CN114819777A

CN114819777A - Enterprise sales business analysis and management system based on digital twin technology

Info

Publication number: CN114819777A
Application number: CN202210756666.4A
Authority: CN
Inventors: 刘德利; 郭靖
Original assignee: Huitongda Network Co ltd
Current assignee: Huitongda Network Co ltd
Priority date: 2022-06-30
Filing date: 2022-06-30
Publication date: 2022-07-29

Abstract

The invention provides an enterprise sales business analysis management system based on a digital twin technology, and relates to the technical field of sales business analysis. According to the method and the device, data relation mining processing is carried out through correlation analysis according to historical data with poor sales performance, a sales performance model is built according to functional relations between the sales performance and various influence factors, a virtual digital twin model is built based on the sales performance model, interaction and mapping between a physical scene and the virtual digital twin model are built, the digital twin model is trained according to data collected by a sensor in the physical scene in real time, and the trained digital twin model is obtained. Based on the trained digital twin model, real-time data is used as an input value, the sales disadvantage model can be output in real time, the sales service can be analyzed conveniently, and the response rate and accuracy of the sales service analysis management system are improved.

Description

Enterprise sales business analysis and management system based on digital twin technology

Technical Field

The invention relates to the technical field of sales business analysis, in particular to an enterprise sales business analysis management system based on a digital twin technology.

Background

The quality of the sales service is influenced by various factors, the situation is more complex in the process of analyzing the sales service, and the influence degrees of the same factor on different situations are different. However, the existing analysis method has the problems of incomplete investigation factors, no speciality and the like, and the obtained analysis conclusion is not accurate enough.

Disclosure of Invention

The invention aims to provide an enterprise sales business analysis management system based on a digital twin technology so as to improve the problems. In order to achieve the purpose, the technical scheme adopted by the invention is as follows:

in a first aspect, the present application provides an enterprise sales business analysis management system based on a digital twin technology, including:

obtaining historical data and real-time data of sales;

performing data relation mining processing on the historical data based on correlation analysis to construct a sales disadvantage model, wherein the sales disadvantage model is a sales model with a performance level lower than a first preset threshold value, and the performance level is an assessment level estimated based on the historical data; the first preset threshold value is a self-defined maximum critical value for distinguishing the quality of the sales service;

training a digital twin model by using the sales disadvantage model and the real-time data to obtain the trained digital twin model;

and analyzing the enterprise sales business based on the trained digital twin model and the real-time data.

The invention has the beneficial effects that:

according to the method and the device, data relation mining processing is carried out through correlation analysis according to historical data with poor sales performance, a sales performance model is built according to functional relations between the sales performance and various influence factors, a virtual digital twin model is built based on the sales performance model, interaction and mapping between a physical scene and the virtual digital twin model are built, the digital twin model is trained according to data collected by a sensor in the physical scene in real time, and the trained digital twin model is obtained. Based on the trained digital twin model, real-time data is used as an input value, the sales disadvantage model can be output in real time, the sales service can be analyzed conveniently, and the response rate and accuracy of the sales service analysis management system are improved. On the other hand, the factors in all aspects are comprehensively and professionally considered in the modeling process, time sequence and nonlinear association among the factors is established, and accuracy of association between the factors and sales performance is effectively improved.

Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by the practice of the embodiments of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are required to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained from the drawings without inventive effort.

Fig. 1 is a schematic diagram illustrating the operation steps of an enterprise sales analysis management system based on digital twin technology according to an embodiment of the present invention;

FIG. 2 is a schematic structural diagram of an enterprise sales analysis management system based on digital twin technology according to an embodiment of the present invention;

in the figure, 710, an acquisition module; 720. building a module; 721. a screening unit; 722. a construction unit; 723. a first calculation unit; 724. a second calculation unit; 725. a third calculation unit; 726. a fourth calculation unit; 727. a fifth calculation unit; 7271. a pre-processing unit; 7272. a modeling unit; 730. a training module; 740. an analysis module; 741. a first analysis unit; 742. a fitting unit; 743. a second analysis unit; 744. an intercepting unit; 745. and an output unit.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, 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, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without any inventive step, are within the scope of the present invention.

