt, obtain that the adjustment status of the optimization plan Oopt is not to adjust, and continue to execute the optimization plan Oopt; When the vector deviation factor Hh ≥ the optimization trigger threshold Hopt, obtain that the adjustment status of the optimization plan Oopt is to continue to adjust, and generate the optimization plan Oopt twice.

[0033] In this embodiment, by extracting the historical data set Dhist, using the time series prediction model to calculate the future behavior vector Hfuture, and combining it with the inertial behavior vector Hinert to calculate the comprehensive risk vector Rrisk, it is possible to perceive the trend changes of the enterprise's business in advance, making up for the defect that the traditional system cannot give early warnings of operation risks, resulting in lagging decisions. Secondly, the solution generation unit intelligently evaluates whether optimization and adjustment are needed by comparing the comprehensive risk vector Rrisk with the preset optimization threshold Ymax, and accurately controls the adjustment range of the original enterprise operation business data by calculating the proportional adjustment amount, making the optimization solution more flexible and targeted, and effectively solving the problem that the adjustment method of the traditional system is single and the optimization amplitude is difficult to accurately match the risk degree. In addition, the evaluation unit automatically determines whether the optimization solution needs to be further adjusted by calculating the vector deviation factor Hh between the solution execution inertial behavior vector FHinert and the pre-execution inertial behavior vector Hinert and comparing it with the optimization trigger threshold Hopt, ensuring that the system does not solidify the strategy due to one-time optimization, but forms an adaptive adjustment mechanism, overcoming the disadvantages that the traditional optimization solution cannot verify the effect after adjustment and relies on manual judgment to determine whether secondary optimization is needed. Through the above mechanism, this solution can not only predict the business change trend in advance, quickly respond to anomalies and accurately optimize the operation strategy, but also perform dynamic feedback adjustment after the optimization is executed, enabling the enterprise to achieve continuous optimization and maintain the optimal operation state in the complex and changeable market environment.

[0034] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An enterprise operation supervision system based on data analysis, characterized by: It includes data acquisition and processing module, behavior inertia analysis module, abnormal deviation detection module, trend prediction and evaluation module, decision execution module and evaluation and iteration optimization module; The data collection and processing module collects enterprise operation data, performs data cleaning and data standardization preprocessing, and constructs a time series of the preprocessed enterprise operation data to form a time series data set D; The behavior inertia analysis module calculates the inertial behavior vector Hinert of each business node based on the acquired time series data set D; The abnormal deviation detection module extracts the current business features of the business node to form the current behavior vector Hcurr, and calculates based on the inertial behavior vector Hinert to obtain the behavior deviation vector Dd, and compares it with the deviation threshold Deu to obtain the deviation state result, and then triggers trend analysis based on the deviation state result; When the trend prediction and evaluation module triggers trend analysis, it uses the time series prediction model to predict and obtain the future behavior vector Hfuture, and performs a comprehensive evaluation with the inertial behavior vector Hinert to obtain the comprehensive risk vector Rrisk; The decision execution module compares the obtained comprehensive risk vector Rrisk with the preset optimization threshold Ymax to obtain the optimization solution Oopt; The evaluation iterative optimization module compares the inertial behavior vector Hinert after executing the optimization scheme Oopt to determine the adjustment status of the optimization scheme Oopt.

