Asset evaluation method and system based on enterprise contract invoice driving
By building an artificial intelligence-driven asset appraisal model, the problems of high labor costs and poor accuracy of traditional asset appraisal methods are solved, efficient and accurate asset appraisal is achieved, and the application value of corporate contract invoices and the convenience of appraisal reports are enhanced.
Patent Information
- Application Number
- CN202411159463.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-22
- Publication Date
- 2024-11-29
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing asset appraisal methods rely on manual collection and analysis of financial statements and contract data, resulting in high labor costs, large workload, poor accuracy, and inability to effectively explore the deep relationship between corporate contract invoices and assets, resulting in inaccurate evaluation results.
Using an artificial intelligence-based method, a text recognition model, a corporate portrait generation model and an asset evaluation model are constructed. By pre-processing, text recognition, data verification and feature fusion of enterprise business data and contract invoice files, asset evaluation is carried out using deep learning algorithms. , generate an asset appraisal report.
Improve the intelligence and accuracy of asset appraisal, reduce labor costs, and enhance the application effect of enterprise contract invoices. Through the enterprise portrait generation model combined with the asset appraisal model, the asset appraisal results are visually displayed, which improves the authenticity and convenience of the appraisal. sex.
Smart Images

Figure CN119047880A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of asset evaluation, and in particular relates to an asset evaluation method and system driven by enterprise contract invoices. Background Art
[0002] With the development of the market economy, transactions between enterprises are becoming more and more frequent. Contract invoices, as important vouchers for transactions, carry information on the economic activities of enterprises. Among them, the contract document is a written agreement reached between the enterprise and the counterparty, which stipulates the terms and conditions of the transaction, including the description, quantity, price, delivery time, etc. of the goods or services; the contract document is the basis for the enterprise to carry out business activities, but it does not directly constitute an asset; the invoice document is the financial voucher of the transaction, which records the amount, tax amount and other information of the transaction, and is used for tax declaration and financial accounting. In accounting processing, the invoice document is the key basis for confirming income, calculating taxes and conducting cost accounting. Therefore, the enterprise contract invoice document plays an important role in the enterprise asset evaluation.
[0003] In the existing technology, the traditional asset valuation method mainly relies on manual collection and analysis of the company's financial statements, contracts and other information, which requires large manpower costs and workload, and is easily limited by the subjective judgment of personnel, resulting in poor accuracy of asset valuation results; and in the asset valuation process, it is impossible to dig out the deep relationship between the company's contract invoices and the company's assets, resulting in poor application of the company's contract invoices in asset valuation, and unable to provide effective support for the asset valuation results. Summary of the invention
[0004] In order to solve the problems of high manpower cost, heavy workload, poor accuracy and poor application effect of enterprise contract invoices in the prior art of asset valuation, the present invention aims to provide an asset valuation method and system driven by enterprise contract invoices.
[0005] The technical solution adopted by the present invention is:
[0006] An asset valuation method based on enterprise contract invoice drive includes the following steps:
[0007] Based on the historical business operation data of different first-tier enterprises and several historical enterprise contract and invoice documents, artificial intelligence algorithms are used to build text recognition models, enterprise portrait generation models, and asset evaluation models;
[0008] Collecting the real-time business operation data and several real-time enterprise contract invoice files of the second enterprise, and preprocessing them to obtain the preprocessed real-time business operation data and several preprocessed real-time enterprise contract invoice files;
[0009] According to the pre-processed real-time enterprise operation data, the enterprise portrait generation model is used to generate the enterprise portrait, thereby obtaining the real-time enterprise portrait of the second enterprise;
[0010] Based on a number of pre-processed real-time enterprise contract invoice files, a text recognition model is used to perform text recognition to obtain a number of real-time enterprise contract invoice data;
[0011] According to the enterprise contract invoice database preset by the second enterprise, data verification is performed on a number of real-time enterprise contract invoice data to obtain a number of verified real-time enterprise contract invoice data;
[0012] Based on the real-time enterprise portrait and some verified real-time enterprise contract invoice data, use the asset valuation model to conduct asset valuation and obtain real-time asset valuation results;
[0013] According to the real-time asset evaluation results, a real-time asset evaluation report of the second enterprise is generated using a preset asset evaluation report template.
