Metering asset inventory change recording and analyzing system and method based on big data

Abstract

The invention discloses a metering asset inventory change recording and analyzing system and method based on big data, and relates to the technical field of big data, real-time inventory and historical usage are comprehensively obtained through the Internet of Things and a data management platform, and industry dynamic information and meteorological data are introduced to form an initial feature vector set Fd; and basic data support is provided for the model. And by performing data verification and default value processing on the initial feature vector set Fd, a more accurate processed feature vector set Fp is generated, and the integrity and consistency of the data are ensured. And then, the scheme predicts a future inventory demand trend by using a linear regression model, generates a trend prediction vector Tpred, performs modeling on a residual sequence of trend prediction through an ARIMA model, and finally generates a prediction comprehensive index Pf in combination with a prediction residual value YR and seasonal adjustment, so as to accurately reflect the inventory demand in a period of time in the future. And high precision and adaptability of inventory demand prediction are realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of big data, and in particular relates to a system and method for recording and analyzing inventory changes of metering assets based on big data. Background Art

[0002] In the context of the rapid development of information technology today, big data technology has become one of the important tools for many industries to achieve digital management. By using big data analysis, companies can dig out key trends and valuable decision-making basis from massive data. This technology is particularly widely used in the field of asset management, especially in the management of metering assets. Big data technology helps companies better control inventory changes and asset usage. Specifically, in the inventory management of metering assets, big data analysis can be used to monitor the status of various assets in real time and predict future usage needs.

[0003] However, many current methods of inventory management for measured assets are unable to cope with complex and changing external factors. Most traditional inventory management and forecasting algorithms rely on simple trends and patterns in historical data, and it is difficult to take into account the impact of external environments such as seasons, industry development or weather changes on asset demand. Existing inventory change recording and analysis systems usually only focus on the asset's usage history and internal demand patterns, while ignoring the dynamic changes of external factors. For example, in a fixed forecasting model, some assets with significant seasonal usage fluctuations often fail to adjust inventory in a timely manner due to external influences, resulting in inaccurate forecasts and excessive or insufficient inventory. This will not only increase the difficulty of inventory management, but may also bring unnecessary financial pressure and waste of resources to the company. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a system and method for recording and analyzing inventory changes of metering assets based on big data, which solves the problems mentioned in the background technology.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: A method for recording and analyzing inventory changes of measurement assets based on big data, comprising the following steps:

[0006] S1. Obtain real-time inventory data and historical usage data of metering assets through the Internet of Things and data management platform, and simultaneously obtain industry dynamic information and meteorological data to form the initial feature vector set Fd;

[0007] S2, perform data verification on the initial feature vector set Fd, including checking the historical usage data, and then determine the usage records for default processing to form a processed feature vector set Fp;

[0008] S3. Analyze the feature vector set Fp based on the linear regression model, predict the inventory demand trend in the future, obtain the inventory demand trend Tp, and integrate the inventory demand trend Tp to form a trend prediction vector Tpred;

[0009] S4. Calculate the residual value between the trend prediction vector Tpred and the actual inventory SI, obtain the residual coefficient R, and use the ARIMA model to perform sequence modeling on the residual coefficient R, determine the stable state of the residual coefficient R, and obtain the predicted residual value YR based on the stable state;

[0010] S5, the forecast residual value YR and the trend forecast vector Tpred are combined, and combined with seasonal adjustment to generate a forecast comprehensive index Pf;

[0011] S6. Compare the generated forecast comprehensive index Pf with historical data to obtain the comprehensive error evaluation index Et, match it with the preset total error exceeding threshold WZ, and iteratively optimize the ARIMA model and linear regression model.

[0012] Preferably, said S1 includes S11 and S12;

[0013] S11. Obtain the real-time inventory data of the metering assets at time t and the historical usage data of several time periods through the Internet of Things and the data management platform. By extracting the real-time inventory data, obtain the inventory quantity I(t)={I(t1), I(t2), I(t3), ..., I(tn)} from time t1 to time tn. By extracting the historical usage data, obtain the historical usage U(t)={U(t1), U(t2), U(t3), ..., U(tn)} from time t1 to time tn.

[0014] S12. Obtain industry dynamic information and meteorological data through external data sources. The industry dynamic information is obtained by crawling the changing information of industry demand at time t on the Internet using crawler technology, which is marked as the industry dynamic index Dind(t) to reflect the demand trend of the industry market for metering assets. The meteorological data is obtained by connecting to the meteorological platform to obtain climate information at time t, which is marked as meteorological data Text(t). The inventory I(t), historical usage U(t), industry dynamic index Dind(t) and meteorological data Text(t) are simultaneously integrated to form an initial feature vector set Fd={I(t), U(t), Dind(t), Text(t)}.

