Incoming invoice roll-out method, device and equipment and medium
By combining multiple binary classifiers and dimensionality reduction processing in high-dimensional feature space, the problem of input invoice classification is solved, efficient and accurate invoice transfer is achieved, and the efficiency and accuracy of financial processing is improved.
Patent Information
- Application Number
- CN202510057397.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-14
- Publication Date
- 2025-05-09
AI Technical Summary
The prior art is difficult to efficiently process and classify complex input invoices, especially when facing linear and inseparable data, and traditional support vector machine methods are difficult to achieve accurate classification.
Multi-classifiers are built by combining multiple preset binary classifiers and dimensionality reduction is performed in high-dimensional feature space using nonlinear mapping and principal component analysis. For nonlinear inseparable data, use a multi-classifier for classification; for linear inseparable data, add a penalty term to modify the multi-classifier, thereby achieving accurate data classification and transfer.
It improves the classification accuracy and accuracy of input invoices, reduces the time and cost of manual operations, reduces the error rate, and improves the accuracy and efficiency of financial processing.
Smart Images
Figure CN119961808A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of data processing technology, and in particular to a method, device, equipment and medium for transferring input invoices. Background Art
[0002] In the financial management of enterprises, the management of input invoices is an important link. With the expansion of enterprise scale and the increase of business volume, the number of input invoices has also increased sharply. How to efficiently handle these invoices, especially the accurate classification and rapid processing of input invoices that need to be transferred out, has become a major challenge for enterprises.
[0003] Traditionally, the manual processing of invoices is not only inefficient, but also easily interfered by human factors, resulting in frequent errors. It is obviously unable to meet the high standards of modern enterprises for refined management. Therefore, methods based on machine learning and data mining have attracted much attention due to their efficiency and accuracy. As a method in the field of machine learning, the core idea of linear support vector machine is to find an optimal hyperplane as a decision boundary to clearly divide data points of different categories, while maximizing the minimum interval between each type of data point and the hyperplane. For linear input invoice data sets and binary classification, SVM can be used directly for classification, and a hyperplane can be obtained to separate the two samples. However, there are many types of input invoices, and the data features are complex and diverse. Sometimes even a seemingly linear data set may show linear indivisibility due to the inherent correlation and noise interference of the data. If the data set encountered is linear and indivisible, it is difficult to classify various types of input invoices in this way. Summary of the invention
[0004] In order to solve the above technical problems, one or more embodiments of this specification provide a method, device, equipment and medium for transferring input invoices.
[0005] One or more embodiments of this specification adopt the following technical solutions:
[0006] One or more embodiments of this specification provide a method for transferring out an input invoice, the method comprising:
[0007] Combining multiple preset binary classifiers based on a preset method to construct a multi-classifier;
[0008] Mapping the input invoice to be transferred out to a high-dimensional feature space according to a nonlinear mapping, and performing dimensionality reduction processing on the input invoice to be transferred out based on a principal component analysis strategy in the high-dimensional feature space to obtain a data representation after dimensionality reduction;
[0009] Determining the data type of the input invoice to be transferred out based on the data representation;
[0010] If the data type is non-linear indivisible data, classifying the data representation based on the multiple classifiers, and transferring the input invoice to be transferred out based on the classification result;
[0011] If the data type is linearly inseparable data, a penalty term is added to the objective function of the multi-class classification to obtain a modified multi-classifier, and the data representation is classified based on the modified multi-classifier, and the input invoice to be transferred out is transferred out based on the classification result.
[0012] Optionally, in one or more embodiments of the present specification, before combining multiple preset binary classifiers based on a preset method to construct a multi-classifier, the method further includes:
[0013] The historical input invoices are used as sample points, and the optimization problem corresponding to the initial binary classifier is determined; wherein the optimization problem is to maximize the geometric interval and satisfy the function interval constraint of all sample points while minimizing the function interval constraint;
[0014] Converting the optimization problem into a convex quadratic programming problem to determine a definition expression corresponding to the convex quadratic programming problem;
[0015] Optimizing the variables in the definition expression based on a sequence minimum optimization algorithm to obtain an optimal definition expression;
[0016] According to the optimal definition expression, an optimal separation hyperplane corresponding to the preset binary classifier is determined.
