A method and system for online management of differential electricity charges based on a shared service model

Abstract

This invention relates to an online management method and system for differential electricity charges based on a shared service model, belonging to the field of data processing. The method includes: regional segmentation of invoice element layout information to determine element regions and attributes; statistical analysis of element difference frequency ratios to obtain element attribute splitting probabilities; furthermore, construction of an invoice matching binary tree; and sorting of invoices from the invoice pool to obtain full information on target invoices. Through the aforementioned automated and intelligent technical means, efficient matching and differential electricity charge verification are performed on invoices in the shared platform's invoice pool, thereby improving the efficiency and accuracy of electricity charge management and achieving real-time feedback on payment correction prompts. This solves the technical problems of existing methods that rely on manual verification of the difference between invoices and payment records, resulting in low verification efficiency, low accuracy of invoice matching, lack of dynamic matching mechanisms, and untimely differential electricity charge verification.

Description

Technical Field

[0001] This invention relates to the field of data processing, and in particular to a method and system for online management of differential electricity charges based on a shared service model. Background Technology

[0002] In traditional electricity bill management, businesses and organizations typically need to manually verify the difference between payment records and invoices to ensure the accuracy of electricity bill settlement. With the popularization of shared service models, centralized management and automated processing of invoice data have become a trend. Currently, invoice data processing still relies on manual labor or simple automated tools. Traditional methods struggle to efficiently match invoices with pending settlement tasks, resulting in low efficiency in verifying electricity bill discrepancies.

[0003] The main drawbacks of existing technologies are that traditional methods rely on manual verification of the difference between invoices and payment records, which is time-consuming, labor-intensive, and prone to errors. Manual verification is inefficient and lacks an automated matching mechanism, leading to settlement delays or errors at the user end and untimely reconciliation of electricity bill differences. Summary of the Invention

[0004] This invention addresses the technical problems of existing electricity bill reconciliation methods, which require manual verification of the difference between payment and invoice, resulting in extremely low efficiency and untimely reconciliation of electricity bill differences. It provides an online management method and system for electricity bill differences based on a shared service model to solve these problems.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0006] In a first aspect, the present invention provides an online management method for differential electricity charges based on a shared service model, comprising:

[0007] Based on the layout information of invoice elements, the invoice is divided into regions to determine the first element region and the first element attribute, and so on up to the Nth element region and the Nth element attribute.

[0008] Based on the first element region and the first element attribute, up to the Nth element region and the Nth element attribute, the frequency ratio of invoice element differences between each pair of invoices in the shared platform invoice pool is statistically analyzed to obtain the first element attribute splitting probability up to the Nth element attribute splitting probability.

[0009] Based on the splitting probability of the first element attribute up to the splitting probability of the Nth element attribute in descending order, the first element attribute up to the Nth element attribute are deployed step by step to construct an invoice matching binary tree;

[0010] Based on the invoice matching binary tree, the invoice pool of the shared platform is sorted according to the elements of the invoices to be matched, and the full information of the target invoices is obtained.

[0011] When there is a discrepancy in electricity charges between the full information of the target invoice and the pending settlement task, a payment correction prompt will be sent to the user.

[0012] Secondly, the present invention provides an online management system for differential electricity charges based on a shared service model, comprising:

[0013] The invoice element area and attribute module is used to divide the invoice into areas based on the invoice element layout information, and determine the first element area and the first element attribute, up to the Nth element area and the Nth element attribute;

[0014] The element attribute splitting probability module is used to perform pairwise invoice element difference frequency ratio statistics on invoices in the shared platform invoice pool based on the first element region and the first element attribute, up to the Nth element region and the Nth element attribute, to obtain the first element attribute splitting probability up to the Nth element attribute splitting probability.

[0015] The invoice matching binary tree module is used to deploy the first element attribute up to the Nth element attribute level by level according to the splitting probability of the first element attribute up to the splitting probability of the Nth element attribute in descending order, and to construct the invoice matching binary tree.

[0016] The full invoice information module is used to sort invoices in the shared platform invoice pool based on the elements of the invoices to be matched according to the invoice matching binary tree, and obtain the full information of the target invoices.

[0017] The user feedback module is used to provide a payment correction prompt to the user when there is a difference in electricity charges between the full information of the target invoice and the task to be settled. When there is no difference in electricity charges between the full information of the target invoice and the task to be settled, the task to be settled is confirmed to be executed.

[0018] Thirdly, the present invention provides an electronic device, the electronic device comprising:

[0019] processor;

[0020] Memory used to store processor-executable instructions;

[0021] The processor implements the online management method for differential electricity charges as described in the first aspect above by running the executable instructions.

[0022] Fourthly, the present invention provides a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the steps of the online differential electricity fee management method described in the first aspect above.

[0023] The beneficial effects of this invention are as follows: This technology aims to solve the technical problems of existing methods that rely on manual verification of the difference between invoices and payment records, resulting in low verification efficiency, low accuracy of invoice matching, lack of dynamic matching mechanism, and untimely reconciliation of electricity bill discrepancies. This invention divides the layout information of invoice elements into regions to determine the element regions and element attributes, and then performs statistical analysis of the frequency ratio of element differences to obtain the probability of element attribute splitting. Furthermore, it constructs an invoice matching binary tree to sort invoices in the invoice pool to obtain the full information of target invoices. Through the above-mentioned automated and intelligent technical means, invoices in the shared platform invoice pool are efficiently matched and electricity bill discrepancies are reconciled, thereby improving the efficiency and accuracy of electricity bill management and achieving the technical effect of real-time feedback of payment correction prompts. Attached Figure Description

[0024] Figure 1 A flowchart illustrating an online management method for differential electricity charges based on a shared service model provided by the present invention;

[0025] Figure 2 This invention provides a schematic diagram of the structure of an online differential electricity fee management system based on a shared service model.

[0026] Attached reference numerals: 11. Invoice element area and attribute module; 12. Element attribute splitting probability module; 13. Invoice matching binary tree module; 14. Invoice full information module; 15. User feedback module. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0029] In the description of this invention, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed herein.

