Material supply intelligent settlement management system

Through the intelligent settlement management system for material supply, the CNN-GRU hybrid prediction model is used to predict material storage and demand, which solves the problem of inaccurate prediction of traditional systems, realizes dynamic adjustment strategies, and improves the efficiency and accuracy of inventory management.

CN119067699APending Publication Date: 2024-12-03ANHUI JIYUAN SOFTWARE CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411234765.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-04
Publication Date
2024-12-03

Smart Images

  • Figure CN119067699A_ABST
    Figure CN119067699A_ABST
Patent Text Reader

Abstract

The invention discloses a material supply intelligent settlement management system, and relates to the field of material supply settlement management, and the system comprises a data obtaining module which is used for obtaining source domain data, carrying out the feature screening and standardization of the source domain data, and obtaining a data storage module; the source domain data comprises historical material storage amount data, material demand data, supply chain data, a price fluctuation proportion and a seasonal index; the model prediction module is used for inputting the source domain data into a pre-obtained prediction model to predict the material storage amount and demand amount in the next quarter, and the prediction model is constructed through a CNN-GRU hybrid prediction model; and the intelligent settlement module is used for integrating the predicted values of the storage amount and the demand amount of the materials in the next quarter, and adjusting the existing settlement order according to the ratio of the predicted values of the storage amount and the demand amount.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of material supply settlement management, and more specifically, to an intelligent settlement management system for material supply. Background Art

[0002] In modern industry and manufacturing, intelligent settlement management of material supply chains is crucial. Enterprises must ensure the efficient operation of material supply chains to cope with ever-changing market demands and complex supply networks. Traditional material supply settlement systems are based on historical data and simple statistical models for prediction, which cannot fully consider factors such as market changes, seasonal fluctuations and emergencies, resulting in inaccurate predictions, which in turn affects inventory management and procurement decisions. At the same time, because traditional material supply settlement systems usually rely on manual experience and static inventory management methods, they consume a lot of time and manpower and cannot cope with the modern complex and dynamic supply chain environment. In response to the above problems, the present invention proposes a material supply intelligent settlement management system. Summary of the invention

[0003] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides an intelligent settlement management system for material supply, through which the problems raised in the above-mentioned background technology are solved.

[0004] To achieve the above object, the present invention provides the following technical solutions: A material supply intelligent settlement management system comprises a data acquisition module, a model prediction module and an intelligent settlement module, wherein the modules are connected: the data acquisition module is used to acquire source domain data, and perform feature screening and standardization on the source domain data, wherein the source domain data comprises historical material storage quantity data, material demand data, supply chain data, price fluctuation ratio and seasonal index; the model prediction module is used to input the source domain data into a pre-acquired prediction model to predict the material storage quantity and demand quantity of the next quarter, wherein the prediction model is constructed by a CNN-GRU hybrid prediction model; and the intelligent settlement module is used to integrate the predicted values ​​of the material storage quantity and demand quantity of the next quarter, and calculate the material demand response evaluation index.

[0005] In a preferred embodiment, the feature screening and standardization of the source domain data are specifically as follows: the source domain data is converted into a feature vector; a distance correlation coefficient is calculated based on a preset distance relationship calculation formula in combination with the feature vector; feature screening is performed based on the distance correlation coefficient and preset feature screening rules; and the feature vector after feature screening is subjected to data standardization.

[0006] In a preferred embodiment, the source domain data is input into a pre-acquired prediction model to predict the material storage and demand for the next quarter, specifically: constructing a CNN-GRU hybrid prediction model; dividing the source domain data into a training set and a test set, inputting the training set into the CNN-GRU hybrid prediction model, initializing the training parameters and training the model; using the test set to test and verify the trained model to obtain a trained prediction model; The current source domain data to be predicted is input into the trained prediction model to predict the material storage and demand in the next quarter.

[0007] In a preferred embodiment, the existing settlement order is adjusted according to the ratio of the predicted values ​​of storage volume and demand volume, specifically: when the ratio of the predicted values ​​of storage volume and demand volume is greater than the preset demand warning threshold and less than the preset storage warning threshold, the settlement order is passed and no adjustment is required; when the ratio of the predicted values ​​of storage volume and demand volume is greater than the preset storage warning threshold, the order quantity is reduced; when the ratio of the predicted values ​​of storage volume and demand volume is less than the preset demand warning threshold, the order quantity is increased.

[0008] In a preferred embodiment, the construction of the CNN-GRU hybrid prediction model is specifically as follows: the CNN model includes a first input layer, a convolution layer and a maximum pooling layer; the GRU model includes a second input layer, a hidden layer, a fully connected layer and a regression output layer; the output of the CNN model is used as the input of the GRU model to construct the CNN-GRU hybrid prediction model; the CNN-GRU hybrid prediction model is iteratively optimized according to a preset cross entropy loss function with a regularization term of the L2 norm added.