[0015] Preferably, said S2 includes S21;

[0016] S21. Perform data validity check on the initial feature vector set Fd, including checking the historical usage U(t), specifically checking whether the historical usage data exists within a fixed time period. When there is no usage record for several consecutive time periods, the predicted value is set to zero to eliminate the impact of invalid data. The specific verification logic is to check the usage records within three time periods. When there is no usage data in the three time periods, the historical usage U(t) is default processed, including marking the usage as 0, and synchronously marking the verified historical usage U(t) as the verified historical usage Up(t), and re-substituting it into the initial feature vector set Fd={I(t), Up(t), Dind(t, Text(t)}.

[0017] Preferably, said S3 includes S31 and S32;

[0018] S31. Based on the verified feature vector set Fp as the input data of the established linear regression model, by establishing the relationship between time t and the industry dynamic index Dind(t) and meteorological data Text(t) in the input data feature vector set Fp, and by correlating the industry dynamic index Dind(t) and meteorological data Text(t) in the linear regression model to jointly influence the trend of inventory demand, the inventory demand trend in the future is predicted to obtain the inventory demand trend Tp.

[0019] Preferably, S32, after obtaining the inventory demand trend Tp at different times t by calculation, an integration process is performed to generate a smooth trend prediction vector Tpred, including using a moving average method to smooth the predicted inventory demand trend Tp at different times t and adjust the impact of short-term fluctuations on long-term trends.

[0020] Preferably, said S4 includes S41 and S42;

[0021] S41. The residual coefficient R is obtained by calculating the residual value between the trend prediction vector Tpred and the actual inventory SI. The residual coefficient R is used to reflect the error of the linear regression model, which specifically reflects the deviation ratio between the trend prediction result and the time inventory. The residual coefficient R of several times t is calculated, that is, the time inventory SI is subtracted from the trend prediction vector Tpred, to generate a complete residual sequence vector XR={R(t1), R(t2), ..., R(tn)} of the residual coefficient R, which records the prediction error information of the linear regression model at different times t, and is then used to compensate for the error part in the trend prediction.

[0022] Preferably, S42, stationarity detection and ARIMA model building are performed according to the residual sequence vector XR, and the stationarity of the residual sequence vector XR is judged. When the residual sequence vector XR is judged to be non-stationary, it is converted into a stationary sequence through differential processing, and the discontinuous residual sequence vector XR converted into a stationary sequence is obtained, and the ARIMA model is applied for modeling. The ARIMA model modeling captures the time series characteristics of the discontinuous residual sequence vector XR of the stationary sequence through the combination of three components: natural regression, difference and moving average, and generates a predicted residual value YR={YR(t1), YR(t2),...,YR(tn)} within a future fixed time period.

[0023] Preferably, S5, the forecast residual value YR and the trend forecast vector Tpred are combined to generate a preliminary inventory forecast result vector Pb(t) at time t, and seasonal adjustment is performed in combination with the meteorological data Text(t) in the initial feature vector set Fd to generate a forecast comprehensive index Pf.

[0024] Preferably, said S6 includes S61 and S62;

[0025] S61, compare the generated forecast comprehensive index Pf with historical data to obtain a comprehensive error evaluation index Et,

[0026] And it is matched with the preset total error exceeding threshold WZ. When the comprehensive error evaluation index Et is greater than the total error exceeding threshold WZ, it is determined that the ARIMA model and the linear regression model need to be iteratively optimized. When the comprehensive error evaluation index Et is less than or equal to the total error exceeding threshold WZ, the ARIMA model and the linear regression model are not iteratively optimized.