[0017] Optionally, in one or more embodiments of the present specification, a plurality of preset binary classifiers are combined based on a preset method to construct a multi-classifier, specifically including:
[0018] Determine the classification category of the multi-classifier to be constructed based on the transferable interface corresponding to the input invoice to be transferred out;
[0019] Determine the preset method corresponding to constructing the multi-classifier according to the current hardware resources corresponding to the input invoice to be transferred out and the data volume of the input invoice to be transferred out;
[0020] If it is determined that the preset method is a one-to-one method, training a preset binary classifier corresponding to any two classification categories within the classification category, so as to combine the preset binary classifiers to obtain a multi-classifier;
[0021] If it is determined that the preset method is a one-to-many method, a preset binary classification type is trained corresponding to each classification type to combine the preset classifiers to obtain a multi-classifier.
[0022] Optionally, in one or more embodiments of the present specification, the input invoice to be transferred out is mapped to a high-dimensional feature space according to a nonlinear mapping, and the input invoice to be transferred out is subjected to dimensionality reduction processing based on a principal component analysis strategy in the high-dimensional feature space to obtain a data representation after dimensionality reduction, specifically including:
[0023] Mapping the input invoices to be transferred out to a high-dimensional feature space based on a preset kernel function, so as to calculate the covariance matrix of each input invoice to be transferred out in the high-dimensional feature space;
[0024] Performing eigendecomposition on the covariance matrix to obtain eigenvalues and eigenvectors corresponding to the eigenvalues;
[0025] Sorting the eigenvalues according to their sizes, screening the eigenvalues based on the ratio of the sum of the eigenvalues to the sum of the total eigenvalues, and obtaining the principal component;
[0026] Project each of the input invoices to be transferred out in the high-dimensional space onto the principal component to obtain a data representation after dimensionality reduction.
[0027] Optionally, in one or more embodiments of the present specification, determining the data type of the input invoice to be transferred out based on the data representation specifically includes:
[0028] Inputting the data representation into the multi-classifier to obtain the decision boundary of the multi-classifier;
[0029] If it is determined that the decision boundary has a nonlinear shape, then the data type of the input invoice to be transferred out is determined to be linearly inseparable data;
[0030] If it is determined that the decision boundary appears in a linear shape, then the data type of the input invoice to be transferred out is determined to be non-linear indivisible data.
[0031] Optionally, in one or more embodiments of the present specification, if the data type is non-linear indivisible data, classifying the data representation based on the multiple classifiers, and transferring the input invoice to be transferred out based on the classification result, specifically includes:
[0032] Inputting the data representation into the multi-classifier, predicting the data representation based on a plurality of preset binary classifiers in the multi-classifier, and obtaining a predicted category;
[0033] Voting is performed according to the predicted categories of the preset binary classifier to determine the classification result represented by the data;
[0034] Determine the transfer interface corresponding to the classification result, so as to transfer the input invoice to be transferred out to the corresponding transfer interface.
[0035] Optionally, in one or more embodiments of the present specification, classifying the data representation based on the modified multi-classifier, and transferring the input invoice to be transferred out based on the classification result, specifically includes:
[0036] inputting the data representation into the modified multi-classifier to construct an optimal separation hyperplane in a high-dimensional feature space based on the multi-classifier;
[0037] Based on the classification result corresponding to the optimal separating hyperplane, the transfer-out interface corresponding to the classification result is determined, so as to transfer the input invoice to be transferred out to the corresponding transfer-out interface.
[0038] One or more embodiments of this specification provide a device for transferring out incoming invoices, the device comprising:
[0039] A construction unit, used for combining a plurality of preset binary classifiers based on a preset method to construct a multi-classifier;
[0040] A dimensionality reduction unit, used for mapping the input invoice to be transferred out to a high-dimensional feature space according to a nonlinear mapping, and performing dimensionality reduction processing on the input invoice to be transferred out based on a principal component analysis strategy in the high-dimensional feature space to obtain a data representation after dimensionality reduction;
[0041] A determination unit, configured to determine the data type of the input invoice to be transferred out based on the data representation;
[0042] A first transfer unit, configured to classify the data representation based on the multiple classifiers if the data type is non-linear data, and transfer the input invoice to be transferred out based on the classification result;
[0043] The second transfer unit is used to add a penalty term to the objective function of the multi-class classification if the data type is linear data to obtain a modified multi-classifier, classify the data representation based on the modified multi-classifier, and transfer the input invoice to be transferred out based on the classification result.