[0030] Example 1, as Figure 1 As shown, this embodiment of the invention provides an online management method for differential electricity charges based on a shared service model, the execution steps of which include:

[0031] S10: Based on the invoice element layout information, divide the invoice into regions, determine the first element region and the first element attribute, and so on up to the Nth element region and the Nth element attribute;

[0032] Furthermore, this invention performs refined region segmentation and attribute recognition on invoices. By utilizing the layout information of invoice elements, the system can accurately identify the various element regions and their attributes on the invoice, providing structured data support for subsequent invoice matching and electricity bill reconciliation. This method utilizes Optical Character Recognition (OCR) technology and image processing algorithms to perform two-dimensional coordinate positioning and text recognition on the invoice, thereby achieving precise extraction of invoice elements.

[0033] Furthermore, invoices are extracted from the shared platform's invoice pool and positioned using two-dimensional coordinates to ensure the accuracy of invoice layout information. OCR technology is used to recognize the text on the invoices, converting the text content into structured text data. The recognized text boxes are located using coordinate information to determine their specific positions on the invoice. The recognized text content is then categorized by attribute to identify the corresponding invoice element attributes for each text box. The identified and categorized data is then stored in a structured manner to provide basic data support for subsequent invoice matching and electricity bill reconciliation. The entire process is completed automatically by the system, reducing manual intervention and improving processing efficiency and accuracy.

[0034] Furthermore, based on the invoice element layout information, the invoice is divided into regions to determine the first element region and the first element attribute, up to the Nth element region and the Nth element attribute. Step S10 includes:

[0035] S11: Extract the first invoice from the shared platform invoice pool;

[0036] Furthermore, the shared platform invoice pool is a system that centrally stores and manages invoice data. Invoice images or PDF files can be obtained through API interfaces or database queries.

[0037] S12: Perform two-dimensional coordinate positioning on the first invoice to obtain the full-domain coordinate information of the first invoice;

[0038] Furthermore, the global coordinate information of the first invoice is obtained through image processing algorithms in OCR technology. For example, deep learning algorithms such as PSENet or DBNet are used to detect text regions in the invoice image and locate the text regions in the invoice.

[0039] S13: Perform OCR text recognition on the first invoice to obtain the first text box and the first text, up to the Pth text box and the Pth text;

[0040] Furthermore, OCR technology can recognize the text in invoices and convert it into structured text data. Commonly used OCR tools include Tencent Cloud OCR and PaddleOCR. These tools can efficiently recognize the text in invoices, obtain the first text box and the first text through OCR text recognition, and so on up to the Pth text box and the Pth text, and record the data.

[0041] S14: Locate the first text box up to the Pth text box using the first invoice global coordinate information to obtain the first element area up to the Pth element area;

[0042] Furthermore, the identified text boxes are located using global coordinate information to obtain the element area of ​​the invoice. The position information of each text box (such as the coordinates of the top left corner and the bottom right corner) is recorded for subsequent element attribute classification.

[0043] S15: Classify the text attributes of the first text up to the Pth text to obtain the first element attributes up to the Pth element attributes;

[0044] Furthermore, the identified text content is classified according to its text attributes to determine the element attributes of each text box (such as amount, invoice number, invoice date, etc.). This step can be achieved using supervised learning algorithms, such as CRNN or Transformer models, to classify the text content.

[0045] Furthermore, invoice element layout information refers to the arrangement and distribution information of various elements on the invoice (such as amount, invoice number, invoice date, etc.), which usually includes position, size, and relative relationship; two-dimensional coordinate positioning refers to determining the specific position of each element on the invoice through a coordinate system, usually using the coordinate points of the upper left and lower right corners to define an area; OCR text recognition refers to using optical character recognition technology to convert the text content in an image into editable text data; text attribute classification refers to classifying text content and identifying the category or attribute to which the text belongs, such as the amount, invoice number, and invoice date on an invoice.

[0046] Furthermore, text attribute classification is performed on the first text up to the Pth text to obtain the first element attributes up to the Pth element attributes. Step S15 includes:

[0047] S151: Configure multiple sets of data through the management terminal, wherein any one of the multiple sets of data includes: a text record data set and a text attribute identifier dataset;

[0048] S152: Using the text attribute identifier dataset as supervision and the text record data set as input, the first classifier is trained by equally weighting the multiple sets of data.

[0049] S153: Extract the first loss dataset from the first classifier that is not classified accurately;

[0050] S154: When the amount of data in the first loss dataset is greater than or equal to the convergence data amount threshold, extract the weights of the first loss dataset in the multiple sets of data, and train the second classifier.

[0051] S155: Until the amount of data in the Qth loss dataset is less than the convergence data amount threshold, the first classifier, the second classifier up to the Qth classifier are fused to obtain a text attribute classifier. The first text up to the Pth text are classified into text attributes to obtain the first feature attributes up to the Pth feature attributes. The output of the text attribute classifier is equal to the mode value of the first classifier, the second classifier up to the Qth classifier.

[0052] Furthermore, text attribute classification refers to classifying text content and identifying its category or attribute, such as the amount, invoice number, and invoice date on an invoice; supervised learning is a machine learning method that trains a model using labeled training data, enabling the model to learn the mapping relationship between input data and output labels; the loss dataset refers to the set of data samples that were misclassified during classifier training. These data samples are used to adjust model parameters to improve classifier performance; the convergence data threshold refers to the specific operation triggered during training when the amount of data in the loss dataset reaches or exceeds a certain threshold, such as adjusting weights or stopping training; classifier fusion refers to integrating the output results of multiple classifiers, usually through voting or other strategies, to improve the accuracy and stability of classification results.

[0053] Furthermore, this invention relates to a text attribute classification method based on supervised learning and ensemble learning, used for attribute classification of text content in invoices. By configuring multiple sets of training data containing text record datasets and text attribute identifier datasets, the system can train multiple classifiers and improve classification accuracy by fusing these classifiers. This method utilizes supervised learning and ensemble learning strategies in machine learning, continuously optimizing the weights of the loss dataset to improve classifier performance, ultimately achieving high-precision classification of text attributes.

[0054] Furthermore, multiple sets of training data, including text record datasets and text attribute label datasets, are configured through the management terminal to provide a foundation for classifier training. Supervised learning methods are used, with the text attribute label dataset as supervision, to train the initial classifier, and the performance of the classifier is gradually optimized by adjusting the weights of the loss dataset. By fusing multiple classifiers and using strategies such as mode voting, the stability and accuracy of classification results are improved.