[0027] A measurement asset inventory change recording and analysis system based on big data, including a data acquisition module, a feature extraction module, a trend prediction module, a residual judgment module, a season fitting module and an iterative optimization module;

[0028] The data acquisition module obtains real-time inventory data and historical usage data of metering assets through the Internet of Things and the data management platform, and simultaneously obtains industry dynamic information and meteorological data to form an initial feature vector set Fd;

[0029] The feature extraction module performs data verification on the initial feature vector set Fd, including checking the historical usage data, and then determines the usage records for default processing to form a processed feature vector set Fp;

[0030] The trend prediction module analyzes the feature vector set Fp based on the linear regression model, predicts the inventory demand trend in the future, obtains the inventory demand trend Tp, and integrates the inventory demand trend Tp to form a trend prediction vector Tpred;

[0031] The residual judgment module calculates the residual value between the trend prediction vector Tpred and the actual inventory SI, obtains the residual coefficient R, and uses the ARIMA model to perform sequence modeling on the residual coefficient R, determines the stable state of the residual coefficient R, and obtains the predicted residual value YR according to the stable state;

[0032] The seasonal fitting module combines the prediction residual value YR and the trend prediction vector Tpred, and combines seasonal adjustment to generate a prediction comprehensive index Pf;

[0033] The iterative optimization module compares the generated forecast comprehensive index Pf with historical data to obtain the comprehensive error evaluation index Et, and matches it with the preset total error exceeding threshold WZ to iteratively optimize the ARIMA model and the linear regression model.

[0034] Compared with the prior art, the beneficial effects are:

[0035] (1) Through the Internet of Things and data management platform, the real-time inventory and historical usage are fully obtained, and industry dynamic information and meteorological data are introduced to form the initial feature vector set Fd to provide basic data support for the model. By performing data verification and default value processing on the initial feature vector set Fd, a more accurate processed feature vector set Fp is generated to ensure data integrity and consistency. Subsequently, the solution uses a linear regression model to predict future inventory demand trends, generates a trend prediction vector Tpred, and models the residual sequence of trend prediction through the ARIMA model. Combined with the predicted residual value YR and seasonal adjustment, it finally generates a forecast comprehensive index Pf, which accurately reflects the inventory demand in the future. High accuracy and adaptability of inventory demand forecasting are achieved. While improving the accuracy of forecasts, the solution fully considers the impact of industry dynamics and meteorological data, making inventory management more forward-looking and able to respond more promptly to market demand fluctuations and climate seasonal changes, avoiding the common defects of large forecast deviations and untimely adjustments in traditional inventory management.

[0036] (2) Through comprehensive multivariate regression and time series analysis, high-precision trend forecasting and residual correction of inventory demand are achieved. This process first uses the industry dynamic index Dind(t) and meteorological data Text(t) in the verified feature vector set Fp to establish a linear regression model to dynamically capture the impact of industry demand and climate change on inventory trends, thereby obtaining the inventory demand trend Tp and generating a trend forecast vector Tpred. This trend forecast not only takes into account the time change trend of actual inventory demand, but also smoothes short-term fluctuations to make long-term trends more stable. This makes the inventory forecast results more accurate and solves the problem that traditional methods cannot handle the impact of industry volatility and climate change on inventory. This solution significantly improves the stability and accuracy of the forecast, providing enterprises with more reliable inventory demand trend analysis and dynamic adjustment support.

[0037] (3) The trend prediction vector Tpred and the prediction residual value YR are combined to generate the preliminary inventory prediction result vector Pb(t), and seasonal adjustment is performed in combination with the meteorological data Text(t) in the initial feature vector set Fd to generate a forecast comprehensive index Pf that is more in line with actual demand fluctuations. Seasonal adjustment improves the accuracy and adaptability of inventory forecasts by combining the sine wave function with meteorological data. By comparing the forecast comprehensive index Pf with historical data, the comprehensive error evaluation index Et is obtained and matched with the preset total error threshold WZ. The ARIMA model and the linear regression model are iteratively optimized to further improve the forecast accuracy and ensure that the model has stable accuracy; the model error is dynamically controlled on the basis of being able to adapt to seasonal changes and multi-factor fluctuations, so that the forecast results not only adapt to cyclical demand, but also can be autonomously iterated and optimized according to the error situation. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a schematic diagram of the steps of a method for recording and analyzing inventory changes of measurement assets based on big data according to the present invention;

[0039] Figure 2 This is a schematic diagram of a block diagram of a measurement asset inventory change recording and analysis system based on big data according to the present invention. DETAILED DESCRIPTION

[0040] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0041] Example 1

[0042] The present invention provides a method for recording and analyzing inventory changes of measurement assets based on big data. Figure 1 , including the following steps:

[0043] S1. Obtain real-time inventory data and historical usage data of metering assets through the Internet of Things and data management platform, and simultaneously obtain industry dynamic information and meteorological data to form the initial feature vector set Fd;

[0044] S2, perform data verification on the initial feature vector set Fd, including checking the historical usage data, and then determine the usage records for default processing to form a processed feature vector set Fp;