It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined and explained in subsequent figures. Meanwhile, in the description of the present invention, the terms "first", "second", and the like are used only for distinguishing the description, and are not to be construed as indicating or implying relative importance.

Example 1:

referring to fig. 1, fig. 1 is a schematic diagram illustrating operation steps of an enterprise sales service analysis management system based on a digital twin technology, and the operation method of the enterprise sales service analysis management system includes steps S1, S2, S3, and S4.

And step S1, acquiring historical data and real-time data of sales.

It is understood that, in this step, sales information of the product is collected by various types of sensors, and history data of the sensors and real-time collected data are respectively acquired.

And when the sensors collect data, a large number of parameters irrelevant to sales performance can be collected, so that not only is storage resources wasted, but also the operation process is increased. Therefore, the collected data needs to be preprocessed, and most of the data without actual influence is filtered out, so that the occupation and the operation process of data storage are reduced, and the response rate is improved. Therefore, step S1 further includes step S11, step S12, step S13, step S14, and step S15.

And step S11, analyzing all factors based on the historical data or the real-time data, and establishing a hierarchical structure model.

It can be understood that, in this step, classification analysis is performed on all the factors in the historical data, and the classification analysis is divided into three layers of hierarchical structure models of a target layer, a criterion layer and a scheme layer which are formed from top to bottom. For example, in steel selling, the target layer includes selling grade, total selling amount, etc., the criterion layer includes market, industry, etc., and the scheme layer includes product type, etc.

And S12, comparing the importance of the factors layer by layer based on the hierarchical structure model, and constructing a discrimination matrix.

It can be understood that, in this step, two factors are compared layer by layer based on the hierarchical structure model to obtain a relationship of relative importance, each index is scored by using a 1-9 scale method, and a discrimination matrix is obtained after normalization processing, and the discrimination matrix is as shown in formula (1):

（1）

wherein: a is a discrimination matrix;

the importance ratio scale of the element i and the element j of the current level to the previous level; i and j are different kinds of factors respectively; and n is the dimension of the hierarchical structure model.

And step S13, respectively calculating eigenvectors and maximum eigenvalues based on the discrimination matrix.

It can be understood that, in this step, normalization processing is performed according to column vectors based on the discrimination matrix to obtain a normalized matrix; and then adding the normalized matrixes according to rows to obtain a characteristic vector, and calculating the maximum characteristic value based on the characteristic vector.

And step S14, calculating based on the maximum characteristic value to obtain a consistency index.

In order to ensure the reasonability of the scheme weight, the consistency check is carried out on the discrimination matrix, and an index for measuring the deviation consistency of the discrimination matrix is obtained based on the maximum eigenvalue and a formula (2), wherein the formula (2) is as follows:

（2）

wherein: r is a consistency index;

the maximum eigenvalue of the discrimination matrix; n is the order of the discrimination matrix; e is an average random consistency index;

and step S15, sorting the factors of each level based on the characteristic vectors and the maximum characteristic values under the condition of meeting the consistency index to obtain a factor sorting model.

It can be understood that, in this step, when the consistency index R is less than 0.1, it indicates that the decision matrix satisfies the consistency index, and indicates that the scheme weight is reasonable. If not, redesigning the discrimination matrix. And finally, solving the index weight by adopting a geometric mean method, firstly, after the consistency of the discrimination matrix is checked, solving the eigenvector corresponding to the maximum eigenvalue, normalizing the eigenvector, wherein the corresponding value of each element is the weight of each factor of the layer, sorting the weight values based on the weight values, and modeling based on the steps S11 to S15 to obtain a factor sorting model. And (3) selecting 20 factors with the front weight values (the corresponding number is set according to different application scenes) by using a factor sorting model as the basis of subsequent data analysis and processing, simplifying the complex data processing into the processing of relevance factors, improving the response rate, reducing the occupied space of data storage and the like.