[0055] For example, suppose a text record dataset contains 1,000 text records, each corresponding to an invoice element (such as amount, invoice number, etc.); and a text attribute label dataset contains an attribute label (such as "amount", "invoice number", etc.) for each text record.

[0056] Initial training: Train the first classifier using the initial data, assuming an initial accuracy of 85%.

[0057] Loss data processing: Identify inaccurately classified text records to form the first loss dataset, assuming a data volume of 150.

[0058] Weight adjustment: Increase the weights of the first loss dataset and retrain to obtain the second classifier, assuming the accuracy is improved to 90%.

[0059] Iterative training: Continue processing new loss datasets until the amount of data in the loss dataset is less than the convergence data size threshold (assumed to be 50).

[0060] Classifier fusion: Fuse all trained classifiers to obtain the final text attribute classifier, assuming a final accuracy of 95%.

[0061] The above steps achieved the following classification effect:

[0062] Classification accuracy: improved from 85% initially to 95% ultimately; Processing time: average processing time per text record was 0.1 seconds.

[0063] S16: Add the first element region up to the Pth element region and the first element attribute up to the Pth element attribute into the first element region and the first element attribute, up to the Nth element region and the Nth element attribute;

[0064] Furthermore, the identified feature areas and attributes are added to the database of the shared platform's invoice pool to form structured invoice data.

[0065] S17: When the invoices in the shared platform invoice pool have been traversed, the first element area and the first element attribute are output, until the Nth element area and the Nth element attribute are output.

[0066] Furthermore, repeat the above steps until all invoices in the shared platform invoice pool have been processed, and output the element regions and attributes of all invoices, that is, obtain the Nth element region and the Nth element attribute.

[0067] Through the above implementation steps, this invention can efficiently identify and classify the element regions and attributes of invoices in the shared platform's invoice pool, forming structured invoice data. This provides a solid foundation for subsequent invoice matching and electricity bill reconciliation, significantly improving the efficiency and accuracy of electricity bill management, and enabling real-time feedback and payment correction prompts.

[0068] S20: Based on the first element region and the first element attribute, up to the Nth element region and the Nth element attribute, perform pairwise invoice element difference frequency ratio statistics on the invoices in the shared platform invoice pool to obtain the first element attribute splitting probability up to the Nth element attribute splitting probability.

[0069] Furthermore, based on the first element region and the first element attribute, up to the Nth element region and the Nth element attribute, the frequency ratio of pairwise invoice element differences in the invoices of the shared platform invoice pool is statistically analyzed to obtain the splitting probability of the first element attribute up to the splitting probability of the Nth element attribute. Step S20 includes:

[0070] S21: Obtain the first and second shared invoices from the shared platform's invoice pool;

[0071] S22: Based on the first element region and the first element attribute, up to the Nth element region and the Nth element attribute, extract the first feature value of the first element attribute of the first shared invoice, wherein the first feature value includes the first text and the first distribution location;

[0072] S23: Based on the first element region and the first element attribute, up to the Nth element region and the Nth element attribute, extract the second feature value of the first element attribute of the second shared invoice, wherein the second feature value includes the second text and the second distribution position;

[0073] S24: When the first text is different from the second text, or the first distribution position is different from the second distribution position, the splitting frequency of the first element attribute is incremented by one; otherwise, the consistency frequency of the first element attribute is incremented by one.

[0074] S25: When the invoices in the shared platform invoice pool have completed the pairwise invoice element comparison, the first element attribute splitting frequency and the first element attribute consistency frequency are obtained, and the first element attribute frequency is summed.

[0075] S26: Calculate the ratio of the splitting frequency of the first element attribute to the sum of the frequencies of the first element attributes, and set it as the splitting probability of the first element attribute.

[0076] Furthermore, the eigenvalue refers to the specific form of the invoice element, including information such as text content and distribution location; the split frequency refers to the number of times the element attribute causes the invoice to be split during pairwise comparison; the consistency frequency refers to the number of times the element attribute remains consistent during pairwise comparison; the frequency summation refers to the sum of the split frequency and the consistency frequency, used to calculate the split probability; the split probability refers to the ratio of the split frequency to the frequency summation, reflecting the effectiveness of the element attribute in distinguishing different invoices.

[0077] Furthermore, traditional invoice matching methods often lack a quantitative assessment of the distinguishing power of elements, leading to low matching efficiency and a high risk of errors. This invention provides a scientific basis for invoice matching by quantitatively assessing the distinguishing power of each element attribute, thereby improving matching efficiency and accuracy. Specifically, the solution aims to address the problem of traditional methods' inability to dynamically adjust matching strategies by statistically analyzing the frequency of differences in elements to optimize the matching process.

[0078] Furthermore, this invention relates to a statistical analysis-based method for calculating the frequency of differences in invoice elements, used to evaluate the distinguishing ability of invoice elements in the matching process. By comparing invoices in the shared platform's invoice pool pairwise, the frequency of differences and the frequency of similarities among the elements are statistically analyzed, and the splitting probability of each element attribute is calculated. The splitting probability reflects the effectiveness of the element attribute in distinguishing different invoices, providing data support for the subsequent construction of an invoice matching binary tree.

[0079] For example, suppose a company's shared platform invoice pool contains 5 invoices (in real-world scenarios, this can be expanded to more). The following are some key details about these invoices:

[0080]

[0081] The invoices in the invoice pool are extracted one by one and compared in pairs. First, invoice FP001 (first shared invoice) and invoice FP002 (second shared invoice) are extracted.

[0082] For invoices FP001 and FP002, extract the feature values ​​of each element attribute, including text and distribution location information:

[0083] Invoice Number | Text: FP001, Location: (50, 50, 200, 100) | Text: FP002, Location: (50, 50, 200, 100); Invoice Date | Text: 2024-07-01, Location: (50, 150, 200, 200) | Text: 2024-07-01, Location: (50, 150, 200, 200); Buyer Name | Text: Company A, Location: (50, 250, 300, 300) | Text: Company A, Location: (50, 250, 300, 300); Seller Name | Text: Company B, Location: (50, 350, 300, 400) | Text: Company C, Location: (50, 350, 300, 400); Amount | Text: ¥1000.00, Location: (50, 450, 200, 500) | Text: ¥1500.00, Location: (50, 450, 200, 500).