[0014] Furthermore, based on the historical business operation data of different first enterprises and several historical enterprise contract invoice documents, artificial intelligence algorithms are used to build a text recognition model, an enterprise portrait generation model, and an asset evaluation model, including the following steps:
[0015] Collecting historical business operation data and several historical enterprise contract invoice files of different first enterprises, performing preprocessing, and obtaining several preprocessed historical business operation data and several preprocessed historical enterprise contract invoice files;
[0016] Based on several pre-processed historical enterprise contract invoice documents, an image text recognition algorithm is used to build a text recognition model and generate several historical enterprise contract invoice data;
[0017] Based on some pre-processed historical business data, a deep learning algorithm is used to build a business portrait generation model and generate some historical business portraits;
[0018] Based on the historical corporate portraits of different first-tier companies and the corresponding historical corporate contract invoice data, an asset valuation model is constructed using a deep learning algorithm.
[0019] Furthermore, preprocessing includes standardization processing, data dimensionality reduction and enterprise portrait label adding of historical enterprise operating data, as well as data cleaning, format processing, image processing and label adding of historical enterprise contract invoice documents.
[0020] Furthermore, the text recognition model is constructed based on the CTPN-CRNN algorithm.
[0021] Furthermore, the enterprise portrait generation model is constructed based on the N-GAN-MLP algorithm, where N is the preset enterprise portrait focus dimension.
[0022] Furthermore, the asset valuation model is constructed based on the CNN-RF-MLP algorithm.
[0023] Further, based on the pre-processed real-time enterprise operation data, using the enterprise portrait generation model, the enterprise portrait is generated to obtain the real-time enterprise portrait of the second enterprise, including the following steps:
[0024] Inputting the preprocessed real-time enterprise operation data into the enterprise portrait generation model, and using the enterprise portrait generation model to extract the real-time enterprise operation data focus feature group of the preprocessed real-time enterprise operation data;
[0025] Perform feature fusion on N real-time enterprise management data focus features in the real-time enterprise management data focus feature group to obtain the real-time enterprise management data focus fusion feature;
[0026] Based on the real-time enterprise operation data, the enterprise portrait is generated to obtain the real-time enterprise portrait of the second enterprise.
[0027] Further, based on a number of pre-processed real-time enterprise contract invoice files, a text recognition model is used to perform text recognition to obtain a number of real-time enterprise contract invoice data, including the following steps:
[0028] Input the pre-processed real-time enterprise contract invoice file into the text recognition model to locate the text, and obtain a number of real-time text location areas and corresponding real-time text location positions;
[0029] According to the real-time text positioning position, a plurality of real-time text positioning areas of the pre-processed real-time enterprise contract invoice file are cropped to obtain a plurality of real-time text positioning area images;
[0030] According to a number of real-time text positioning area images, text recognition is performed, and according to the corresponding real-time text positioning positions, text reorganization is performed to obtain the corresponding real-time enterprise contract invoice data;
[0031] All pre-processed real-time enterprise contract invoice files are traversed to obtain a number of real-time enterprise contract invoice data.
[0032] Furthermore, based on the real-time enterprise portrait and a number of verified real-time enterprise contract invoice data, an asset valuation model is used to perform an asset valuation to obtain a real-time asset valuation result, including the following steps:
[0033] Input the real-time enterprise portrait and a number of verified real-time enterprise contract invoice data into the asset valuation model, and use the asset valuation model to extract the real-time enterprise portrait features of the real-time enterprise portrait;
[0034] Performing matrix conversion on a plurality of verified real-time enterprise contract invoice data to obtain a corresponding real-time enterprise contract invoice data matrix, and extracting a real-time enterprise contract invoice data feature group of the real-time enterprise contract invoice data matrix;
[0035] Perform feature fusion on the real-time enterprise portrait feature and the real-time enterprise contract invoice data feature group to obtain the real-time enterprise portrait contract invoice fusion feature;
[0036] Based on the fusion features of real-time enterprise portrait contract and invoice, asset appraisal is carried out to obtain real-time asset appraisal results.
[0037] An asset evaluation system driven by enterprise contract invoices is used to implement an asset evaluation method. The system includes a model building unit, a preprocessing unit, an enterprise portrait generation unit, a text recognition unit, a data verification unit, an asset evaluation unit and a report generation unit which are connected in sequence.