[0045] S3. Analyze the feature vector set Fp based on the linear regression model, predict the inventory demand trend in the future, obtain the inventory demand trend Tp, and integrate the inventory demand trend Tp to form a trend prediction vector Tpred;

[0046] S4. Calculate the residual value between the trend prediction vector Tpred and the actual inventory SI, obtain the residual coefficient R, and use the ARIMA model to perform sequence modeling on the residual coefficient R, determine the stable state of the residual coefficient R, and obtain the predicted residual value YR based on the stable state;

[0047] S5, the forecast residual value YR and the trend forecast vector Tpred are combined, and combined with seasonal adjustment to generate a forecast comprehensive index Pf;

[0048] S6. Compare the generated forecast comprehensive index Pf with historical data to obtain the comprehensive error evaluation index Et, match it with the preset total error exceeding threshold WZ, and iteratively optimize the ARIMA model and linear regression model.

[0049] In this embodiment, the real-time inventory and historical usage are fully obtained through the Internet of Things and the data management platform, and industry dynamic information and meteorological data are introduced to form the initial feature vector set Fd to provide basic data support for the model. By performing data verification and default value processing on the initial feature vector set Fd, a more accurate processed feature vector set Fp is generated to ensure the integrity and consistency of the data. Subsequently, the scheme uses a linear regression model to predict future inventory demand trends, generates a trend prediction vector Tpred, and models the residual sequence of trend prediction through the ARIMA model, combines the predicted residual value YR and seasonal adjustment, and finally generates a predicted comprehensive index Pf, which accurately reflects the inventory demand in the future. High precision and adaptability of inventory demand forecasting are achieved. While improving the accuracy of prediction, the scheme fully considers the impact of industry dynamics and meteorological data, making inventory management more forward-looking, able to respond more promptly to market demand fluctuations and climate seasonal changes, and avoiding the common defects of large prediction deviations and untimely adjustments in traditional inventory management.

[0050] Example 2

[0051] This embodiment is explained in Example 1, please refer to Figure 1 , specifically: S1 includes S11 and S12;

[0052] S11. Obtain the real-time inventory data of the metering assets at time t and the historical usage data of several time periods through the Internet of Things and the data management platform. By extracting the real-time inventory data, obtain the inventory quantity I(t)={I(t1), I(t2), I(t3), ..., I(tn)} from time t1 to time tn. By extracting the historical usage data, obtain the historical usage U(t)={U(t1), U(t2), U(t3), ..., U(tn)} from time t1 to time tn.

[0053] S12. Obtain industry dynamic information and meteorological data through external data sources. The industry dynamic information is obtained by crawling the changing information of the network industry demand at time t using crawler technology, marked as the industry dynamic index Dind(t), reflecting the demand trend of the industry market for metering assets. The meteorological data obtains the climate information at time t by connecting to the meteorological platform, marked as meteorological data Text(t). The inventory I(t), historical usage U(t), industry dynamic index Dind(t) and meteorological data Text(t) are simultaneously integrated to form an initial feature vector set Fd={I(t), U(t), Dind(t), Text(t)}.

[0054] The industry dynamics index Dind(t) collects data from industry reports published on industry exchange websites, market research websites, and industry databases by using API or web crawler technology. In addition, it combines the big data analysis platform to process large-scale data from different sources, integrates and extracts representative indicators to generate an industry demand index.

[0055] Among them, the industry dynamic index Dind(t) is obtained by the following calculation formula:

[0056]

[0057] Where W represents the total number of data sources, including industry reports and market research, w represents the w-th data source, Vw(t) represents the industry demand value from the w-th data source at time t, Vmax and Vmin represent the maximum and minimum values ​​in the total number of data sources W, respectively;

[0058] Meteorological data Text(t) connects to the API interface of real-time and historical meteorological data of various places, and the system obtains meteorological data by calling these interfaces. In addition, the meteorological data is updated through regular push by data suppliers to obtain information on temperature, humidity, wind speed and seasonal changes in the area;

[0059] Among them, the meteorological data Text(t) is obtained by the following calculation formula:

[0060] Text(t)=e1*T(t)+e2*H(t)+P(t);

[0061] In the formula, T(t) represents the temperature data at time t, H(t) represents the humidity data at time t, P(t) represents the wind speed data at time t, and e1, e2 and e3 represent the preset weight values ​​of the temperature data T(t) at time t, the humidity data H(t) at time t and the wind speed data P(t) at time t, respectively.