2. According to claim 1, the enterprise operation supervision system based on data analysis is characterized by: The data acquisition and processing module includes an acquisition unit and a processing unit; The collection unit collects original enterprise operation business data from multiple enterprise management systems and calculates the original enterprise operation business data to form an original data set Draw; Among them, the enterprise management system includes ERP system, CRM system, SCM system and OMS system. The original data set Draw includes customer transaction data Cfreq, supply chain data Sstb, order fulfillment data Ofolw, and inventory data Icons; Among them, the customer transaction data Cfreq is obtained by counting the number of transactions of customers in the past N days and calculating the average value through the OMS system and CRM system; The customer transaction data Cfreq is obtained through the following calculation formula: ； Where U represents the number of customer orders, and U(i) represents the number of customer orders on the i-th day; Supply chain data Sstb is obtained by calculating the ratio of the standard deviation to the mean of supplier delivery time through the SCM system; The supply chain data Sstb is obtained through the following calculation formula: ； In the formula, σdel represents the standard deviation of supplier delivery time, and μdel represents the mean of supplier delivery time; Order fulfillment data Ofolw obtains by calculating the average time from order creation to delivery through the OMS system; The order fulfillment data Ofolw is obtained through the following calculation formula: ； Where TcreateD represents the time from order creation to delivery, and TcreateP represents the time from order creation to payment; Inventory data Icons is obtained by calculating the ratio of inventory consumption rate to sales rate through the ERP system; The inventory data Icons are obtained through the following calculation formula: ； Where Iout represents the inventory consumption rate, specifically the daily inventory reduction, and Iin represents the inventory replenishment rate, specifically the daily storage increase.

3. The enterprise operation supervision system based on data analysis according to claim 2 is characterized by: The processing unit performs data cleaning and data standardization preprocessing on the collected original data set Draw, and then arranges the preprocessed original data set Draw in chronological order to form a time series data set D, and simultaneously stores the time series data set D as a historical data set Dhist. The time series data set D includes standard customer transaction data Cnorm, standard supply chain data Snorm, standard order fulfillment data Onorm, and standard inventory data Inorm; Among them, data cleaning includes removing missing values ​​using the interpolation method of the previous and next time points and filtering data outside 3 times the standard deviation using the Z-score; Data standardization preprocessing includes using the Z-score standardization method to normalize the data in the original data set Draw.

4. The enterprise operation supervision system based on data analysis according to claim 1 is characterized by: The behavior inertia analysis module includes a time aggregation unit and a vector generation unit; The time aggregation unit calculates the historical behavior trends of different business nodes based on the acquired time series data set D, including using an exponential decay weighted calculation method to perform weighted aggregation processing on the time series data set D at different time points, and obtains a feature vector Xw after weighted aggregation processing, wherein the feature vector Xw includes feature customer transaction data Cwe, feature supply chain data Swe, feature order fulfillment data Owe, and feature inventory data Iwe; The eigenvector Xw is obtained by the following calculation method: ； Where M represents the length of the historical time window, w(i) represents the time decay weight, specifically the decay weight at the i-th time point, and D(i, j) represents the j-th standard data in the time series data set at the i-th time point; Among them, the time decay weight w(i) is specifically , where λ represents the time decay factor, which is used to control the influence of past data.

5. The enterprise operation supervision system based on data analysis according to claim 4 is characterized by: The vector generation unit calculates the inertial behavior vector Hinert of each service node based on the acquired feature vector Xw; The inertial behavior vector Hinert includes vector customer transaction data Cin, vector supply chain data Sin, vector order fulfillment data Oin, and vector inventory data Iin; The vector customer transaction data Cin is obtained through the following calculation formula: ； Where Cwe(i) represents the characteristic customer transaction data at the i-th time point; The vector supply chain data Sin is obtained through the following calculation formula: ； In the formula, Swe(i) represents the characteristic supply chain data at the i-th time point; The vector order fulfillment data Oin is obtained through the following calculation formula: ； Where Owe(i) represents the characteristic order fulfillment data at the i-th time point; The vector inventory data Iin is obtained by the following calculation formula: ； Where Iwe(i) represents the characteristic inventory data at the i-th time point.