[0038] The beneficial effects of the present invention are:
[0039] The present invention discloses an asset evaluation method and system driven by enterprise contract invoices. By performing image conversion and text recognition, the enterprise contract invoice documents are digitally converted. As important supporting data for asset evaluation, the application effect and data value of the enterprise contract invoice are enhanced, while the intelligence level, evaluation efficiency and practicality of asset evaluation are improved, the manpower cost investment is reduced, and the workload is alleviated. Through the enterprise portrait generation model, asset evaluation is performed in combination with the enterprise portrait, which improves the authenticity and comprehensiveness. Through the asset evaluation model constructed by the deep learning algorithm, the deep relationship between the enterprise portrait, the enterprise contract invoice data and the enterprise assets is excavated, which improves the accuracy of the evaluation. In addition, through the generated asset evaluation report, the asset evaluation results of the enterprise are intuitively displayed, which improves the convenience.
[0040] Other beneficial effects of the present invention will be further described in the specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] Figure 1 It is a flowchart of the asset valuation method driven by enterprise contract invoices in the present invention.
[0042] Figure 2 It is a structural block diagram of the asset evaluation system driven by enterprise contract invoices in the present invention. DETAILED DESCRIPTION
[0043] The present invention will be further explained below in conjunction with the accompanying drawings and specific embodiments.
[0044] Embodiment 1:
[0045] like Figure 1 As shown, this embodiment provides an asset valuation method driven by enterprise contract invoices, including the following steps:
[0046] S1: Based on the historical business operation data of different first-tier enterprises and several historical enterprise contract and invoice documents, use artificial intelligence algorithms to build a text recognition model, an enterprise portrait generation model, and an asset evaluation model, including the following steps:
[0047] S1-1: Collecting historical business operation data and several historical enterprise contract invoice files of different first enterprises, preprocessing them, and obtaining several preprocessed historical business operation data and several preprocessed historical enterprise contract invoice files, including the following steps:
[0048] S1-1-1: Collect historical business operation data of different first-class enterprises and several historical enterprise contract invoice documents;
[0049] The historical enterprise contract and invoice files include historical enterprise contract files and historical enterprise invoice files;
[0050] S1-1-2: Standardize some historical business data, reduce data dimension, and add enterprise portrait labels to obtain some pre-processed historical business data;
[0051] Business operation data includes various public or non-public data of the enterprise, including but not limited to financial statements, market behavior, legal proceedings, patent information, news reports, management background, etc.;
[0052] S1-1-3: Perform data cleansing on several historical enterprise contract invoice files to obtain several cleaned historical enterprise contract invoice files;
[0053] Data cleaning includes screening duplicate historical enterprise contract invoice documents and removing incorrect historical enterprise contract invoice documents to improve data quality and the speed and accuracy of model training;
[0054] S1-1-4: Perform format conversion and image format unification processing on a number of cleaned historical enterprise contract invoice files to obtain a number of format-processed and pre-processed historical enterprise contract invoice files;
[0055] The data formats of historical enterprise contract invoice files include images, tables, text, etc. In order to facilitate the subsequent asset evaluation, the file formats are uniformly converted into image formats, and different image formats, such as jpg and png, are unified to improve the efficiency of subsequent model analysis;
[0056] S1-1-5: Gaussian denoising and grayscale processing are performed on several format-processed and pre-processed historical enterprise contract invoice files to obtain several image-pre-processed historical enterprise contract invoice files;
[0057] S1-1-6: marking historical text location areas, adding text labels, and adding asset evaluation labels to a number of historical enterprise contract invoice files after image preprocessing, to obtain a number of preprocessed historical enterprise contract invoice files;
[0058] S1-2: Based on several pre-processed historical enterprise contract invoice files, a text recognition model is constructed using the Detecting Text in Natural Image with Connectionist Text Proposal Network (CTPN)-Convolutional Recurrent Neural Network (CRNN) algorithm, and several historical enterprise contract invoice data are generated;
[0059] The text recognition model is constructed based on the CTPN-CRNN algorithm, and the text recognition model includes a text positioning module constructed based on the CTPN algorithm and a text recognition module constructed based on the CRNN algorithm, including the following steps:
[0060] S1-2-1: Divide a number of pre-processed historical enterprise contract invoice files into the first model training set and the first model test set in a ratio of 7:3;