And step S2, mining the data relationship of the historical data based on the correlation analysis, and constructing a sales disadvantage model, wherein the sales disadvantage model is a sales model with a performance level lower than a first preset threshold, and the performance level is an assessment level based on the historical data, and the first preset threshold is a maximum critical value for self-defining the quality of the sales business.

It should be understood that, in this step, the performance level in this embodiment is divided into four levels, i.e., good, medium, and poor, and in other scenarios, a specific numerical range may also be used, and the details are not limited. And if the first preset threshold value is good, calling out historical data with medium and poor performance levels in the historical data, then carrying out data relation mining processing on the historical data based on correlation analysis, finding out a functional relation between the sales performance and the individual factors, and constructing a sales disadvantage model.

In order to further identify the main influence factors of the sales performance, find the degree of association between each factor and the sales performance, extract the main control factors influencing the sales performance, and avoid other factors from influencing the accuracy of the prediction result, step S2 includes step S21, step S22, step S23, step S24, step S25, step S26, and step S27.

And S21, screening out the data to be analyzed which is lower than the first preset threshold value according to the historical data.

Step S22, determining a first time series and a second time series based on the data to be analyzed corresponding to the same performance level, respectively, where the first time series is based on sales performance in the data to be analyzed as a dependent variable; the second time series is a sequence of arguments affecting the first time series.

It is understood that in this step, the processing is performed separately for different performance levels, and based on the data to be analyzed of the same performance level, the steel product sales amount is set as the dependent variable G, and the first time series is determined as

Wherein G is a first time series,

the steel sales amount in a period k, and k is time (year/month, etc. as a measurement unit). Using the influencing factor as an independent variable, and determining the second time series as

Wherein P is a second time series,

is the influence factor under the k period

I is the type of influencing factor and k is time (in years/months, etc. as a unit of measure).

Step S23, calculating the first time series and the second time series based on a mean value conversion method, respectively, to obtain the updated first time series and the updated second time series.

It can be understood that, in this step, to avoid the excessive magnitude difference to cause the excessive subsequent processing error, each sample element is subjected to non-dimensionalization before the data is calculated, in this embodiment, the first time series and the second time series are respectively calculated based on a mean value transformation method to eliminate the dimensional difference between different data, and the mean value transformation method is shown in formula (3):

（3）

wherein:

for the parameters after the non-dimensionalization process,

for a sample in a factor in the first time series or the second time series,

a sample mean value of a factor in the first time series or the second time series;

is the standard deviation of the sample for a factor in the first time series or the second time series.

Step S24, calculating based on the updated first time series and each updated second time series, respectively, to obtain a relationship coefficient.

It is understood that, in this step, the relationship coefficient of each variable factor and dependent variable factor is calculated according to formula (4), formula (4) is as follows;

（4）

wherein:

is a relation coefficient of an independent variable f relative to a dependent variable k; f is an independent variable; k is a dependent variable;

is a first time series;

is a second time series;

for the resolution factor, take 0-1.

And step S25, calculating based on each relation coefficient to obtain the association degree.

It is understood that, in this step, the association degree of the respective variable factors with the dependent variable factors is calculated based on the formula (5), and the formula (5) is as follows;

（5）

wherein:

the correlation degree corresponding to the independent variable t; t is the factor type of the independent variable; h is the factor type of the independent variable; n is the total number of samples of the independent variable factor;

is a relation coefficient of the independent variable f relative to the dependent variable h.

And step S26, determining a correlation index based on all the correlation degrees and a second preset threshold value, wherein the second preset threshold value is a self-defined minimum critical screening value.

It can be understood that, in this step, the second preset threshold is taken as 0.7, and the factors greater than 0.7 are used as the correlation index to filter out the factors with very low correlation degree.

In order to reduce the coupling relation between independent variables, data with high contribution rate is reserved, and dimension reduction processing is carried out on the association index. Step S26 further includes step S261, step S262, step S263, step S264, and step S265.

And S261, establishing a correlation matrix based on all the correlation indexes.

It will be appreciated that in this step, the correlation matrix is constructed according to equation (6), where equation (6) is as follows:

（6）

wherein: r is a correlation matrix;

is a correlation index

And associating

Correlation coefficient between indexes; w is the order of the correlation matrix;

is a correlation index

Average value of (d);

is a correlation index

Average value of (d); n are different data samples.