[0084] Furthermore, by comparing eigenvalues ​​and updating frequencies, the attribute elements of invoices FP001 and FP002 are compared:

[0085] Invoice Number: If the text is different, the splitting frequency increases by one; Invoice Date: If both the text and the distribution position are the same, the consistency frequency increases by one; Buyer Name: If both the text and the distribution position are the same, the consistency frequency increases by one; Seller Name: If the text is different, the splitting frequency increases by one; Amount: If the text is different, the splitting frequency increases by one.

[0086] Continuing to compare other invoices pairwise, the final statistical results are as follows:

[0087] List using the format of element | split frequency | consistent frequency, invoice number | 4 | 1; invoice date | 2 | 3; buyer's name | 3 | 2; seller's name | 3 | 2; amount | 4 | 1.

[0088] The frequency sum of each attribute is calculated as follows: Invoice number: 4 + 1 = 5; Invoice date: 2 + 3 = 5; Buyer's name: 3 + 2 = 5; Seller's name: 3 + 2 = 5; Amount: 4 + 1 = 5.

[0089] Calculate the probability of splitting: Invoice number: 4 / 5 = 0.8; Invoice date: 2 / 5 = 0.4; Buyer's name: 3 / 5 = 0.6; Seller's name: 3 / 5 = 0.6; Amount: 4 / 5 = 0.8.

[0090] Through the above process, the splitting probability of each element attribute was calculated. The splitting probability of invoice number and amount was the highest, both at 0.8, indicating that these two elements differ significantly in the invoice pool and can be used as key elements for invoice matching. The splitting probability of invoice date was the lowest, at 0.4, indicating that the invoice dates of most invoices are relatively concentrated. The splitting probability of buyer name and seller name was the same, both at 0.6, showing that they have a certain degree of distinguishability in the invoice pool.

[0091] These splitting probabilities will be used to construct the invoice matching binary tree, prioritizing elements with high splitting probabilities for invoice matching to improve matching efficiency and accuracy. For example, in the invoice matching process, the system first filters by invoice number, and then combines other elements such as amount to quickly locate the target invoice, achieving efficient reconciliation of electricity bill discrepancies and payment correction prompts.

[0092] S30: Based on the splitting probability of the first element attribute up to the splitting probability of the Nth element attribute in descending order, deploy the first element attribute up to the Nth element attribute level by level to construct an invoice matching binary tree;

[0093] Furthermore, based on the splitting probability of the first element attribute up to the splitting probability of the Nth element attribute in descending order, the first element attribute up to the Nth element attribute are deployed step by step to construct an invoice matching binary tree. Step S30 further includes:

[0094] S31: Perform binary element combination on the first element attribute up to the Nth element attribute to obtain a binary element attribute combination set;

[0095] S32: Traverse the set of combinations of the two element attributes, count the frequency percentage of the split frequency plus one instruction when comparing two invoices in the shared platform invoice pool, and obtain the set of trigger probabilities of the combination of the two element attributes.

[0096] S33: Based on the set of trigger probabilities of the two element attributes, traverse the set of two element attributes, and for combinations whose trigger probability is greater than or equal to the trigger probability threshold, select the attribute with the highest non-split probability and add it to the set of element attributes to be deleted, wherein the trigger probability threshold is ≥0.95.

[0097] S34: Perform three-item attribute combination cleaning on the first element attribute up to the Nth element attribute, until N-item attribute combination cleaning is performed on the first element attribute up to the Nth element attribute, to obtain a set of element attributes to be deleted;

[0098] S35: After deleting the set of element attributes to be deleted from the first element attribute to the Nth element attribute, deploy them step by step according to the splitting probability of the first element attribute to the Nth element attribute from large to small, and construct an invoice matching binary tree.

[0099] Furthermore, by deploying element attributes hierarchically and combining trigger probability analysis of binomial and multi-element combinations, the system can dynamically optimize the structure of the matching tree, improving matching efficiency and accuracy. Specifically, the solution uses splitting probabilities to sort element attributes and identifies redundant attributes by combining trigger probabilities, thereby constructing an efficient matching model.

[0100] Furthermore, a binary tree system is a system based on a binary tree data structure used to efficiently organize, store, and retrieve data. A binary tree is a tree-like data structure in which each node has at most two child nodes, usually called the left child node and the right child node. Binary tree systems construct and manipulate binary trees through specific rules and algorithms to achieve efficient data management and processing.

[0101] Furthermore, the basic structure of a binary tree includes:

[0102] Node: Each element in a binary tree is called a node, which contains a data part and two pointers to its child nodes; Root node: The topmost node of the binary tree, which has no parent node; Parent node: A node that has child nodes; Child node: A node pointed to by its parent node, which is divided into left child node and right child node; Leaf node: A node that has no child nodes.

[0103] Furthermore, the working principle of a binary tree is as follows:

[0104] Inserting data: Inserting new data into the appropriate position in the binary tree according to specific rules (such as size order); Searching for data: Searching for target data along the path of the tree by comparing node data; Deleting data: Finding the target node and adjusting the structure of the tree to maintain the properties of the tree.

[0105] For example, the splitting probability statistics calculated according to the above embodiments are as follows:

[0106]

[0107] By combining each element attribute in pairs, a set of binary element combinations is generated, resulting in a total of C(5,2) = 10 combinations:

[0108] 1. Invoice number & invoice date; 2. Invoice number & buyer's name; 3. Invoice number & seller's name; 4. Invoice number & amount; 5. Invoice date & buyer's name; 6. Invoice date & seller's name; 7. Invoice date & amount; 8. Buyer's name & seller's name; 9. Buyer's name & amount; 10. Seller's name & amount.

[0109] The system iterates through the shared platform's invoice pool, comparing each pair of invoices pairwise, and calculates the percentage of times each combination of two elements simultaneously triggers the "split frequency plus one" instruction. The results are as follows:

[0110] Combination 1: Trigger probability = 0.85; Combination 2: Trigger probability = 0.96; Combination 3: Trigger probability = 0.75; Combination 4: Trigger probability = 0.92; Combination 5: Trigger probability = 0.60; Combination 6: Trigger probability = 0.68; Combination 7: Trigger probability = 0.88; Combination 8: Trigger probability = 0.55; Combination 9: Trigger probability = 0.97; Combination 10: Trigger probability = 0.80.