[0044] One or more embodiments of this specification provide a device for transferring out incoming invoices, the device comprising:
[0045] at least one processor; and,
[0046] a memory communicatively connected to the at least one processor; wherein,
[0047] The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:
[0048] Combining multiple preset binary classifiers based on a preset method to construct a multi-classifier;
[0049] Mapping the input invoice to be transferred out to a high-dimensional feature space according to a nonlinear mapping, and performing dimensionality reduction processing on the input invoice to be transferred out based on a principal component analysis strategy in the high-dimensional feature space to obtain a data representation after dimensionality reduction;
[0050] Determining the data type of the input invoice to be transferred out based on the data representation;
[0051] If the data type is non-linear data, classifying the data representation based on the multiple classifiers, and transferring the input invoice to be transferred out based on the classification result;
[0052] If the data type is linear data, a penalty term is added to the objective function of the multi-class classification to obtain a modified multi-classifier, and the data representation is classified based on the modified multi-classifier, and the input invoice to be transferred out is transferred out based on the classification result.
[0053] One or more embodiments of this specification provide a non-volatile computer storage medium storing computer executable instructions, wherein the computer executable instructions are configured to:
[0054] Combining multiple preset binary classifiers based on a preset method to construct a multi-classifier;
[0055] Mapping the input invoice to be transferred out to a high-dimensional feature space according to a nonlinear mapping, and performing dimensionality reduction processing on the input invoice to be transferred out based on a principal component analysis strategy in the high-dimensional feature space to obtain a data representation after dimensionality reduction;
[0056] Determining the data type of the input invoice to be transferred out based on the data representation;
[0057] If the data type is non-linear data, classifying the data representation based on the multiple classifiers, and transferring the input invoice to be transferred out based on the classification result;
[0058] If the data type is linear data, a penalty term is added to the objective function of the multi-class classification to obtain a modified multi-classifier, and the data representation is classified based on the modified multi-classifier, and the input invoice to be transferred out is transferred out based on the classification result.
[0059] At least one of the above technical solutions adopted in the embodiments of this specification can achieve the following beneficial effects:
[0060] By combining multiple binary classifiers, we can make full use of the advantages of different classifiers in different categories or features, so as to more accurately capture the inherent laws and differences of the data and improve the precision and accuracy of the overall classification. Then, the original input invoice to be transferred out is mapped to a higher-dimensional space through the preset kernel function. This mapping can capture the complex nonlinear relationship in the input invoice to be transferred out, thereby extracting more representative features. When there is linearly inseparable data, by adding a penalty term to the original objective function, the optimal separating hyperplane is made to have a large interval while the number of misclassified samples is as small as possible, thereby improving the accuracy of classification. Based on the classification results corresponding to the optimal separating hyperplane, the input invoice to be transferred out can be transferred out accurately, and the automatic transfer of the input invoice can be realized. The automated classification and transfer process reduces the time and cost of manual operation, while reducing the error rate and improving the accuracy and efficiency of financial processing. BRIEF DESCRIPTION OF THE DRAWINGS
[0061] In order to more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art description. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. In the drawings:
[0062] Figure 1 A schematic diagram of a method flow of a method for transferring an input invoice provided in an embodiment of this specification;
[0063] Figure 2 A schematic diagram of the classification of a two-classification input invoice provided in the embodiment of this specification;
[0064] Figure 3 A schematic diagram of the classification of linearly inseparable input invoices provided in the embodiments of this specification;
[0065] Figure 4 A schematic diagram of an optimal separation hyperplane corresponding to a preset binary classifier provided in an embodiment of this story;
[0066] Figure 5 A schematic diagram of an optimal separation hyperplane in a high-dimensional feature space provided in an embodiment of this specification;
[0067] Figure 6 A schematic diagram of the structure of a transfer device for an input invoice provided in an embodiment of this specification;
[0068] Figure 7 A schematic diagram of the structure of a transfer device for an input invoice provided in an embodiment of this specification;
[0069] Figure 8 A schematic diagram of the structure of a non-volatile storage medium provided in an embodiment of this specification. DETAILED DESCRIPTION
[0070] The embodiments of this specification provide a method, device, equipment and medium for transferring input invoices.