[0061] S1-2-2: Use the CTPN algorithm to construct an initial text positioning module, input the first model training set, optimize the initial text positioning module, obtain an optimized text positioning module, and obtain several historical text positioning areas and corresponding historical text positioning positions of the pre-processed historical enterprise contract invoice files;
[0062] S1-2-3: According to the historical text positioning positions, several historical text positioning areas of the historical enterprise contract invoice files preprocessed in the first model training set are cropped to obtain several historical text positioning area images;
[0063] S1-2-4: Use the CRNN algorithm to build an initial text recognition module, and input several historical text positioning area images to optimize the initial text recognition module and obtain an optimized text recognition module;
[0064] S1-2-5: Integrate the optimized text positioning module and the optimized text recognition module to obtain an optimized text recognition model, input a first model test set, perform a model test on the optimized text recognition model, and obtain a first model test accuracy rate;
[0065] S1-2-6: If the test accuracy of the first model is greater than the first accuracy threshold, output the final text recognition model, otherwise, continue to perform optimization training;
[0066] S1-3: Based on some pre-processed historical business data, use the N-Generative Adversarial Network (GAN)-Multilayer Perceptron (MLP) algorithm to build a corporate portrait generation model and generate some historical corporate portraits;
[0067] The enterprise portrait generation model is constructed based on the N-GAN-MLP algorithm, where N is the preset enterprise portrait focus dimension, and the enterprise portrait generation model includes N focus feature extraction modules constructed based on the GAN algorithm and a first fusion prediction module constructed based on the MLP algorithm, including the following steps:
[0068] S1-3-1: Divide a number of pre-processed historical enterprise operation data into a second model training set and a second model test set in a ratio of 7:3;
[0069] S1-3-2: Extract the features of the enterprise business data according to the preset N enterprise portrait focus dimensions, such as financial status, market position, business activities, management team, industry trends, etc., use the GAN algorithm to build N initial focus feature extraction modules, input the second model training set, optimize and train the N initial focus feature extraction modules, obtain N optimized focus feature extraction modules, and obtain several historical enterprise business data focus feature groups; the historical enterprise business data focus feature groups are composed of the historical enterprise business data focus features output by the N focus feature extraction modules belonging to the same pre-processed historical enterprise business data;
[0070] The feature extraction module includes a generator and a discriminator. The generator is responsible for generating enterprise features from the latent space, and the discriminator is responsible for judging the authenticity of the features. The discriminator is added with the recognition of enterprise portrait features, so that it can not only judge the authenticity of the data, but also judge whether it conforms to the enterprise characteristics.
[0071] S1-3-3: Use the MLP algorithm to construct an initial first fusion prediction module, and input a number of historical enterprise operation data focus feature groups to optimize and train the initial first fusion prediction module to obtain an optimized first fusion prediction module;
[0072] S1-3-4: Integrate N optimized focus feature extraction modules and the optimized first fusion prediction module to obtain an optimized enterprise portrait generation model, input the second model test set, perform model testing on the optimized enterprise portrait generation model, and obtain the second model test accuracy;
[0073] S1-3-5: If the test accuracy of the second model is greater than the second accuracy threshold, the final enterprise portrait generation model is output; otherwise, optimization training continues;
[0074] S1-4: Based on the historical corporate portraits of different first-tier enterprises and the corresponding contract invoice data of several historical enterprises, the CNN-RF-MLP algorithm is used to build an asset valuation model;
[0075] The asset evaluation model is constructed based on the Convolutional Neural Networks (CNN)-Random Forest (RF)-MLP algorithm, and the asset evaluation model includes a portrait feature extraction model constructed based on the CNN algorithm, a data feature extraction model constructed based on the RF algorithm, and a second fusion prediction module constructed based on the MLP algorithm, including the following steps:
[0076] S1-4-1: Integrate the historical enterprise portraits of the same first enterprise and the corresponding contract invoice data of several historical enterprises into historical asset evaluation samples to obtain several historical asset evaluation samples;
[0077] S1-4-2: Divide a number of historical asset evaluation samples into a third model training set and a third model test set in a ratio of 7:3;
[0078] S1-4-3: Use the CNN algorithm to build an initial portrait feature extraction model, input the third model training set, optimize the initial portrait feature extraction model, obtain an optimized portrait feature extraction model, and obtain several historical enterprise portrait features of the first enterprise;