[0062] S2 includes S21;

[0063] S21. Check the data validity of the initial feature vector set Fd, including checking the historical usage U(t), specifically checking whether the historical usage data exists within a fixed time period. When there is no usage record for several consecutive time periods, the predicted value is set to zero to eliminate the influence of invalid data. The specific verification logic is to check the usage records within three time periods. When there is no usage data in the three time periods, the historical usage U(t) is defaulted, including marking the usage as 0, and synchronously marking the verified historical usage U(t) as the verified historical usage Up(t), and re-substituting it into the initial feature vector set Fd={I(t), Up(t), Dind(t, Text(t)};

[0064] The verification formula is as follows:

[0065]

[0066] In the formula, U(t-1)=U(t-2)=U(t-3)=0 represents the verification condition, indicating that there is no usage record in the last three cycles.

[0067] S2. Perform data verification on the initial feature vector set Fd, including checking the historical usage data, and then determine the usage records for default processing to form a processed feature vector set Fp.

[0068] In this embodiment, the real-time inventory data I(t) and the historical usage data U(t) within a certain time period are automatically obtained through the Internet of Things and the data management platform, and the industry dynamic index Dind(t) and the meteorological data Text(t) are obtained through the API or web crawler technology combined with the big data analysis platform. This step not only obtains the changing trend of the industry dynamics, but also combines the influence of meteorological factors, and fully reflects the fluctuations of external environmental factors that inventory demand may be affected by. Further through data validity checking and default value processing, the credibility of the data is improved to ensure that invalid data in the data source does not affect the model prediction. Through the fixed period verification of the historical usage data U(t), when there is no usage record within three periods, the predicted value is set to zero, marked as the verification of the historical usage Up(t), and the complete initial feature vector set Fd is reconstructed, which not only ensures the diversity and external adaptability of the data, but also makes the data more accurate through the verification mechanism, especially in the case of drastic changes in industry demand and climate environment, it can eliminate the interference of invalid data on the inventory forecasting model, and provide a solid data foundation for the stability and accuracy of the subsequent model.

[0069] Example 3

[0070] This embodiment is explained in Example 2. Please refer to Figure 1 , specifically: S3 includes S31 and S32;

[0071] S31. Based on the verified feature vector set Fp as the input data of the established linear regression model, by establishing the relationship between time t and the industry dynamic index Dind(t) and meteorological data Text(t) in the input data feature vector set Fp, and by correlating the industry dynamic index Dind(t) and meteorological data Text(t) in the linear regression model to jointly influence the trend of inventory demand, the inventory demand trend in the future is predicted to obtain the inventory demand trend Tp.

[0072] The inventory demand trend Tp is obtained through the following multivariate linear regression equation calculation formula:

[0073] Tp=α0+α1*t+α2*Dind(t)+α3*Text(t);

[0074] In the formula, α0 represents the constant term of the multivariate linear regression equation, α1 represents the regression coefficient of the time factor, specifically representing the changing trend of the inventory demand trend Tp with time t, α2 and α3 represent the regression coefficients of the industry dynamic index Dind(t) and the meteorological data Text(t), respectively, and 0<α1+α2+α3≤1, and the specific value is set by the user.

[0075] S32. After obtaining the inventory demand trend Tp at different times t by calculation, an integration process is performed to generate a smooth trend prediction vector Tpred, including using a moving average method to smooth the predicted inventory demand trend Tp at different times t and adjust the impact of short-term fluctuations on long-term trends.

[0076] S4 includes S41 and S42;

[0077] S41. The residual coefficient R is obtained by calculating the residual value between the trend prediction vector Tpred and the actual inventory SI. The residual coefficient R is used to reflect the error of the linear regression model, which specifically reflects the deviation ratio between the trend prediction result and the time inventory. The residual coefficient R of several times t is calculated, that is, the time inventory SI is subtracted from the trend prediction vector Tpred, to generate a complete residual sequence vector XR={R(t1), R(t2), ..., R(tn)} of the residual coefficient R, which records the prediction error information of the linear regression model at different times t, and is then used to compensate for the error part in the trend prediction.

[0078] S42. Perform stationarity detection and ARIMA model building based on the residual sequence vector XR. By judging the stationarity of the residual sequence vector XR, when it is judged that the residual sequence vector XR is not stationary, convert it into a stationary sequence through differential processing, obtain the discontinuous residual sequence vector XR converted into a stationary sequence, and apply the ARIMA model for modeling. The ARIMA model building combines the three components of natural regression, difference and moving average to capture the time series characteristics of the discontinuous residual sequence vector XR of the stationary sequence, and generate the predicted residual value YR={YR(t1), YR(t2), ..., YR(tn)} in the future fixed time period.