And step S262, calculating an eigenvalue and an eigenvector corresponding to the eigenvalue based on the correlation matrix.

Step S263, calculating based on the eigenvalue to obtain a first parameter and a second parameter, where the first parameter is an information contribution rate of the principal component, and the second parameter is an accumulated contribution rate of the principal component.

It can be understood that, in this step, the correlation indexes are arranged from large to small according to the feature values, a preset number of correlation indexes are determined to be selected, and the selected correlation indexes are respectively calculated according to the formulas (7) and (8), where the formulas (7) and (8) are as follows:

（7）

（8）

wherein:

the information contribution rate of the jth principal component, i.e. the first parameter;

the characteristic value of the jth principal component;

the characteristic value of the kth principal component; m is the total number of the characteristic values, namely the total number of the main components;

is the cumulative contribution rate, i.e. the second parameter; y is the kind of the main component; p is the total number of principal components.

And step S264, calculating based on the feature vector and the correlation matrix to obtain a third parameter, wherein the third parameter is a principal component matrix.

It is understood that, in this step, the principal component is calculated according to equation (9), and equation (9) is as follows:

（9）

wherein:

is a principal component matrix, i.e. a third parameter;

is a transposed matrix of the eigenvectors; r is a correlation matrix.

Step S265, determining the updated association index based on the analysis of the first parameter, the second parameter, and the third parameter.

It is understood that, in this step, the principal component evaluation score is calculated according to the formula (10), and the formula (10) is as follows:

（10）

wherein: a is the evaluation score of the main component;

an information contribution rate for the s-th principal component;

a principal component matrix; u is the total number of principal components. Rearranging according to the size of the principal component evaluation score and the size of the accumulated contribution rateAnd comprehensively selecting the correlation indexes with the preset number values ahead for subsequent processing, reducing the complexity and redundancy of data, and reserving the correlation indexes with higher values to improve the calculation accuracy.

And step S27, constructing a sales disadvantage model based on the correlation index and the performance level.

It is understood that, in this step, various algorithms such as a neural network model and a regression model are used to simulate a functional relationship between the correlation index and the performance level, thereby constructing a sales disadvantage model.

In this embodiment, based on generating a functional relationship between the associated indicators and the performance level of the confrontation network model simulation, the step S27 includes a step S271 and a step S272, in which the step S271 performs normalization preprocessing based on each of the associated indicators to obtain a preprocessed value.

Step S272, alternately training a generator model and a discriminator model of a countermeasure network model respectively based on the preprocessing numerical value and the performance grade, and taking the preprocessing numerical value as an input value of the generator model to obtain generated data; and then taking the generated data and the performance level as input values of the discriminator model, performing weight updating by using a loss value and a residual value fed back by the discriminator model, and minimizing the residual value and the loss value to obtain the sales disadvantage model, wherein the residual value is a distribution distance between the preprocessed numerical value and the generated data.

It can be understood that, in this step, each parameter of the discriminator model is fixed, the preprocessed values are input into the generation countermeasure network model to train the generator model, and the generated data is obtained after the up-sampling operation. And inputting the generated data and the performance level into a discriminator model for true and false recognition, performing weight updating by using a loss value fed back by the discriminator model and a difference value between the generated data and the performance level, and minimizing a residual value, wherein the weight updating and optimizing algorithms of the generator model and the discriminator model are both RMSProp algorithms.

Due to the unreasonable equivalent optimization distance measurement mode, phenomena such as disappearance of gradients, instability of training models and the like easily occur in the generated confrontation network model, and the Wasserstein distance is adopted as the distance measurement of equivalent optimization in the method for measuring the distance, so that the problems are solved. The method for calculating the difference includes step S2721, step S2722, step S2723, and step S2724.

Step S2721, obtaining a first sample distribution based on the preprocessing numerical value, wherein the first sample distribution is a real distribution probability among the associated indexes.