[0009] In a preferred embodiment, the current source domain data to be predicted is input into the trained prediction model to predict the material storage and demand in the next quarter. Specifically, the CNN model is combined with the convolution layer and the maximum pooling layer to mine the deep information in the source domain data and construct a deep feature vector; the deep feature vector is input into the GRU model for time feature extraction, and the deep feature vector is fitted and predicted; the CNN-GRU hybrid prediction model outputs the predicted value of the material storage and demand in the next quarter through the fully connected layer.

[0010] In a preferred embodiment, the feature vectors after feature screening are standardized, specifically: the mean and standard deviation of all feature vectors are calculated; and each feature vector is standardized based on a standardization formula according to the mean and standard deviation.

[0011] The technical effects and advantages of the intelligent settlement management system for material supply of the present invention are as follows: 1. The present invention obtains source domain data including historical material storage data, material demand data, supply chain data, price fluctuation ratio and seasonal index through a data acquisition module, and performs feature screening and standardization on the source domain data, which can effectively extract a set of representative feature sets from the source domain data, and effectively improve the generalization ability and prediction accuracy of the model; a CNN-GRU hybrid prediction model is constructed through a model prediction module to perform material storage and demand for the next quarter, wherein CNN can not only effectively extract local features of time series data, but also reduce the number of model parameters, reduce computational complexity and training time, on the other hand, GRU is a new type of recurrent neural network. Compared with the LSTM network structure, GRU simplifies the internal unit structure, so that it can better control the update and output of the state vector, thereby extracting important information of the time series data, while filtering out redundant information, and ultimately improving the accuracy and efficiency of the model.

[0012] 2. The present invention can realize a dynamic adjustment strategy by adjusting the existing settlement orders according to the ratio of the predicted values ​​of storage volume and demand volume through the intelligent settlement module. Enterprises can optimize the intelligent settlement management of material supply, ensure the balance between inventory levels and market demand, and improve operational efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 The present invention is a structural schematic diagram of a material supply intelligent settlement management system. DETAILED DESCRIPTION

[0014] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0015] Embodiment 1, Figure 1 The present invention provides a material supply intelligent settlement management system, including a data acquisition module, a model prediction module and an intelligent settlement module, and there are connections between the modules: A data acquisition module is used to acquire source domain data and perform feature screening and standardization on the source domain data, wherein the source domain data includes historical material storage data, material demand data, supply chain data, price fluctuation ratio, and seasonal index; A model prediction module is used to input the source domain data into a pre-acquired prediction model to predict the material storage and demand in the next quarter, wherein the prediction model is constructed by a CNN-GRU hybrid prediction model; The intelligent settlement module is used to obtain the forecast values ​​of material storage and demand for the next quarter, and adjust the existing settlement orders according to the ratio of the forecast values ​​of storage and demand.

[0016] In this example, the source domain data is feature screened and standardized, specifically: Convert source domain data into feature vectors; Calculate the distance correlation coefficient based on the preset distance relationship calculation formula in combination with the feature vector; Perform feature screening based on distance correlation coefficient and preset feature screening rules; The feature vector after feature screening is normalized.

[0017] In this example, the specific expression for feature screening based on the distance correlation coefficient and the preset feature screening rule is:

[0018] in, is the screening discriminant value of the k-th dimension feature vector, which is a Boolean value, and w is the screening threshold. is the distance correlation coefficient between the k-th dimension feature vector and the output label Y.

[0019] It should be noted that when the absolute value of the distance correlation coefficient is greater than the threshold w, The value of is 1, then the k-th dimension feature is retained; when the absolute value of the distance correlation coefficient is less than the threshold w, If the value of is 0, the k-th dimension feature vector is deleted. After the new feature vector is screened, it is uniformly normalized.

[0020] In this example, the preset distance relationship calculation formula is specifically:

[0021] in, is the k-th dimension feature vector, Y is the output label, and They are The distance variance from Y, is the distance covariance between the k-th dimension feature vector and the output label Y, is the distance correlation coefficient between the k-th dimension feature vector and the output label Y.

[0022] It should be noted that when =0, there is no correlation between the two. =1, the two are strongly correlated.