[0079] In this embodiment, high-precision trend prediction and residual correction of inventory demand are achieved through comprehensive multivariate regression and time series analysis. This process first uses the industry dynamic index Dind(t) and meteorological data Text(t) in the verified feature vector set Fp to establish a linear regression model to dynamically capture the impact of industry demand and climate change on inventory trends, thereby obtaining the inventory demand trend Tp and generating a trend prediction vector Tpred. This trend prediction not only takes into account the time change trend of actual inventory demand, but also smoothes short-term fluctuations to make long-term trends more stable. This makes the inventory forecast results more accurate and solves the problem that traditional methods cannot handle the impact of industry volatility and climate change on inventory. This solution significantly improves the stability and accuracy of the forecast, and provides enterprises with more reliable inventory demand trend analysis and dynamic adjustment support.

[0080] Example 4

[0081] This embodiment is explained in Example 3, please refer to Figure 1 Specifically: S5. The forecast residual value YR and the trend forecast vector Tpred are combined to generate the preliminary inventory forecast result vector Pb(t) at time t, and seasonally adjusted in combination with the meteorological data Text(t) in the initial feature vector set Fd to generate the forecast comprehensive index Pf as the forecast value of the final inventory demand, which can more accurately reflect the inventory demand trend in the future and provide a key reference for inventory management and decision-making.

[0082] The forecast comprehensive index Pf is obtained by the calculation formula Pf = Pb(t) + S(t), where S(t) represents the seasonal adjustment at time t;

[0083] The seasonal adjustment S(t) at time t is obtained by the following calculation formula:

[0084]

[0085] In the formula, b represents the seasonal adjustment coefficient, which specifically represents the amplitude of the sine wave, and adjusts the degree of influence of seasonal fluctuations. T represents the length of the seasonal cycle, which specifically represents the fluctuation cycle, including settings of 12 months or 4 quarters. β represents the influence coefficient of meteorological data, which specifically represents the degree of direct influence of temperature or other seasonal factors on inventory demand.

[0086] S6 includes S61 and S62;

[0087] S61. Compare the generated forecast comprehensive index Pf with the historical data to obtain the comprehensive error evaluation index Et, and match it with the preset total error exceeding threshold WZ. When the comprehensive error evaluation index Et is greater than the total error exceeding threshold WZ, it is determined that the ARIMA model and the linear regression model need to be iteratively optimized. When the comprehensive error evaluation index Et is less than or equal to the total error exceeding threshold WZ, the ARIMA model and the linear regression model are not iteratively optimized.

[0088] The comprehensive error evaluation index Et is obtained by the following calculation formula:

[0089]

[0090] In the formula, n represents the constant of the time series, and t represents time t.

[0091] In this embodiment, the trend prediction vector Tpred and the prediction residual value YR are combined to generate a preliminary inventory prediction result vector Pb(t), and seasonal adjustment is performed in combination with the meteorological data Text(t) in the initial feature vector set Fd to generate a forecast comprehensive index Pf that is more in line with actual demand fluctuations. Seasonal adjustment improves the accuracy and adaptability of inventory forecasts by combining the sine wave function with meteorological data. By comparing the forecast comprehensive index Pf with historical data, the comprehensive error evaluation index Et is obtained and matched with the preset total error threshold WZ. The ARIMA model and the linear regression model are iteratively optimized to further improve the prediction accuracy and ensure that the model has stable accuracy; the dynamic control of the model error is achieved on the basis of being able to adapt to seasonal changes and multi-factor fluctuations, so that the prediction results not only adapt to periodic demand, but also can be autonomously iterated and optimized according to the error situation.

[0092] Example 5

[0093] A system for recording and analyzing inventory changes of measurement assets based on big data, please refer to Figure 2 ,Specifically: including data acquisition module, feature extraction module, trend prediction module, residual judgment module, season fitting module and iterative optimization module;

[0094] The data acquisition module obtains the real-time inventory data and historical usage data of metering assets through the Internet of Things and data management platform, and simultaneously obtains industry dynamic information and meteorological data to form the initial feature vector set Fd;

[0095] The feature extraction module performs data verification on the initial feature vector set Fd, including checking the historical usage data, and then determines the usage records for default processing to form a processed feature vector set Fp;