Step S2722, obtaining a second sample distribution based on the generated data, wherein the second sample distribution is a false distribution probability generated corresponding to the correlation index.

Step S2723, obtaining a third sample distribution based on the first sample distribution and the second sample distribution, where the third sample distribution is a set of all possible joint distributions of the first sample distribution and the second sample distribution.

Step S2724 of calculating Wasserstein distances based on the first sample distribution, the second sample distribution and the third sample distribution to obtain the residual values.

And step S3, training a digital twin model by using the sales disadvantage model and the real-time data to obtain the trained digital twin model.

It can be understood that, in this step, a virtual digital twin model is constructed based on the sales disadvantage model, an interactive mapping relationship between the real physical scene and the virtual digital twin model is established, and then the digital twin model is trained according to real-time data acquired by each sensor in the real physical scene.

The training method of the digital twin model includes step S31, step S32, step S33, and step S34. Wherein step S31, constructing a digital twin model in the virtual space based on the sales disadvantage model.

It can be understood that in the step, a virtual digital twin model is constructed based on the sales disadvantage model, and the interaction and mapping relation between the real physical scene and the virtual digital twin model is established.

And step S32, inputting the real-time data into the sales disadvantage model and the digital twin model respectively for calculation to obtain physical data and simulation data respectively.

It can be understood that, in this step, the real-time data collected by the sensor is respectively input into the sales disadvantage model and the digital twin model for calculation, and physical data and simulation data are respectively obtained.

And step S33, calculating based on the physical data and the simulation data to obtain a difference value.

It is understood that, in this step, the difference value is calculated according to equation (11), equation (11) is as follows:

（11）

wherein:

difference values of the q data;

outputting a fitting value for the kth of the qth data, namely simulation data;

the true value, i.e. the physical data, is output for the kth of the qth data.

And step S34, updating all parameters of the digital twin model based on the difference value to obtain the trained digital twin model.

It can be understood that, in this step, in the training process, the model is fitted according to different models, and the difference value is calculated to form error feedback. Under the action of a plurality of training samples, the model continuously adjusts relevant parameters of the model according to training error feedback and carries out iterative computation until the training is finished, so as to obtain the trained digital twin model.

And step S4, analyzing the enterprise sales service based on the trained digital twin model and the real-time data.

It can be understood that, in this step, the trained digital twin model has learned the relationship between the performance level and the influencing factors based on the inferior marketing service, and real-time data acquired by the sensor in the real physical scene is input into the trained digital twin model, and the inferior marketing model is output to further analyze the enterprise marketing service.

In order to quickly determine the factors affecting the sales service, the above step S4 includes steps S41, S42, S43, S44, and S45.

And step S41, performing stepwise regression analysis based on the output result of the trained digital twin model, and screening out significant influence parameters, wherein the significant influence parameters are factors having significant influence on the performance grade.

It can be understood that, in this step, a stepwise regression analysis method is adopted for modeling based on the output result (inferior sales model) of the digital twin model, all independent variables are rechecked every time one independent variable is added from the first independent variable, and insignificant independent variables are removed to screen out significant influence parameters.

In order to avoid the correlation between the factors having significant influence, step S41 includes steps S411, S412, and S413, in which:

s411, based on the significant influence parameters and the regression model, multiple collinearity tests are carried out, and whether collinearity exists is judged.

And S412, if the significant influence parameters are collinear, analyzing the significant influence parameters by adopting a relevancy sorting method or a principal component analysis method to obtain key characteristic factors, wherein the key characteristic factors are the factors with the largest influence degree in the significant influence parameters.

It can be understood that, in this step, the significant influence parameters with collinearity are analyzed by using a relevance ranking method or a principal component analysis method, and key characteristic factors are selected.

And S413, inputting the key characteristic factors and the non-collinear significant influence parameters into the regression model to perform multiple collinearity detection again and judge whether the parameters are collinear or not until the detection results are not collinear.

It can be understood that, in this step, the remaining significant impact parameters (excluding the significant impact parameters with collinearity) and the key feature factors are input into the regression model each time, multiple collinearity tests are performed and whether collinearity exists is determined, and iteration is performed according to the method of this step until the test results are not collinearity.