[0111] Furthermore, the trigger probability threshold is set to 0.95. The set of trigger probabilities for two-element attribute combinations is traversed, and combinations with a trigger probability greater than or equal to 0.95 are selected: combination 2 (invoice number & buyer name), trigger probability 0.96; combination 9 (buyer name & amount), trigger probability 0.97.

[0112] For combination 2, the probability of splitting the invoice number (0.9) is greater than that of the buyer's name (0.3), so the buyer's name is selected and added to the set of attributes to be deleted.

[0113] For combination 9, the probability of splitting the amount (0.8) is greater than that of the buyer's name (0.3), so the buyer's name is selected and added to the set of attributes to be deleted.

[0114] At this point, the set of attributes for the element to be deleted is {buyer name}.

[0115] The remaining element attributes (invoice number, invoice date, seller name, and amount) are cleaned using three-element combinations, generating a set of four possible combinations:

[0116] 1. Invoice number & invoice date & seller's name; 2. Invoice number & invoice date & amount; 3. Invoice number & seller's name & amount; 4. Invoice date & seller's name & amount.

[0117] The trigger probabilities for each of the three combinations were calculated, and the results are as follows:

[0118] Combination 1: Trigger probability = 0.80; Combination 2: Trigger probability = 0.85; Combination 3: Trigger probability = 0.90; Combination 4: Trigger probability = 0.75.

[0119] The trigger probability threshold for the three-item combination is set to 0.90. Combination 3 (invoice number & seller name & amount) is selected with a trigger probability of 0.90.

[0120] In combination 3, the splitting probabilities of each feature attribute are compared, and the attribute with the highest non-split probability (seller name, splitting probability 0.4) is selected and added to the set of feature attributes to be deleted.

[0121] At this point, the set of attributes for the element to be deleted is {buyer's name, seller's name}.

[0122] The remaining element attributes (invoice number, invoice date, amount) are cleaned with four or more combinations. The cleaning ends when the trigger probability of all combinations is below the threshold.

[0123] Furthermore, the buyer's name and seller's name are deleted from the original element attributes, leaving the invoice number, invoice date, and amount as the remaining element attributes.

[0124] Sort by the probability of splitting the remaining element attributes from largest to smallest: Invoice number (0.9); Amount (0.8); Invoice date (0.5).

[0125] Furthermore, a binary tree is constructed by deploying it level by level. The first level selects the invoice number with the highest probability of splitting as the root node; the second level selects the amount with the second highest probability of splitting as the left child node; and the third level selects the invoice date with the lowest probability of splitting as the right child node.

[0126] Through the above process, the constructed invoice matching binary tree prioritizes invoice number matching, and then further filters based on amount and invoice date. This structure can efficiently distinguish different invoices and improve matching efficiency. For example, when there is an invoice to be matched, it is first quickly filtered at the first level of the binary tree based on the invoice number, greatly narrowing the matching range. Then, it is further filtered at the second level based on the amount, and finally, the target invoice is determined at the third level based on the invoice date. This dynamic adjustment mechanism based on splitting probability and combination trigger probability makes the invoice matching process more efficient and accurate. S40: Based on the invoice matching binary tree, the invoice pool of the shared platform is sorted based on the elements of the invoice to be matched to obtain the full information of the target invoice;

[0127] Furthermore, based on the invoice matching binary tree, the shared platform invoice pool is sorted according to the elements of the invoices to be matched to obtain the full information of the target invoices. Step S40 includes:

[0128] S41: Based on the invoice, match the binary tree and extract the first-level element attributes;

[0129] S42: Based on the first-level element attributes, extract the matching positions of the first-level element attributes from the invoice elements to be matched;

[0130] S43: By comparing the location to be matched with the invoice pool of the shared platform through the first-level element attribute, a set of first-level location sorted invoices with consistent characteristics is obtained;

[0131] S44: When the number of the first-level location sorting invoice set is not equal to 1, extract the first-level element attribute text to be matched;

[0132] S45: The text to be matched by the first-level element attribute is compared with the first-level location sorting invoice set to obtain a first-level text sorting invoice set with consistent features;

[0133] S46: When the number of invoices in the first-level text sorting set is not equal to 1, the invoices are then matched with the second-level sorting of the binary tree;

[0134] S47: Output the full information of the target invoice when the number of sorted invoice sets is equal to 1.

[0135] Furthermore, the first-level element attribute refers to the element attribute of the root node in the invoice matching binary tree, which is usually the attribute with the highest splitting probability; the position to be matched refers to the specific location information of the element attribute in the invoice to be matched, which is used to compare the position with the invoices in the invoice pool; the first-level position sorted invoice set refers to the invoice set obtained by comparing the position of the first-level element attributes; the first-level text sorted invoice set refers to the invoice set obtained by comparing the text of the first-level element attributes; the second-level sorting refers to the process of further sorting using the element attributes of the second-level nodes of the invoice matching binary tree; and the full information of the target invoice refers to all the detailed information of the finally matched unique invoice, including all element attributes and content.

[0136] Furthermore, traditional invoice matching methods often rely on simple linear searches or fixed rules, making it difficult to efficiently process large-scale invoice data and prone to errors due to non-unique matching conditions. This invention proposes an invoice sorting method based on an invoice matching binary tree, which is used to efficiently and accurately match target invoices from the shared platform's invoice pool. By using the element attributes of the invoice matching binary tree level by level for sorting, the system can gradually narrow down the matching range and finally find the unique target invoice. This method combines the efficiency of tree data structures with the accuracy of multi-level filtering mechanisms, ensuring the efficiency and accuracy of the matching process.

[0137] For example, assume that the following invoice matching binary tree has been constructed through the steps of the above embodiments:

[0138] Root node (first-level element attribute): Invoice number (split probability 0.9); Left child node (second-level element attribute): Amount (split probability 0.8); Left child node (third-level element attribute): Invoice date (split probability 0.5); Right child node (third-level element attribute): Buyer name (split probability 0.3); Right child node (second-level element attribute): Seller name (split probability 0.4).

[0139] Suppose there is an invoice to be matched, with the following information: Invoice Number: FP005; Amount: ¥3000.00; Invoice Date: 2024-07-03; Buyer: Company A; Seller: Company B.