[0071] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.
[0072] like Figure 1 As shown, in the embodiment of this specification, a flow chart of a method for transferring out an input invoice is provided. Figure 1 It can be seen that in one or more embodiments of this specification, a method for transferring an input invoice includes the following process:
[0073] S101: Combining multiple preset binary classifiers based on a preset method to construct a multi-classifier.
[0074] In order to adapt to the problem of various types of input invoices and complex and diverse data features, and to solve the problem that the binary classification is difficult to adapt to the complex scenario of the current input invoice, in the embodiments of this specification, multiple preset binary classifiers are combined according to the preset method to construct a multi-classifier. Further, in one or more embodiments of this specification, before combining multiple preset binary classifiers based on the preset method to construct a multi-classifier, the method also includes:
[0075] Take the historical input invoices as sample points and determine the optimization problem corresponding to the initial binary classifier; it can be understood that the optimization problem is to maximize the geometric interval and satisfy the function interval constraints of all sample points to minimize the function interval constraints, where the geometric interval can be analogous to the distance from the midpoint to the straight line in plane geometry, or the distance from the midpoint to the plane in solid geometry. Then convert the optimization problem into a convex quadratic programming problem to determine the definition expression corresponding to the convex quadratic programming problem. Optimize the variables in the definition expression based on the sequence minimum optimization algorithm to obtain the optimal definition expression. Then, based on the optimal definition expression, the optimal separation hyperplane corresponding to the binary classifier can be preset. For example Figure 4 Shown is the optimal separation hyperplane corresponding to the preset binary classifier.
[0076] In this process, by defining the optimization problem to maximize the geometric interval and satisfy the function interval constraints of all sample points, the classifier can achieve a high accuracy on the training set. Maximizing the geometric interval helps the classifier maintain good generalization ability when facing new data and reduce the risk of overfitting. Converting the optimization problem into a convex quadratic programming problem and solving it using a sequential minimum optimization algorithm can ensure that the global optimal solution is found rather than the local optimal solution. This enhances the robustness of the classifier, allowing it to maintain stable performance even when facing complex or noisy data.
[0077] For a linear invoice data set with two categories, SVM can be used directly for classification, and a hyperplane can be obtained. Figure 2 As shown, the two samples can be separated. However, there are many types of input invoices, and the data features are complex and diverse. Sometimes even a seemingly linear data set may show linear inseparable characteristics due to the inherent correlation and noise interference of the data. If the data set encountered is linear and inseparable, the classification result when classifying based on the preset binary classifier will be as follows Figure 3 It is difficult to determine as shown, so specifically, in one or more embodiments of this specification, multiple preset binary classifiers are combined based on a preset method to construct a multi-classifier, specifically including:
[0078] First, according to the transferable interface corresponding to the input invoice to be transferred out, the classification category of the multi-classifier to be constructed is determined. Then, according to the current hardware resources corresponding to the input invoice to be transferred out and the data volume of the input invoice to be transferred out, the preset method corresponding to the construction of the multi-classifier is determined. If the hardware resources and the amount of data are moderate, or the difference between the categories is large, then the preset method is determined to be a one-to-one method, and a preset binary classifier corresponding to any two classification categories within the classification category is trained to combine the preset binary classifiers to obtain a multi-classifier. If the hardware resources are limited or the amount of data is very large, then the preset method is determined to be a one-to-many method, and a preset binary type is trained corresponding to each classification type to combine the preset classifiers to obtain a multi-classifier.