[0096] The trend prediction module analyzes the feature vector set Fp based on the linear regression model, predicts the inventory demand trend in the future, obtains the inventory demand trend Tp, and integrates the inventory demand trend Tp to form a trend prediction vector Tpred;

[0097] The residual judgment module calculates the residual value between the trend prediction vector Tpred and the actual inventory SI, obtains the residual coefficient R, and uses the ARIMA model to perform sequence modeling on the residual coefficient R, determines the stable state of the residual coefficient R, and obtains the predicted residual value YR based on the stable state;

[0098] The seasonal fitting module combines the forecast residual value YR and the trend forecast vector Tpred, and combines it with seasonal adjustment to generate the forecast comprehensive index Pf;

[0099] The iterative optimization module compares the generated forecast comprehensive index Pf with historical data to obtain the comprehensive error evaluation index Et, and matches it with the preset total error exceeding threshold WZ to iteratively optimize the ARIMA model and linear regression model.

[0100] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for recording and analyzing inventory changes of measurement assets based on big data, characterized in that: The following steps are involved: S1. Obtain real-time inventory data and historical usage data of metering assets through the Internet of Things and data management platform, and simultaneously obtain industry dynamic information and meteorological data to form the initial feature vector set Fd; S2, perform data verification on the initial feature vector set Fd, including checking the historical usage data, and then determine the usage records for default processing to form a processed feature vector set Fp; S3. Analyze the feature vector set Fp based on the linear regression model, predict the inventory demand trend in the future, obtain the inventory demand trend Tp, and integrate the inventory demand trend Tp to form a trend prediction vector Tpred; S4. Calculate the residual value between the trend prediction vector Tpred and the actual inventory SI, obtain the residual coefficient R, and use the ARIMA model to perform sequence modeling on the residual coefficient R, determine the stable state of the residual coefficient R, and obtain the predicted residual value YR based on the stable state; S5, the forecast residual value YR and the trend forecast vector Tpred are combined, and combined with seasonal adjustment to generate a forecast comprehensive index Pf; S6. Compare the generated forecast comprehensive index Pf with historical data to obtain the comprehensive error evaluation index Et, match it with the preset total error exceeding threshold WZ, and iteratively optimize the ARIMA model and linear regression model.

2. According to the method for recording and analyzing changes in inventory of measurement assets based on big data according to claim 1, it is characterized in that: Said S1 includes S11 and S12; S11. Obtain the real-time inventory data of the metering assets at time t and the historical usage data of several time periods through the Internet of Things and the data management platform. By extracting the real-time inventory data, obtain the inventory quantity I(t)={I(t1), I(t2), I(t3), ..., I(tn)} from time t1 to time tn. By extracting the historical usage data, obtain the historical usage U(t)={U(t1), U(t2), U(t3), ..., U(tn)} from time t1 to time tn. S12. Obtain industry dynamic information and meteorological data through external data sources. The industry dynamic information is obtained by crawling the changing information of industry demand at time t on the Internet using crawler technology, which is marked as the industry dynamic index Dind(t) to reflect the demand trend of the industry market for metering assets. The meteorological data is obtained by connecting to the meteorological platform to obtain climate information at time t, which is marked as meteorological data Text(t). The inventory I(t), historical usage U(t), industry dynamic index Dind(t) and meteorological data Text(t) are simultaneously integrated to form an initial feature vector set Fd={I(t), U(t), Dind(t), Text(t)}.

3. The method for recording and analyzing inventory changes of measurement assets based on big data according to claim 2 is characterized in that: Said S2 includes S21; S21. Perform data validity check on the initial feature vector set Fd, including checking the historical usage U(t), specifically checking whether the historical usage data exists within a fixed time period. When there is no usage record for several consecutive time periods, the predicted value is set to zero to eliminate the impact of invalid data. The specific verification logic is to check the usage records within three time periods. When there is no usage data in the three time periods, the historical usage U(t) is default processed, including marking the usage as 0, and synchronously marking the verified historical usage U(t) as the verified historical usage Up(t), and re-substituting it into the initial feature vector set Fd={I(t), Up(t), Dind(t, Text(t)}.

4. The method for recording and analyzing inventory changes of measurement assets based on big data according to claim 3 is characterized in that: The S3 includes S31 and S32; S31. Based on the verified feature vector set Fp as the input data of the established linear regression model, by establishing the relationship between time t and the industry dynamic index Dind(t) and meteorological data Text(t) in the input data feature vector set Fp, and by correlating the industry dynamic index Dind(t) and meteorological data Text(t) in the linear regression model to jointly influence the trend of inventory demand, the inventory demand trend in the future is predicted to obtain the inventory demand trend Tp.