[0140] Extract primary element attributes and perform position comparison: Based on the invoice matching binary tree, extract the primary element attribute "invoice number". Extract the matching position of "invoice number" from the invoice elements to be matched, assuming it is (50,50,200,100). Compare it with the invoice pool of the shared platform to find the invoice with the invoice number position (50,50,200,100) and obtain the primary position sorted invoice set, assuming there are 10 invoices.

[0141] Extract the primary feature attribute text to be matched and perform text comparison: Extract the text "FP005" of the invoice number to be matched and compare it with the invoices in the primary location sorting invoice set to obtain the primary text sorting invoice set. Assume there are 2 invoices.

[0142] Entering the second-level sorting (amount): Enter the left child node "amount" of the invoice matching binary tree, extract the "amount" position of the invoice to be matched, assuming it is (50,450,200,500), compare it with the invoices in the first-level text sorting invoice set, and obtain the second-level position sorting invoice set, assuming there are 2 invoices.

[0143] Extract the "Amount" text "￥3000.00" from the invoice to be matched, and compare it with the invoices in the secondary location sorting invoice set to obtain the secondary text sorting invoice set. Assume there is 1 invoice.

[0144] Since the number of sorted invoices is equal to 1, output the full information of the target invoice: Invoice number: FP005; Amount: ¥3000.00; Invoice date: 2024-07-03; Buyer name: Company A; Seller name: Company B.

[0145] This invention uses a hierarchical sorting mechanism based on a binary tree for invoice matching to gradually narrow down the matching range, ultimately achieving efficient and accurate invoice matching. This method not only improves matching efficiency but also reduces manual intervention, significantly enhancing both matching efficiency and accuracy.

[0146] S50: When there is a discrepancy in electricity charges between the full information of the target invoice and the pending settlement task, a payment correction prompt is sent to the user.

[0147] S51: When there is no difference in electricity charges between the full information of the target invoice and the pending settlement task, confirm the execution of the pending settlement task.

[0148] The online management method for differential electricity charges based on a shared service model provided in this invention has at least the following technical effects:

[0149] This invention addresses the limitations of traditional methods in refining invoice segmentation and attribute identification by segmenting invoice elements based on their layout information, thus enabling structured processing of invoice data. It overcomes the lack of a dynamic matching mechanism through element difference frequency statistics and splitting probability calculation, improving matching accuracy and adaptability. Furthermore, it addresses the low matching efficiency of traditional methods by constructing an invoice matching binary tree and rapidly narrows the matching range through a step-by-step sorting mechanism, significantly improving efficiency and reducing unnecessary comparison steps. The multi-level sorting mechanism, starting from the root node and using element attributes to sort step-by-step until a unique matching invoice is found, solves the problem of inaccurate target invoice identification in existing technologies, ensuring accurate matching results. The automated and intelligent matching process overcomes the low efficiency and high error rate of traditional methods relying on manual verification. Finally, the real-time feedback of payment correction prompts addresses the inability of traditional methods to provide real-time correction prompts when electricity bill discrepancies are discovered, helping users promptly identify and address such discrepancies and improving the timeliness and accuracy of electricity bill management.

[0150] Through the above-mentioned technical means, this invention solves the technical problems in traditional electricity fee management, such as low efficiency of manual verification, insufficient invoice element processing capabilities, lack of dynamic matching mechanism, and untimely reconciliation of electricity fee differences. Ultimately, it achieves the technical effects of efficient invoice matching, accurate reconciliation of electricity fee differences, and real-time feedback of payment correction prompts, significantly improving the efficiency and accuracy of electricity fee management.

[0151] Example 2, as Figure 2 As shown, based on the same inventive concept as the online management method for differential electricity charges based on a shared service model provided in Embodiment 1, this embodiment of the invention also provides an online management system for differential electricity charges based on a shared service model, comprising:

[0152] The invoice element area and attribute module 11 is used to divide the invoice into areas based on the invoice element layout information, and determine the first element area and the first element attribute, up to the Nth element area and the Nth element attribute;

[0153] The element attribute splitting probability module 12 is used to perform pairwise invoice element difference frequency ratio statistics on invoices in the shared platform invoice pool based on the first element region and the first element attribute, up to the Nth element region and the Nth element attribute, to obtain the first element attribute splitting probability up to the Nth element attribute splitting probability.

[0154] The invoice matching binary tree module 13 is used to deploy the first element attribute up to the Nth element attribute level by level according to the splitting probability of the first element attribute up to the splitting probability of the Nth element attribute in descending order, and to construct the invoice matching binary tree.

[0155] The full invoice information module 14 is used to sort invoices in the shared platform invoice pool based on the elements of the invoices to be matched according to the invoice matching binary tree, and obtain the full invoice information of the target invoices.

[0156] User feedback module 15 is used to provide payment correction prompts to the user when there is a difference in electricity charges between the full information of the target invoice and the pending settlement task.

[0157] Furthermore, the user feedback module 15 executes the following steps: when there is no difference in electricity charges between the full information of the target invoice and the pending settlement task, confirm the execution of the pending settlement task.

[0158] Furthermore, the execution steps of the invoice element area and attribute module 11 include:

[0159] Retrieve the first invoice from the shared platform's invoice pool;

[0160] Perform two-dimensional coordinate positioning on the first invoice to obtain the full-domain coordinate information of the first invoice;

[0161] Perform OCR text recognition on the first invoice to obtain the first text box and the first text, up to the Pth text box and the Pth text;

[0162] The first text box up to the Pth text box is located using the global coordinate information of the first invoice to obtain the first element area up to the Pth element area;

[0163] Classify the text attributes of the first text up to the Pth text to obtain the first element attributes up to the Pth element attributes;

[0164] Add the first feature region up to the Pth feature region, and the first feature attribute up to the Pth feature attribute, into the first feature region and the first feature attribute, up to the Nth feature region and the Nth feature attribute;

[0165] When the invoices in the shared platform invoice pool have been traversed, the first element area and the first element attribute are output, until the Nth element area and the Nth element attribute are output.

[0166] Furthermore, the execution steps of the invoice element area and attribute module 11 also include:

[0167] Through the management terminal, multiple sets of data can be configured, wherein any one of the multiple sets of data includes: a set of text record data and a set of text attribute identifiers;

[0168] Using the text attribute identifier dataset as supervision and the text record dataset as input, the first classifier is trained by equally weighting the multiple sets of data.