[0079] By combining multiple binary classifiers in this process, we can make full use of the advantages of different classifiers in different categories or features, so as to more accurately capture the inherent laws and differences of the data and improve the precision and accuracy of the overall classification. This process can flexibly select preset methods according to different hardware resources and data volumes, so that the system can be optimized according to actual conditions, which not only ensures the classification effect but also avoids unnecessary resource consumption. For data sets such as invoices, which are of various types and have complex and diverse data features, a single classifier is often unable to cope with it. Multi-classifiers can help process linearly inseparable or nonlinearly separable data sets by combining multiple binary classifiers, and improve the robustness and stability of classification.
[0080] S102: Mapping the input invoice to be transferred out to a high-dimensional feature space according to nonlinear mapping, and performing dimensionality reduction processing on the input invoice to be transferred out in the high-dimensional feature space based on a principal component analysis strategy to obtain a data representation after dimensionality reduction.
[0081] In order to improve the feature expression capability of the input invoices to be transferred out, and to achieve the effect of retaining important features while reducing redundant information and thus improving classification accuracy, in the embodiment of this specification, the input invoices to be transferred out are mapped to a high-dimensional feature space according to a nonlinear mapping, and the input invoices to be transferred out are subjected to dimensionality reduction processing in the high-dimensional feature space based on the principal component analysis strategy to obtain the data representation after dimensionality reduction.
[0082] Specifically, in one or more embodiments of the present specification, the input invoice to be transferred out is mapped to a high-dimensional feature space according to a nonlinear mapping, and the input invoice to be transferred out is subjected to dimensionality reduction processing based on a principal component analysis strategy in the high-dimensional feature space to obtain a data representation after dimensionality reduction, which specifically includes:
[0083] First, the input invoices to be transferred out are mapped to a high-dimensional feature space according to a preset kernel function, so as to calculate the covariance matrix of each input invoice to be transferred out in the high-dimensional feature space. Then, the covariance matrix is eigen-decomposed to obtain the eigenvalues and the eigenvectors corresponding to the eigenvalues. The eigenvalues are sorted according to the size, and the eigenvalues are screened based on the preset ratio of the sum of the eigenvalues to the sum of the total eigenvalues to obtain the principal component. Then, each of the input invoices to be transferred out in the high-dimensional space is projected onto the principal component to obtain the data representation after dimensionality reduction. Through the preset kernel function such as RBF, polynomial kernel, etc., the original input invoice to be transferred out can be mapped to a higher-dimensional space. This mapping can capture the complex nonlinear relationship in the input invoice to be transferred out, thereby extracting more representative features. In addition, by selecting the principal component whose sum of eigenvalues accounts for a certain proportion of the sum of the total eigenvalues, it can be ensured that the data after dimensionality reduction still retains the key information in the original data, thereby ensuring the accuracy of subsequent analysis.
[0084] S103: Determine the data type of the input invoice to be transferred out based on the data representation.
[0085] Based on the data identifier determined in the above step S102, the data type of the input invoice to be transferred out is determined to facilitate better classification of the input invoice to be transferred out. Specifically, in one or more embodiments of the present specification, the data type of the input invoice to be transferred out is determined based on the data representation, which specifically includes the following process: First, the data representation is input into a multi-classifier to obtain the decision boundary of the multi-classifier. If it is determined that the decision boundary appears to have a nonlinear shape, then the data type of the input invoice to be transferred out is determined to be linearly inseparable data. If it is determined that the decision boundary appears to have a linear shape, then the data type of the input invoice to be transferred out is determined to be nonlinearly inseparable data.
[0086] S104: If the data type is non-linear inseparable data, classify the data representation based on the multiple classifiers, and transfer the input invoice to be transferred out based on the classification result.
[0087] If the data type is non-linear indivisible data, the data representation is classified based on the multiple classifiers, and the invoice to be transferred out is transferred out based on the classification result. Specifically, in one or more embodiments of the present specification, if the data type is non-linear indivisible data, the data representation is classified based on the multiple classifiers, and the input invoice to be transferred out is transferred out based on the classification result, specifically including:
[0088] The data representation is input into a multi-classifier, and the data representation is predicted according to multiple preset binary classifiers in the multi-classifier to obtain the predicted category. Then, voting is performed according to the predicted category of the preset binary classifier to determine the classification result of the data representation. Then, the transfer interface corresponding to the classification result is determined to transfer the input invoice to be transferred to the corresponding transfer interface. By using multiple classifiers, especially in combination with multiple preset binary classifiers, this method can handle complex, nonlinearly separable data sets more flexibly. Each binary classifier focuses on different aspects or features of the data. By integrating the prediction results of multiple classifiers, the accuracy of classification can be significantly improved.