[0023] In this example, the feature vector after feature screening is subjected to data standardization, specifically: Calculate the mean and standard deviation of all eigenvectors; Standardize each feature vector based on the standardization formula according to the mean and standard deviation; The calculation formulas for the mean and standard deviation are:

[0024] in, is the mean, N is the row of the feature matrix, is the i-th eigenvalue of the j-th row, is the standard deviation; The standardized formula is as follows:

[0025] in, is the mean, is the i-th eigenvalue of the j-th row, is the standard deviation, is the standardized eigenvalue.

[0026] In this example, the source domain data is input into the pre-acquired forecasting model to forecast the material storage and demand for the next quarter, specifically: Construct a CNN-GRU hybrid prediction model; Dividing the source domain data into a training set and a test set, inputting the training set into the CNN-GRU hybrid prediction model, initializing training parameters and training the model; Use the test set to test and verify the trained model to obtain the trained prediction model; The current source domain data to be predicted is input into the trained prediction model to predict the material storage and demand in the next quarter.

[0027] In this example, a CNN-GRU hybrid prediction model is constructed, specifically: The CNN model includes the first input layer, convolutional layer, and maximum pooling layer; The GRU model includes a second input layer, a hidden layer, a fully connected layer, and a regression output layer; The output of the CNN model is used as the input of the GRU model to build a CNN-GRU hybrid prediction model; The CNN-GRU hybrid prediction model is iteratively optimized according to the preset cross entropy loss function with the L2 norm regularization term.

[0028] In this example, the regularization term of the L2 norm is:

[0029] in, is the cross entropy loss of the CNN-GRU hybrid prediction model, ‖𝑤‖ is the L2 norm of the parameter vector w, is the regularization coefficient and L is the regularization term.

[0030] It should be noted that the three convolutional layers in the CNN model are used to extract deeper features, and the role of the maximum pooling layer is to reduce the size of the model, increase the calculation speed of the model, and improve the robustness of the extracted features.

[0031] In this example, the current source domain data to be predicted is input into the trained prediction model to predict the material storage and demand in the next quarter, specifically: The CNN model combines convolutional layers and maximum pooling layers to mine deep information in the source domain data and construct deep feature vectors. Inputting the deep feature vector into the GRU model to extract temporal features, and performing fitting prediction on the deep feature vector; The CNN-GRU hybrid forecasting model outputs the forecast values ​​of material storage and demand for the next quarter through the fully connected layer.

[0032] The source domain data is divided into a training set and a test set, the training set is input into the CNN-GRU hybrid prediction model, the training parameters are initialized and the model is trained, specifically: The CNN-GRU hybrid prediction model first performs feature screening and normalization on the source domain data, and then inputs it into the input layer; uses the convolution layer and the maximum pooling layer to extract and construct features of the time series data to obtain the feature vector of the time series data; and uses the expanded time series feature vector as the input of the GRU to complete the modeling and processing of the sequence data.

[0033] It should be noted that with the rise of deep learning and the improvement of computing power, CNN has been deeply studied and widely used, and its model is also constantly evolving and developing. CNN can not only effectively extract local features of time series data, but also reduce the amount of model parameters, reduce computational complexity and training time. On the other hand, GRU is a new type of recurrent neural network. Compared with the LSTM network structure, GRU simplifies the internal unit structure, so that it can better control the update and output of the state vector, thereby extracting important information of time series data, while filtering out redundant information, and ultimately improving the accuracy and efficiency of the model. The CNN-GRU hybrid prediction model combines the local feature extraction capability of CNN and the time series information processing capability of GRU to achieve the advantages of reduced parameters, improved computational efficiency, accurate information extraction, redundant information filtering and improved model accuracy.

[0034] In this example, the existing settlement order is adjusted according to the ratio of the predicted value of storage volume to demand volume, specifically: When the ratio of the predicted storage volume to the demand volume is greater than the preset storage warning threshold, it means that the predicted storage volume matches the expected demand volume well, indicating that the inventory level is moderate and can meet market demand, and the settlement order is passed; When the ratio of the predicted storage volume to the demand volume is greater than the preset storage warning threshold, it indicates that there may be over-storage and it is necessary to consider reducing inventory or optimizing procurement plans to reduce order quantities. When the ratio of the predicted value of storage volume to demand volume is greater than the preset storage warning threshold, it indicates that there may be insufficient inventory and it is necessary to increase inventory to avoid the risk of out-of-stock; increase the order quantity; Dynamic adjustment strategies can be implemented through the ratio of predicted storage volume to demand volume. Enterprises can optimize intelligent settlement management of material supply, ensure the balance between inventory levels and market demand, and improve operational efficiency.