[0169] Extract the first loss dataset from the first classifier that is inaccurately classified;

[0170] When the amount of data in the first loss dataset is greater than or equal to the threshold of the amount of convergent data, the weights of the first loss dataset in the multiple sets of data are extracted, and a second classifier is trained.

[0171] Until the amount of data in the Qth loss dataset is less than the convergence data amount threshold, the first classifier, the second classifier up to the Qth classifier are fused to obtain a text attribute classifier. The first text up to the Pth text are classified into text attributes to obtain the first feature attributes up to the Pth feature attributes. The output of the text attribute classifier is equal to the mode value of the first classifier, the second classifier up to the Qth classifier.

[0172] Furthermore, the feature attribute splitting probability module 12 includes:

[0173] Obtain the first and second shared invoices from the shared platform's invoice pool;

[0174] Based on the first element region and the first element attribute, up to the Nth element region and the Nth element attribute, extract the first feature value of the first element attribute of the first shared invoice, wherein the first feature value includes the first text and the first distribution location;

[0175] Based on the first element region and the first element attribute, up to the Nth element region and the Nth element attribute, extract the second feature value of the first element attribute of the second shared invoice, wherein the second feature value includes the second text and the second distribution position;

[0176] When the first text is different from the second text, or the first distribution position is different from the second distribution position, the splitting frequency of the first element attribute is incremented by one; otherwise, the consistency frequency of the first element attribute is incremented by one.

[0177] Once the invoices in the shared platform's invoice pool have completed pairwise invoice element comparison, the frequency of splitting the first element attribute and the frequency of consistency of the first element attribute are obtained, and the sum of the frequencies of the first element attribute is calculated.

[0178] Calculate the ratio of the split frequency of the first element attribute to the sum of the frequencies of the first element attributes, and set it as the split probability of the first element attribute.

[0179] Furthermore, the invoice matching binary tree module 13 includes:

[0180] Perform binary element combination on the first element attribute up to the Nth element attribute to obtain a binary element attribute combination set;

[0181] Traverse the set of combinations of the two element attributes, and count the frequency percentage of the split frequency increment instruction triggered simultaneously when comparing pairs of invoices in the shared platform invoice pool, to obtain the set of trigger probabilities of the two element attribute combinations.

[0182] Based on the set of trigger probabilities for the combination of the two element attributes, the set of the two element attribute combinations is traversed. For combinations whose combination trigger probability is greater than or equal to the trigger probability threshold, the attribute with the highest non-split probability is selected and added to the set of element attributes to be deleted, wherein the trigger probability threshold is ≥0.95.

[0183] Perform three-item attribute combination cleaning on the first element attribute up to the Nth element attribute, and then perform N-item attribute combination cleaning on the first element attribute up to the Nth element attribute to obtain a set of element attributes to be deleted;

[0184] After deleting the set of element attributes to be deleted from the first element attribute to the Nth element attribute, the invoice matching binary tree is constructed by deploying the elements level by level according to the splitting probability of the first element attribute down to the splitting probability of the Nth element attribute.

[0185] Furthermore, the full invoice information module 14 includes:

[0186] Based on the invoice, a binary tree is matched to extract the first-level element attributes;

[0187] Based on the primary element attributes, extract the matching positions of the primary element attributes from the invoice elements to be matched;

[0188] By comparing the location to be matched with the invoice pool of the shared platform using the first-level element attributes, a set of first-level location sorted invoices with consistent characteristics is obtained;

[0189] When the number of invoices in the first-level location sorting set is not equal to 1, extract the first-level element attribute text to be matched;

[0190] By comparing the text to be matched with the first-level element attributes with the first-level location sorting invoice set, a first-level text sorting invoice set with consistent features is obtained.

[0191] When the number of invoices in the first-level text sorting set is not equal to 1, the invoices are then matched with the second-level sorting of the binary tree;

[0192] When the number of sorted invoice sets equals 1, output the full information of the target invoice.

[0193] Furthermore, the user feedback module 15 includes:

[0194] Used to send a payment correction prompt to the user when there is a difference in electricity charges between the full information of the target invoice and the pending settlement task.

[0195] If there is no difference in electricity charges between the full information of the target invoice and the pending settlement task, confirm the execution of the pending settlement task.

[0196] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0197] This invention provides a computer-readable storage medium storing computer instructions that, when executed by a processor, implement the steps of the differential electricity charge online management method described in the above-described method embodiments. Those skilled in the art will understand that embodiments of this invention can be provided as methods, systems, or computer program products. Therefore, this invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this invention can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0198] This invention provides an electronic device comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor implements the differential electricity charge online management method of the above-described method embodiments by executing the executable instructions. This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0199] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0200] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0201] Although preferred embodiments of the invention have been described, those skilled in the art, once they have learned the basic inventive concept, can make other changes and modifications to these embodiments.

[0202] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of this invention and its equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for online management of differential electricity charges based on a shared service model, characterized in that, The method comprises the following steps: According to the invoice element layout information, the invoice is regionally segmented to determine the first element region and the first element attribute, and the Nth element region and the Nth element attribute; Based on the first element region and the first element attribute, the Nth element region and the Nth element attribute, the invoice of the shared platform invoice pool is counted for invoice element difference frequency proportion by two, and the first element attribute splitting probability to the Nth element attribute splitting probability is obtained; According to the first element attribute splitting probability to the Nth element attribute splitting probability from large to small, the first element attribute to the Nth element attribute is deployed step by step to construct an invoice matching binary tree; According to the invoice matching binary tree, the shared platform invoice pool is sorted based on the to-be-matched invoice element to obtain the target invoice full information; When there is a difference in electricity charges between the target invoice full information and the to-be-settled task, the user end is fed back with a payment correction prompt.

2. The method of claim 1, wherein, Further comprising: When there is no difference in electricity charges between the target invoice full information and the to-be-settled task, the to-be-settled task is confirmed to be executed.