[0089] S105: If the data type is linearly inseparable data, a penalty term is added to the objective function of the multi-class classification to obtain a modified multi-classifier, and the data representation is classified based on the modified multi-classifier, and the input invoice to be transferred out is transferred out based on the classification result.
[0090] If the data type is determined to be linearly inseparable data, then it is necessary to add a penalty term to the objective function of the multi-class classification to obtain a modified multi-classifier, and classify the data representation based on the modified multi-classifier, and transfer out the incoming invoices to be transferred out based on the classification results. Because when the data type is linearly inseparable data, there are samples that do not meet the constraints, which will result in no feasible solution to the optimization problem. Therefore, in order to make SVM suitable for linearly inseparable data sets, it is necessary to modify the objective function and constraints. By adding a penalty term to the original objective function, the more misclassified samples, the greater the penalty, which makes the optimal separation hyperplane as large as possible while minimizing the number of misclassified samples.
[0091] Specifically, in one or more embodiments of the present specification, classifying the data representation based on the modified multi-classifier, and transferring out the input invoice to be transferred out based on the classification result, specifically includes:
[0092] The data representation is input into the modified multi-classifier, so as to construct the optimal separation hyperplane in the high-dimensional feature space according to the multi-classifier. Then, based on the classification result corresponding to the optimal separation hyperplane, the transfer interface corresponding to the classification result is determined, and the input invoice to be transferred out is transferred to the corresponding transfer interface. The modified multi-classifier can construct the optimal separation hyperplane in the high-dimensional feature space, which means that it can more accurately capture the complex patterns and boundaries in the data, thereby achieving effective classification of nonlinear inseparable data. This ability is particularly important for processing complex and changeable financial data. By looking for such in the high-dimensional space Figure 5 The optimal separation hyperplane shown in the figure shows that multiple classifiers can more accurately classify data and reduce the possibility of misclassification. This is crucial for financial operations that require high accuracy, such as invoice transfers, to ensure that invoices are correctly and timely transferred to the corresponding interface. The automated classification and transfer process reduces the time and cost of manual operations, while reducing the error rate and improving the accuracy and efficiency of financial processing.
[0093] like Figure 6 As shown, the embodiment of this specification provides a schematic diagram of the internal structure of the transfer device of the input invoice, which is composed of Figure 6 It can be seen that in one or more embodiments of this specification, a device for transferring invoices includes:
[0094] A construction unit 601 is used to combine multiple preset binary classifiers based on a preset method to construct a multi-classifier;
[0095] A dimensionality reduction unit 602 is used to map the input invoice to be transferred out to a high-dimensional feature space according to a nonlinear mapping, and perform dimensionality reduction processing on the input invoice to be transferred out based on a principal component analysis strategy in the high-dimensional feature space to obtain a data representation after dimensionality reduction;
[0096] A determination unit 603, configured to determine the data type of the input invoice to be transferred out based on the data representation;
[0097] A first transfer unit 604 is configured to classify the data representation based on the multiple classifiers if the data type is non-linear data, and transfer the input invoice to be transferred out based on the classification result;
[0098] The second transfer unit 605 is used to add a penalty term to the objective function of the multi-class classification to obtain a modified multi-classifier if the data type is linear data, and classify the data representation based on the modified multi-classifier, and transfer the input invoice to be transferred out based on the classification result.