[0035] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0036] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0037] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0038] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0039] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0040] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A material supply intelligent settlement management system, characterized in that: It includes data acquisition module, model prediction module and intelligent settlement module, and there are connections between the modules: A data acquisition module is used to acquire source domain data and perform feature screening and standardization on the source domain data. The source domain data includes historical material storage data, material demand data, supply chain data, price fluctuation ratio, and seasonal index; A model prediction module is used to input the source domain data into a pre-acquired prediction model to predict the material storage and demand in the next quarter, wherein the prediction model is constructed by a CNN-GRU hybrid prediction model; The intelligent settlement module is used to obtain the forecast values ​​of material storage and demand for the next quarter, and adjust the existing settlement orders according to the ratio of the forecast values ​​of storage and demand.

2. The intelligent settlement management system for material supply according to claim 1 is characterized in that: The source domain data is subjected to feature screening and standardization, specifically: Convert source domain data into feature vectors; Calculate the distance correlation coefficient based on the preset distance relationship calculation formula in combination with the feature vector; Perform feature screening based on distance correlation coefficient and preset feature screening rules; The feature vector after feature screening is normalized.

3. The intelligent settlement management system for material supply according to claim 1 is characterized in that: The source domain data is input into the pre-acquired forecasting model to forecast the material storage and demand for the next quarter, specifically: Construct a CNN-GRU hybrid prediction model; Dividing the source domain data into a training set and a test set, inputting the training set into the CNN-GRU hybrid prediction model, initializing training parameters and training the model; Use the test set to test and verify the trained model to obtain the trained prediction model; The current source domain data to be predicted is input into the trained prediction model to predict the material storage and demand in the next quarter.

4. The intelligent settlement management system for material supply according to claim 1 is characterized in that: As described above, the existing settlement order is adjusted according to the ratio of the predicted value of storage volume to demand volume, specifically: When the ratio of the predicted value of storage volume to demand volume is greater than the preset demand warning threshold and less than the preset storage warning threshold, the settlement order is passed without adjustment; When the ratio of the predicted value of storage quantity to demand quantity is greater than the preset storage warning threshold, the order quantity is reduced; When the ratio of the predicted value of storage volume to demand volume is less than the preset demand warning threshold, the order volume is increased.

5. The intelligent settlement management system for material supply according to claim 3 is characterized in that: The construction of the CNN-GRU hybrid prediction model is specifically as follows: The CNN model includes the first input layer, convolutional layer, and maximum pooling layer; The GRU model includes a second input layer, a hidden layer, a fully connected layer, and a regression output layer; The output of the CNN model is used as the input of the GRU model to build a CNN-GRU hybrid prediction model; The CNN-GRU hybrid prediction model is iteratively optimized according to the preset cross entropy loss function with the L2 norm regularization term.

6. The intelligent settlement management system for material supply according to claim 5 is characterized in that: The regularization term of the L2 norm is specifically: in, is the cross entropy loss of the CNN-GRU hybrid prediction model, ‖𝑤‖ is the L2 norm of the parameter vector w, is the regularization coefficient and L is the regularization term.

7. The intelligent settlement management system for material supply according to claim 5 is characterized in that: The current source domain data to be predicted is input into the trained prediction model to predict the material storage and demand in the next quarter, specifically: The CNN model combines convolutional layers and maximum pooling layers to mine deep information in the source domain data and construct deep feature vectors. Inputting the deep feature vector into the GRU model to extract temporal features, and performing fitting prediction on the deep feature vector; The CNN-GRU hybrid forecasting model outputs the forecast values ​​of material storage and demand for the next quarter through the fully connected layer.

8. The intelligent settlement management system for material supply according to claim 2 is characterized in that: The specific expression for feature screening based on the distance correlation coefficient and the preset feature screening rule is: in, is the screening discriminant value of the k-th dimension feature vector, which is a Boolean value, and w is the screening threshold. is the distance correlation coefficient between the k-th dimension feature vector and the output label Y.

9. The intelligent settlement management system for material supply according to claim 2 is characterized in that: The preset distance relationship calculation formula is specifically: in, is the k-th dimension feature vector, Y is the output label, and They are The distance variance from Y, is the distance covariance between the k-th dimension feature vector and the output label Y, is the distance correlation coefficient between the k-th dimension feature vector and the output label Y.

10. The intelligent settlement management system for material supply according to claim 2 is characterized in that: The feature vector after feature screening is standardized, specifically: Calculate the mean and standard deviation of all eigenvectors; Standardize each feature vector based on the standardization formula according to the mean and standard deviation; The calculation formulas for the mean and standard deviation are: in, is the mean, N is the row of the feature matrix, is the i-th eigenvalue of the j-th row, is the standard deviation; The standardized formula is as follows: in, is the mean, is the i-th eigenvalue of the j-th row, is the standard deviation, is the standardized eigenvalue.