5. The method for recording and analyzing inventory changes of measurement assets based on big data according to claim 4 is characterized in that: S32. After obtaining the inventory demand trend Tp at different times t by calculation, an integration process is performed to generate a smooth trend prediction vector Tpred, including using a moving average method to smooth the predicted inventory demand trend Tp at different times t and adjust the impact of short-term fluctuations on long-term trends.

6. A method for recording and analyzing inventory changes of measurement assets based on big data according to claim 5, characterized in that: The S4 includes S41 and S42; S41. The residual coefficient R is obtained by calculating the residual value between the trend prediction vector Tpred and the actual inventory SI. The residual coefficient R is used to reflect the error of the linear regression model, which specifically reflects the deviation ratio between the trend prediction result and the time inventory. The residual coefficient R of several times t is calculated, that is, the time inventory SI is subtracted from the trend prediction vector Tpred, to generate a complete residual sequence vector XR={R(t1), R(t2), ..., R(tn)} of the residual coefficient R, which records the prediction error information of the linear regression model at different times t, and is then used to compensate for the error part in the trend prediction.

7. The method for recording and analyzing inventory changes of measurement assets based on big data according to claim 6 is characterized in that: S42. Perform stationarity detection and ARIMA model building based on the residual sequence vector XR. By judging the stationarity of the residual sequence vector XR, when it is judged that the residual sequence vector XR is not stationary, convert it into a stationary sequence through differential processing, obtain the discontinuous residual sequence vector XR converted into a stationary sequence, and apply the ARIMA model for modeling. The ARIMA model building combines the three components of natural regression, difference and moving average to capture the time series characteristics of the discontinuous residual sequence vector XR of the stationary sequence, and generate the predicted residual value YR={YR(t1), YR(t2), ..., YR(tn)} in the future fixed time period.

8. The method for recording and analyzing inventory changes of measurement assets based on big data according to claim 7 is characterized in that: S5. The forecast residual value YR and the trend forecast vector Tpred are combined to generate a preliminary inventory forecast result vector Pb(t) at time t, and seasonal adjustment is performed in combination with the meteorological data Text(t) in the initial feature vector set Fd to generate a forecast comprehensive index Pf.

9. The method for recording and analyzing inventory changes of measurement assets based on big data according to claim 8 is characterized in that: The S6 includes S61 and S62; S61, compare the generated forecast comprehensive index Pf with historical data to obtain a comprehensive error evaluation index Et, And it is matched with the preset total error exceeding threshold WZ. When the comprehensive error evaluation index Et is greater than the total error exceeding threshold WZ, it is determined that the ARIMA model and the linear regression model need to be iteratively optimized. When the comprehensive error evaluation index Et is less than or equal to the total error exceeding threshold WZ, the ARIMA model and the linear regression model are not iteratively optimized.

10. A system for recording and analyzing inventory changes of measurement assets based on big data, applied to a method for recording and analyzing inventory changes of measurement assets based on big data as claimed in any one of claims 1 to 9, characterized in that: It includes data acquisition module, feature extraction module, trend prediction module, residual judgment module, season fitting module and iterative optimization module; The data acquisition module obtains real-time inventory data and historical usage data of metering assets through the Internet of Things and the data management platform, and simultaneously obtains industry dynamic information and meteorological data to form an initial feature vector set Fd; The feature extraction module performs data verification on the initial feature vector set Fd, including checking the historical usage data, and then determines the usage records for default processing to form a processed feature vector set Fp; The trend prediction module analyzes the feature vector set Fp based on the linear regression model, predicts the inventory demand trend in the future, obtains the inventory demand trend Tp, and integrates the inventory demand trend Tp to form a trend prediction vector Tpred; The residual judgment module calculates the residual value between the trend prediction vector Tpred and the actual inventory SI, obtains the residual coefficient R, and uses the ARIMA model to perform sequence modeling on the residual coefficient R, determines the stable state of the residual coefficient R, and obtains the predicted residual value YR according to the stable state; The seasonal fitting module combines the prediction residual value YR and the trend prediction vector Tpred, and combines seasonal adjustment to generate a prediction comprehensive index Pf; The iterative optimization module compares the generated forecast comprehensive index Pf with historical data to obtain the comprehensive error evaluation index Et, and matches it with the preset total error exceeding threshold WZ to iteratively optimize the ARIMA model and the linear regression model.

Images