6. The enterprise operation supervision system based on data analysis according to claim 1 is characterized by: The abnormal deviation detection module includes a construction unit and a deviation judgment unit; The construction unit extracts the time series data set D of the current business characteristics of the business node, marks the real-time time point tnow, and forms the current behavior vector Hcurr, specifically Hcurr={tnow, Cnorm, Snorm, Onorm, Inorm}; The deviation judgment unit calculates the deviation between the current behavior vector Hcurr and the inertial behavior vector Hinert, obtains the behavior deviation vector Dd, and compares the obtained behavior deviation vector Dd with the deviation threshold Deu to obtain the deviation state result, and then triggers trend analysis based on the deviation state result.

7. The enterprise operation supervision system based on data analysis according to claim 6 is characterized by: The behavior deviation vector Dd is obtained by the following calculation formula: ； Where n represents the total number of vectors, Hinert(k) represents the kth vector in the inertial behavior vector, and Hcurr(k) represents the kth vector in the current behavior vector; The deviation status results are obtained by comparing: When the behavior deviation vector Dd ≥ the deviation threshold Deu, the deviation state result is obtained as the trigger result, triggering the trend prediction; When the behavior deviation vector Dd is less than the deviation threshold Deu, the deviation state result is obtained as a non-triggering result, and the trend prediction is not triggered.

8. The enterprise operation supervision system based on data analysis according to claim 7 is characterized by: The trend prediction and assessment module includes a risk prediction unit; When the risk prediction unit triggers trend analysis, it extracts the historical data set Dhist within a fixed period as input data, uses the time series prediction model to predict the future behavior vector Hfuture of the input data, and performs a comprehensive evaluation with the inertial behavior vector Hinert to obtain the comprehensive risk vector Rrisk; The predicted future behavior vector Hfuture is obtained by the following calculation formula: ； Where Hfuture(k) represents the predicted future behavior vector of the kth vector, Dhist(tp, k) represents the kth vector of the historical data set Dhist at the past time tp. Represents the regression coefficient, specifically the regression coefficient at the past time p, represents the error term; The comprehensive risk vector Rrisk is obtained by the following calculation formula: 。 9. The enterprise operation supervision system based on data analysis according to claim 8 is characterized by: The decision execution module includes a solution generation unit; The solution generation unit compares the obtained comprehensive risk vector Rrisk with the preset optimization threshold Ymax to obtain the comprehensive risk optimization evaluation result, and obtains the optimization solution Oopt according to the comprehensive risk optimization evaluation result; When the comprehensive risk vector Rrisk is less than the optimization threshold Ymax, the comprehensive risk optimization assessment result is obtained as an unadjusted result, and the optimization scheme Oopt is not generated; When the comprehensive risk vector Rrisk≥optimization threshold Ymax, obtain the adjustment result of the comprehensive risk optimization assessment result, and generate the optimization plan Oopt, including calculating the comprehensive risk vector Rrisk to obtain the proportional adjustment amount to adjust the original enterprise operation business data, and notify the relevant enterprise management system of the occurrence of the optimization plan Oopt.

10. The enterprise operation supervision system based on data analysis according to claim 1 is characterized by: The evaluation iterative optimization module includes an evaluation unit; The evaluation unit marks the inertial behavior vector Hinert after executing the optimization scheme Oopt as the scheme execution inertial behavior vector FHinert, and then compares it with the inertial behavior vector Hinert before executing the optimization scheme Oopt, obtains the difference and marks it as the vector deviation factor Hh, and compares the vector deviation factor Hh with the preset optimization trigger threshold Hopt to judge the adjustment state of the optimization scheme Oopt, and the adjustment state includes continuing adjustment and not adjusting; The adjustment status of the optimization scheme Oopt is obtained by the following comparison method: When the vector deviation factor Hh is less than the optimization trigger threshold Hopt, the adjustment state of the optimization scheme Oopt is obtained as no adjustment, and the optimization scheme Oopt is continued to be executed; When the vector deviation factor Hh≥optimization trigger threshold Hopt, the adjustment state of the optimization scheme Oopt is obtained as continuing adjustment, and the optimization scheme Oopt is generated for the second time.

Images