[0079] S1-4-4: Use the RF algorithm to build an initial data feature extraction model, input the third model training set, optimize the initial data feature extraction model, obtain an optimized data feature extraction model, and obtain a historical enterprise contract invoice data feature group of several first enterprises; the historical enterprise contract invoice data features of several historical enterprise contract invoice data of the same first enterprise are formed;
[0080] S1-4-5: Use the MLP algorithm to construct an initial second fusion prediction module, input a number of historical enterprise portrait features and corresponding historical enterprise contract invoice data feature groups, optimize and train the initial second fusion prediction module, and obtain an optimized second fusion prediction module;
[0081] S1-4-16: Integrate the optimized portrait feature extraction model, the optimized data feature extraction model and the optimized second fusion prediction module to obtain an optimized asset evaluation model and input the model into the third model test set, perform model testing on the optimized asset evaluation model, and obtain the third model test accuracy;
[0082] S1-4-7: If the test accuracy of the third model is greater than the third accuracy threshold, the final asset evaluation model is output; otherwise, optimization training continues;
[0083] S2: collecting the real-time business operation data and several real-time enterprise contract invoice files of the second enterprise, and preprocessing them to obtain the preprocessed real-time business operation data and several preprocessed real-time enterprise contract invoice files;
[0084] The real-time enterprise contract invoice files include real-time enterprise contract files and real-time enterprise invoice files. The preprocessing includes standardization and data dimension reduction of real-time enterprise operation data, converting unstructured data into structured data, standardizing the data to adapt it to the feature extraction process of GAN, and data cleaning, format processing and image processing of historical enterprise contract invoice files.
[0085] S3: Based on the pre-processed real-time enterprise operation data, using the enterprise portrait generation model, generate an enterprise portrait to obtain a real-time enterprise portrait of the second enterprise, including the following steps:
[0086] S3-1: input the preprocessed real-time enterprise operation data into the enterprise portrait generation model, and use the N focus feature extraction modules of the enterprise portrait generation model to extract the real-time enterprise operation data focus feature group of the preprocessed real-time enterprise operation data; the real-time enterprise operation data focus feature group is composed of the real-time enterprise operation data focus features output by the N focus feature extraction modules belonging to the same preprocessed real-time enterprise operation data;
[0087] S3-2: using the first fusion prediction module, performing feature fusion on N real-time enterprise management data focus features in the real-time enterprise management data focus feature group to obtain real-time enterprise management data focus fusion features;
[0088] S3-3: Using the first fusion prediction module, focusing on fusion features according to real-time enterprise operation data, generating an enterprise portrait, and obtaining a real-time enterprise portrait of the second enterprise;
[0089] S4: Based on a number of pre-processed real-time enterprise contract invoice files, a text recognition model is used to perform text recognition to obtain a number of real-time enterprise contract invoice data, including the following steps:
[0090] S4-1: Input the pre-processed real-time enterprise contract invoice file into the text recognition model, use the text positioning module to perform text positioning, and obtain a number of real-time text positioning areas and corresponding real-time text positioning positions;
[0091] S4-2: According to the real-time text positioning position, a plurality of real-time text positioning areas of the pre-processed real-time enterprise contract invoice file are cropped to obtain a plurality of real-time text positioning area images;
[0092] S4-3: Using a text recognition module, perform text recognition based on a number of real-time text positioning area images, and perform text reorganization based on the corresponding real-time text positioning positions to obtain corresponding real-time enterprise contract invoice data;
[0093] S4-4: traverse all pre-processed real-time enterprise contract invoice files to obtain a number of real-time enterprise contract invoice data;
[0094] S5: According to the enterprise contract invoice database preset by the second enterprise, data verification is performed on a plurality of real-time enterprise contract invoice data to obtain a plurality of verified real-time enterprise contract invoice data, including the following steps:
[0095] S5-1: extracting the real-time enterprise contract number or the real-time enterprise invoice number from the real-time enterprise contract invoice data;
[0096] S5-2: searching the real-time enterprise contract number and the real-time enterprise invoice number according to the enterprise contract invoice database preset by the second enterprise;
[0097] S5-3: If the search and match is successful, the current real-time enterprise contract invoice data is authentic and the process proceeds to the next step. Otherwise, the current real-time enterprise contract invoice data is wrong or forged, and the corresponding real-time enterprise contract invoice data is discarded or the text recognition is performed again.