[0099] like Figure 7 As shown, the embodiment of this specification provides a schematic diagram of the structure of the transfer device of the input invoice, which is composed of Figure 7 It can be seen that in one or more embodiments of this specification, a device for transferring invoices includes:
[0100] at least one processor; and,
[0101] a memory communicatively connected to the at least one processor; wherein,
[0102] The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:
[0103] Combining multiple preset binary classifiers based on a preset method to construct a multi-classifier;
[0104] Mapping the input invoice to be transferred out to a high-dimensional feature space according to a nonlinear mapping, and performing dimensionality reduction processing on the input invoice to be transferred out based on a principal component analysis strategy in the high-dimensional feature space to obtain a data representation after dimensionality reduction;
[0105] Determining the data type of the input invoice to be transferred out based on the data representation;
[0106] If the data type is non-linear data, classifying the data representation based on the multiple classifiers, and transferring the input invoice to be transferred out based on the classification result;
[0107] If the data type is linear data, a penalty term is added to the objective function of the multi-class classification to obtain a modified multi-classifier, and the data representation is classified based on the modified multi-classifier, and the input invoice to be transferred out is transferred out based on the classification result.
[0108] like Figure 8 As shown, the embodiment of this specification provides a schematic diagram of a non-volatile storage medium structure, which is composed of Figure 8 It can be seen that in one or more embodiments of the present specification, a non-volatile storage medium stores computer-executable instructions, and the computer-executable instructions can: execute any of the methods described above.
[0109] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, equipment, and non-volatile computer storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.
[0110] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0111] The above description is only one or more embodiments of this specification and is not intended to limit this specification. For those skilled in the art, one or more embodiments of this specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of one or more embodiments of this specification shall be included in the scope of the claims of this specification.
Claims
1. A method for transferring out incoming invoices, characterized in that: The method comprises: Combining multiple preset binary classifiers based on a preset method to construct a multi-classifier; Mapping the input invoice to be transferred out to a high-dimensional feature space according to a nonlinear mapping, and performing dimensionality reduction processing on the input invoice to be transferred out based on a principal component analysis strategy in the high-dimensional feature space to obtain a data representation after dimensionality reduction; Determining the data type of the input invoice to be transferred out based on the data representation; If the data type is non-linear indivisible data, classifying the data representation based on the multiple classifiers, and transferring the input invoice to be transferred out based on the classification result; If the data type is linearly inseparable data, a penalty term is added to the objective function of the multi-class classification to obtain a modified multi-classifier, and the data representation is classified based on the modified multi-classifier, and the input invoice to be transferred out is transferred out based on the classification result.
2. A method for transferring out incoming invoices according to claim 1, characterized in that: Before combining a plurality of preset binary classifiers based on a preset method to construct a multi-classifier, the method further includes: The historical input invoices are used as sample points, and the optimization problem corresponding to the initial binary classifier is determined; wherein the optimization problem is to maximize the geometric interval and satisfy the function interval constraint of all sample points while minimizing the function interval constraint; Converting the optimization problem into a convex quadratic programming problem to determine a definition expression corresponding to the convex quadratic programming problem; Optimizing the variables in the definition expression based on a sequence minimum optimization algorithm to obtain an optimal definition expression; According to the optimal definition expression, an optimal separation hyperplane corresponding to the preset binary classifier is determined.
3. The method for transferring out an input invoice according to claim 1, characterized in that: Combine multiple preset binary classifiers based on preset methods to build a multi-classifier, including: Determine the classification category of the multi-classifier to be constructed based on the transferable interface corresponding to the input invoice to be transferred out; Determine the preset method corresponding to constructing the multi-classifier according to the current hardware resources corresponding to the input invoice to be transferred out and the data volume of the input invoice to be transferred out; If it is determined that the preset method is a one-to-one method, training a preset binary classifier corresponding to any two classification categories within the classification category, so as to combine the preset binary classifiers to obtain a multi-classifier; If it is determined that the preset method is a one-to-many method, a preset binary classification type is trained corresponding to each classification type to combine the preset classifiers to obtain a multi-classifier.