3. The method of claim 1, wherein, According to the invoice element layout information, the invoice is regionally segmented to determine the first element region and the first element attribute, and the Nth element region and the Nth element attribute, comprising: Extracting the first invoice from the shared platform invoice pool; Positioning the first invoice in two-dimensional coordinates to obtain first global coordinate information of the first invoice; Performing OCR text recognition on the first invoice to obtain the first text box and the first text to the Pth text box and the Pth text; Positioning the first text box to the Pth text box through the first invoice global coordinate information to obtain the first element region to the Pth element region; Text attribute classification is performed on the first text to the Pth text to obtain the first element attribute to the Pth element attribute; The first element region to the Pth element region and the first element attribute to the Pth element attribute are added to the first element region and the first element attribute to the Nth element region and the Nth element attribute; When the invoice of the shared platform invoice pool is traversed, the first element region and the first element attribute to the Nth element region and the Nth element attribute are outputted.

4. The method of claim 3, wherein, Text attribute classification is performed on the first text to the Pth text to obtain the first element attribute to the Pth element attribute, comprising: Configuring multiple sets of data through the management end, wherein any one of the multiple sets of data comprises a text record data set and a text attribute identification data set; Taking the text attribute identification data set as supervision and the text record data set as input, the first classifier is trained by equally configuring the multiple sets of data; Extracting the first loss data set with inaccurate classification in the first classifier; When the data amount of the first loss data set is greater than or equal to the convergence data amount threshold, the weight of the first loss data set in the multiple sets of data is extracted, and the second classifier is trained. Until the data volume of the Qth loss data set is less than the convergence data volume threshold, fusing the first classifier, the second classifier, and the Qth classifier obtains a text attribute classifier, performing text attribute classification on the first text to the Pth text obtains the first element attribute to the Pth element attribute, and the output of the text attribute classifier is equal to the mode value of the first classifier, the second classifier, and the Qth classifier.

5. The method of claim 1, wherein, Based on the first element region and the first element attribute, and the Nth element region and the Nth element attribute, performing two-by-two invoice element difference frequency ratio statistics on invoices in the shared platform invoice pool obtains first element attribute split probability to Nth element attribute split probability, including: Obtaining a first shared invoice and a second shared invoice in the shared platform invoice pool; Based on the first element region and the first element attribute, and the Nth element region and the Nth element attribute, extracting a first feature value of the first element attribute of the first shared invoice, wherein the first feature value includes a first text and a first distribution position; Based on the first element region and the first element attribute, and the Nth element region and the Nth element attribute, extracting a second feature value of the first element attribute of the second shared invoice, wherein the second feature value includes a second text and a second distribution position; When the first text is different from the second text, or the first distribution position is different from the second distribution position, the split frequency of the first element attribute is incremented by one, otherwise, the consistent frequency of the first element attribute is incremented by one; When the two-by-two invoice element comparison of the invoices in the shared platform invoice pool is completed, obtaining the first element attribute split frequency and the first element attribute consistent frequency, and then performing first element attribute frequency sum statistics; Calculating the ratio of the first element attribute split frequency to the first element attribute frequency sum, and setting it as the first element attribute split probability.

6. The method of claim 1, wherein, According to the first element attribute split probability to the Nth element attribute split probability from large to small, performing step-by-step deployment on the first element attribute to the Nth element attribute to construct an invoice matching binary tree, further including: Performing two-item element combination on the first element attribute to the Nth element attribute to obtain a two-item element attribute combination set; Traversing the two-item element attribute combination set to obtain a two-item element attribute combination trigger probability set, and then performing two-by-two invoice comparison on the shared platform invoice pool to obtain a frequency ratio of the split frequency plus one instruction triggered at the same time; Based on the two-item element attribute combination trigger probability set, traversing the two-item element attribute combination set, and then selecting a non-split probability maximum attribute for a combination with a combination trigger probability greater than or equal to a trigger probability threshold to add it to a to-be-deleted element attribute set, wherein the trigger probability threshold is greater than or equal to 0.95; Performing three-item element attribute combination cleaning on the first element attribute to the Nth element attribute until performing N-item element attribute combination cleaning on the first element attribute to the Nth element attribute to obtain the to-be-deleted element attribute set; After the element attribute set to be deleted is deleted from the first element attribute to the Nth element attribute, the first element attribute splitting probability to the Nth element attribute splitting probability is arranged from large to small in a step-by-step manner to build an invoice matching binary tree.

7. The method of claim 1, wherein, According to the invoice matching binary tree, the shared platform invoice pool is sorted based on the invoice elements to be matched to obtain target invoice full information, including: According to the invoice matching binary tree, the first element attribute is extracted; According to the first element attribute, the first element attribute matching position is extracted from the invoice elements to be matched; By comparing the first element attribute matching position with the shared platform invoice pool, a first position sorted invoice set with consistent characteristics is obtained; When the number of the first position sorted invoice set is not equal to 1, the first element attribute matching text is extracted; By comparing the first element attribute matching text with the first position sorted invoice set, a first text sorted invoice set with consistent characteristics is obtained; When the number of the first text sorted invoice set is not equal to 1, the second sorting of the invoice matching binary tree is further performed; Until the number of the sorted invoice set is equal to 1, the target invoice full information is output.

8. A system for online management of differential electricity charges based on a shared service model, characterized in that, It includes: An invoice element area and attribute module is used to divide the invoice into regions according to the invoice element layout information, determine the first element region and the first element attribute, and determine the Nth element region and the Nth element attribute; An element attribute splitting probability module is used to statistically obtain the first element attribute splitting probability to the Nth element attribute splitting probability by comparing the first element region and the first element attribute to the Nth element region and the Nth element attribute; An invoice matching binary tree module is used to arrange the first element attribute to the Nth element attribute in a step-by-step manner from large to small according to the first element attribute splitting probability to the Nth element attribute splitting probability to build an invoice matching binary tree; An invoice full information module is used to sort the shared platform invoice pool based on the invoice elements to be matched according to the invoice matching binary tree to obtain target invoice full information; A user feedback module is used to feedback the amount correction prompt to the user when there is a difference in electricity between the target invoice full information and the task to be settled, and confirm the execution of the task to be settled when there is no difference in electricity between the target invoice full information and the task to be settled.

9. An electronic device, comprising: The electronic device includes: a processor; a memory for storing processor executable instructions; wherein the processor implements the method for online management of the difference in electricity as claimed in any one of claims 1-7 by running the executable instructions.

10. A computer readable storage medium having stored thereon computer instructions, wherein, The instructions are executed by the processor to implement the steps of the method for online management of the difference in electricity as claimed in any one of claims 1-7.

Images