And step S42, calculating by utilizing a multivariate linear regression analysis method based on the significant influence parameters to obtain a regression function, wherein the regression function is a function for describing the relationship between the significant influence parameters and the performance grade.

And step S43, performing monotonicity analysis based on the regression function, and determining the optimal significant influence parameter.

It is understood that, in this step, the optimal significant impact parameter is the upper limit value of each factor affecting the sales performance, and lower than the upper limit value indicates poor sales performance.

And step S44, inputting the performance grade corresponding to the real-time data into the regression function for calculation to obtain the simulated significant influence parameter.

And step S45, judging whether the simulated significant influence parameters are in the optimal range of the significant influence parameters, and if so, outputting the significant influence parameters as the disadvantage factors.

It can be understood that, in this step, the one-to-one comparison is performed according to the simulated significant influence parameters and the optimal significant influences, the disadvantage factors are output, and the rate of the factors that clearly influence the sales service is improved.

Example 2:

as shown in fig. 2, the present embodiment provides a structural schematic diagram of an enterprise sales service analysis management system based on a digital twin technology, including an obtaining module 710, a building module 720, a training module 730, and an analyzing module 740, where:

the obtaining module 710: for obtaining historical and real-time data of sales.

The building block 720: the system comprises a correlation analysis module, a sales performance evaluation module and a data relation mining module, wherein the correlation analysis module is used for mining data relation of the historical data based on the correlation analysis and constructing a sales performance model, the sales performance model is a sales model with a performance grade lower than a first preset threshold, and the performance grade is an assessment grade estimated based on the historical data.

Preferably, the building module 720 includes a screening unit 721, a constructing unit 722, a first calculating unit 723, a second calculating unit 724, a third calculating unit 725, a fourth calculating unit 726, and a fifth calculating unit 727, wherein:

the screening unit 721: the data analysis device is used for screening out data to be analyzed, which are lower than the first preset threshold value, according to the historical data;

the configuration unit 722: the system comprises a data analysis module, a data analysis module and a data analysis module, wherein the data analysis module is used for respectively determining a first time sequence and a second time sequence based on the data to be analyzed corresponding to the same performance level, and the first time sequence is based on the sales performance in the data to be analyzed as a dependent variable; the second time series is a sequence of arguments affecting the first time series;

the first calculation unit 723: the time sequence calculation module is used for calculating the first time sequence and the second time sequence respectively based on a mean value conversion method to obtain the updated first time sequence and the updated second time sequence;

the second calculation unit 724: the time sequence calculation module is used for calculating respectively based on the updated first time sequence and each updated second time sequence to obtain a relation coefficient;

the third calculation unit 725: calculating based on each relation coefficient to obtain a correlation degree;

the fourth calculation unit 726: the correlation index is determined based on all the correlation degrees and a second preset threshold value;

fifth calculation unit 727: for constructing a sales disadvantage model based on the correlation indicator and the performance level.

Preferably, the fifth calculation unit 727 includes a preprocessing unit 7271 and a modeling unit 7272, wherein:

the preprocessing unit 7271: the correlation index is used for carrying out normalization preprocessing on the basis of each correlation index to obtain a preprocessing numerical value;

modeling unit 7272: the generator model and the discriminator model are used for alternately training a generator model and a discriminator model of a countermeasure network model respectively based on the preprocessing numerical value and the performance grade, and the preprocessing numerical value is used as an input value of the generator model to obtain generated data; and then taking the generated data and the performance level as input values of the discriminator model, performing weight updating by using a loss value and a residual value fed back by the discriminator model, and minimizing the residual value and the loss value to obtain the sales disadvantage model, wherein the residual value is a distribution distance between the preprocessed numerical value and the generated data.

The training module 730: and the system is used for training the digital twin model by utilizing the sales disadvantage model and the real-time data to obtain the trained digital twin model.

The analysis module 740: and analyzing the enterprise sales business based on the trained digital twin model and the real-time data.