4. A method for transferring out incoming invoices according to claim 1, characterized in that: The input invoice to be transferred out is mapped to a high-dimensional feature space according to nonlinear mapping, and the input invoice to be transferred out is subjected to dimensionality reduction processing based on a principal component analysis strategy in the high-dimensional feature space to obtain a data representation after dimensionality reduction, specifically including: Mapping the input invoices to be transferred out to a high-dimensional feature space based on a preset kernel function, so as to calculate the covariance matrix of each input invoice to be transferred out in the high-dimensional feature space; Performing eigendecomposition on the covariance matrix to obtain eigenvalues and eigenvectors corresponding to the eigenvalues; Sorting the eigenvalues according to their sizes, screening the eigenvalues based on the ratio of the sum of the eigenvalues to the sum of the total eigenvalues, and obtaining the principal component; Project each of the input invoices to be transferred out in the high-dimensional space onto the principal component to obtain a data representation after dimensionality reduction.
5. The method for transferring out an input invoice according to claim 1, characterized in that: Determining the data type of the input invoice to be transferred out based on the data representation specifically includes: Inputting the data representation into the multi-classifier to obtain the decision boundary of the multi-classifier; If it is determined that the decision boundary has a nonlinear shape, then the data type of the input invoice to be transferred out is determined to be linearly inseparable data; If it is determined that the decision boundary appears in a linear shape, then the data type of the input invoice to be transferred out is determined to be non-linear indivisible data.
6. A method for transferring out incoming invoices according to claim 1, characterized in that: If the data type is non-linear indivisible data, classifying the data representation based on the multiple classifiers, and transferring the input invoice to be transferred out based on the classification result, specifically including: Inputting the data representation into the multi-classifier, predicting the data representation based on a plurality of preset binary classifiers in the multi-classifier, and obtaining a predicted category; Voting is performed according to the predicted categories of the preset binary classifier to determine the classification result represented by the data; Determine the transfer interface corresponding to the classification result, so as to transfer the input invoice to be transferred out to the corresponding transfer interface.
7. The method for transferring out an input invoice according to claim 1, characterized in that: Classifying the data representation based on the modified multi-classifier, and transferring the input invoice to be transferred out based on the classification result, specifically includes: inputting the data representation into the modified multi-classifier to construct an optimal separation hyperplane in a high-dimensional feature space based on the multi-classifier; Based on the classification result corresponding to the optimal separating hyperplane, the transfer-out interface corresponding to the classification result is determined, so as to transfer the input invoice to be transferred out to the corresponding transfer-out interface.
8. The device for transferring out incoming invoices according to claim 1, characterized in that: The device comprises: A construction unit, used for combining a plurality of preset binary classifiers based on a preset method to construct a multi-classifier; A dimensionality reduction unit, used for mapping the input invoice to be transferred out to a high-dimensional feature space according to a nonlinear mapping, and performing dimensionality reduction processing on the input invoice to be transferred out based on a principal component analysis strategy in the high-dimensional feature space to obtain a data representation after dimensionality reduction; A determination unit, configured to determine the data type of the input invoice to be transferred out based on the data representation; A first transfer unit, configured to classify the data representation based on the multiple classifiers if the data type is non-linear data, and transfer the input invoice to be transferred out based on the classification result; The second transfer unit is used to add a penalty term to the objective function of the multi-class classification if the data type is linear data to obtain a modified multi-classifier, classify the data representation based on the modified multi-classifier, and transfer the input invoice to be transferred out based on the classification result.
9. A transfer device for incoming invoices, characterized in that: The device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to: Combining multiple preset binary classifiers based on a preset method to construct a multi-classifier; Mapping the input invoice to be transferred out to a high-dimensional feature space according to a nonlinear mapping, and performing dimensionality reduction processing on the input invoice to be transferred out based on a principal component analysis strategy in the high-dimensional feature space to obtain a data representation after dimensionality reduction; Determining the data type of the input invoice to be transferred out based on the data representation; If the data type is non-linear data, classifying the data representation based on the multiple classifiers, and transferring the input invoice to be transferred out based on the classification result; If the data type is linear data, a penalty term is added to the objective function of the multi-class classification to obtain a modified multi-classifier, and the data representation is classified based on the modified multi-classifier, and the input invoice to be transferred out is transferred out based on the classification result.
10. A non-volatile storage medium storing computer executable instructions, characterized in that: The computer executable instructions can: execute the method described in any one of claims 1 to 7.