Preferably, the analysis module 740 comprises a first analysis unit 741, a fitting unit 742, a second analysis unit 743, a clipping unit 744 and an output unit 745, wherein:

first analysis unit 741: the system is used for performing stepwise regression analysis based on the output result of the trained digital twin model and screening out significant influence parameters which are factors having significant influence on the performance level;

fitting unit 742: the system comprises a regression function, a performance grade calculation function and a data processing function, wherein the regression function is used for calculating by utilizing a multivariate linear regression analysis method based on the significant influence parameters to obtain a regression function, and the regression function is a function for describing the relationship between the significant influence parameters and the performance grade;

second analysis unit 743: the regression function is used for conducting monotonicity analysis on the basis of the regression function, and the optimal significant influence parameter is determined;

a clipping unit 744: the performance grade corresponding to the real-time data is input into the regression function for calculation, and the simulated significant influence parameters are obtained;

output unit 745: and the system is used for judging whether the simulated significant influence parameters are in the optimal range of the significant influence parameters or not, and if so, outputting the significant influence parameters as the disadvantage factors.

It should be noted that, regarding the digital twin technology-based enterprise sales business analysis management system of the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment related to the operation step, and will not be elaborated herein.

Example 3:

corresponding to the above embodiment of the operation steps, a storage medium is further provided in this embodiment, and a storage medium described below and an enterprise sales service analysis management system based on the digital twin technology described above may be referred to correspondingly.

A storage medium, on which a computer program is stored, wherein the computer program, when executed by a processor, implements the operation steps of the digital twin technology-based enterprise sales service analysis management system of the above method embodiments.

The storage medium may be a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, an optical disk, or other storage media capable of storing program codes.

The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

The above description is only an embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of changes or substitutions within the technical scope of the present invention, and all such changes or substitutions are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. An enterprise sales business analysis management system based on a digital twin technology is characterized by comprising:

acquiring historical data and real-time data of sales;

performing data relation mining processing on the historical data based on correlation analysis to construct a sales disadvantage model, wherein the sales disadvantage model is a sales model with a performance level lower than a first preset threshold value, and the performance level is an assessment level estimated based on the historical data; the first preset threshold is a self-defined maximum critical value for distinguishing the quality of the sales service;

2. The system for analyzing and managing enterprise sales business based on digital twin technology according to claim 1, wherein the data relation mining process is performed on the historical data based on the correlation analysis, and a sales disadvantage model is constructed, and the method comprises the following steps:

screening out data to be analyzed which are lower than the first preset threshold value according to the historical data;

respectively determining a first time sequence and a second time sequence based on the data to be analyzed corresponding to the same performance grade, wherein the first time sequence is based on the sales performance in the data to be analyzed as a dependent variable; the second time series is a sequence of arguments that affect the first time series;

calculating the first time sequence and the second time sequence respectively based on a mean value conversion method to obtain the updated first time sequence and the updated second time sequence;

respectively calculating based on the updated first time series and each updated second time series to obtain a relation coefficient;

calculating based on each relation coefficient to obtain a correlation degree;

determining a correlation index based on all the correlation degrees and a second preset threshold; the second preset threshold value is a self-defined minimum critical screening value;

and constructing a sales disadvantage model based on the correlation indexes and the performance grade.

3. The system for analyzing and managing enterprise sales business based on digital twin technology according to claim 2, wherein the determining of the correlation index based on all the correlation degrees and a second preset threshold further comprises:

establishing a correlation matrix based on all the correlation indexes;

calculating an eigenvalue and an eigenvector corresponding to the eigenvalue based on the correlation matrix;

calculating based on the characteristic value to obtain a first parameter and a second parameter, wherein the first parameter is the information contribution rate of the principal component, and the second parameter is the accumulated contribution rate of the principal component;

calculating based on the feature vector and the correlation matrix to obtain a third parameter, wherein the third parameter is a principal component matrix;

and determining the updated correlation index based on the analysis of the first parameter, the second parameter and the third parameter.

4. The system of claim 2, wherein the construction of the sales disadvantage model based on the correlation index and the performance level comprises:

performing normalization pretreatment based on each correlation index to obtain a pretreatment value;

alternately training a generator model and a discriminator model of a generating countermeasure network model respectively based on the preprocessed numerical values and the performance grade, and taking the preprocessed numerical values as input values of the generator model to obtain generated data; and then taking the generated data and the performance level as input values of the discriminator model, performing weight updating by using a loss value and a residual value fed back by the discriminator model, and minimizing the residual value and the loss value to obtain the sales disadvantage model, wherein the residual value is a distribution distance between the preprocessed numerical value and the generated data.

5. The system for analyzing and managing enterprise sales services based on digital twin technology according to claim 4, wherein the residual value calculating method comprises:

obtaining a first sample distribution based on the preprocessing numerical value, wherein the first sample distribution is a real distribution probability among the associated indexes;

obtaining a second sample distribution based on the generated data, wherein the second sample distribution is a false distribution probability generated correspondingly by the correlation index;

obtaining a third sample distribution based on the first sample distribution and the second sample distribution, the third sample distribution being a set of all possible joint distributions of the first sample distribution and the second sample distribution;

and calculating Wasserstein distances based on the first sample distribution, the second sample distribution and the third sample distribution to obtain the residual values.

6. The system for analyzing and managing enterprise sales business based on digital twin technology according to claim 1, wherein the obtaining historical data and real-time data of sales comprises:

analyzing all factors based on the historical data or the real-time data, and establishing a hierarchical structure model;

based on the hierarchical structure model, comparing the importance of the factors layer by layer to construct a discrimination matrix;

respectively calculating a feature vector and a maximum feature value based on the discrimination matrix;

calculating based on the maximum characteristic value to obtain a consistency index;

and sorting the factors of each level based on the characteristic vectors and the maximum characteristic values under the condition of meeting the consistency index to obtain a factor sorting model.

7. The system of claim 6, wherein the calculating the eigenvector and the maximum eigenvalue based on the discriminant matrix respectively comprises:

based on the discrimination matrix, normalization processing is carried out according to column vectors to obtain a normalization matrix;

adding the normalized matrixes in rows to obtain a characteristic vector;

calculating the maximum eigenvalue based on the eigenvector.

8. The system of claim 2, wherein the training of the digital twin model using the sales disadvantage model and the real-time data comprises:

constructing a digital twin model in a virtual space based on the sales disadvantage model;

inputting the real-time data into the sales disadvantage model and the digital twin model respectively for calculation to obtain physical data and simulation data respectively;

calculating based on the physical data and the simulation data to obtain a difference value;

and updating various parameters of the digital twin model based on the difference value to obtain the trained digital twin model.

9. The system of claim 8, wherein the analysis of the enterprise sales business based on the trained digital twin model and the real-time data comprises:

performing stepwise regression analysis based on the output result of the trained digital twin model, and screening out significant influence parameters which are factors having significant influence on the performance level;

calculating by utilizing a multivariate linear regression analysis method based on the significant influence parameters to obtain a regression function, wherein the regression function is a function describing the relationship between the significant influence parameters and the performance level;

performing monotonicity analysis based on the regression function to determine the optimal significant influence parameter;

inputting the performance grade corresponding to the real-time data into the regression function for calculation to obtain the simulated significant influence parameter;

judging whether the simulated significant influence parameters are in the optimal range of the significant influence parameters, and if so, outputting the significant influence parameters as the disadvantage factors.

10. The system for analyzing and managing enterprise sales services based on digital twin technology according to claim 9, wherein the step-by-step regression analysis based on the output results of the trained digital twin model, and after screening out significant impact parameters, comprises:

performing multiple collinearity tests and judging whether to be collinear based on the significant influence parameters and the regression model;

if the significant influence parameters are collinear, analyzing the significant influence parameters by adopting a relevancy sorting method or a principal component analysis method to obtain key characteristic factors, wherein the key characteristic factors are factors with the largest influence degree in the significant influence parameters;

and inputting the key characteristic factors and the non-collinear significant influence parameters into the regression model to perform multiple collinearity detection again and judge whether the parameters are collinear or not until the detection results are not collinear.