[0098] S5-4: If the real-time enterprise contract number of the first real-time enterprise contract invoice data matches the real-time enterprise invoice number of the second real-time enterprise contract invoice data, and the contract amount in the first real-time enterprise contract invoice data is close to the invoice amount in the second real-time enterprise contract invoice data, then the corresponding first real-time enterprise contract invoice data and the second real-time enterprise contract invoice data are regarded as high-quality real-time enterprise contract invoice data; otherwise, the corresponding first real-time enterprise contract invoice data and the second real-time enterprise contract invoice data are regarded as low-quality real-time enterprise contract invoice data;
[0099] S5-5: Set corresponding evaluation weights for high-quality real-time enterprise contract invoice data and low-quality real-time enterprise contract invoice data respectively;
[0100] S5-6: traverse all real-time enterprise contract invoice data to obtain some verified real-time enterprise contract invoice data;
[0101] S6: Based on the real-time enterprise portrait and some verified real-time enterprise contract invoice data, use the asset valuation model to conduct asset valuation and obtain real-time asset valuation results, including the following steps:
[0102] S6-1: input the real-time enterprise portrait and a number of verified real-time enterprise contract invoice data into the asset valuation model, and use the portrait feature extraction model of the asset valuation model to extract the real-time enterprise portrait features of the real-time enterprise portrait;
[0103] S6-2: performing matrix conversion on a plurality of verified real-time enterprise contract invoice data to obtain a corresponding real-time enterprise contract invoice data matrix, and using a data feature extraction model to extract a real-time enterprise contract invoice data feature group of the real-time enterprise contract invoice data matrix, including the following steps:
[0104] S6-2-1: taking the same index data from different verified real-time enterprise contract invoice data in the real-time enterprise contract invoice data matrix as feature input, and obtaining several feature inputs of the real-time enterprise contract invoice data matrix;
[0105] S6-2-2: Use the RF structure trained in the data feature extraction model to obtain the Gini coefficient of the feature corresponding to each feature input quantity of the real-time enterprise contract invoice data matrix, and use the Gini coefficient as the corresponding feature contribution;
[0106] The formula is:
[0107]
[0108] In the formula, is the feature contribution of the jth feature; is the feature contribution of the jth feature in the i-th decision tree in the random forest; i is the decision tree indicator; j is the feature indicator; n is the total number of decision trees; the decision tree is the Classification and Regression Trees (CART) in the random forest;
[0109]
[0110] In the formula, is the change in the Gini index before and after the branching of the mth decision tree node; m is the indicator of the decision tree node; M is the total number of decision tree nodes;
[0111]
[0112]
[0113] In the formula, GI m GI l GI r is the Gini index of the mth, lth and rth decision tree nodes; p mk is the proportion of the kth category in the mth decision tree node; k is the category indicator; |K| is the total number of categories;
[0114] S6-2-3: normalize the feature contribution of each feature to obtain the normalized feature contribution, and obtain the feature selection standard value of each feature of the real-time enterprise contract invoice data matrix according to the normalized feature contribution;
[0115] The formula is:
[0116]
[0117] In the formula, VIM' j is the normalized feature contribution of the jth feature; J is the total number of features; is the feature contribution of the jth feature;
[0118]
[0119] In the formula, CFC j The feature selection standard value for the jth feature; VIM j* is the normalized feature contribution of the j*th feature; j* is the feature indicator;
[0120] S6-2-4: Screen all features of the real-time enterprise contract invoice data matrix according to the feature selection standard value to obtain a real-time enterprise contract invoice data feature group consisting of the first N' real-time enterprise contract invoice data matrix key features, where N' is a preset key feature threshold;
[0121] S6-3: Fusing the real-time enterprise portrait feature and the real-time enterprise contract invoice data feature group to obtain the real-time enterprise portrait contract invoice fusion feature;
[0122] S6-4: Perform asset evaluation based on the real-time enterprise portrait contract invoice fusion features to obtain real-time asset evaluation results;
[0123] S7: Generate a real-time asset appraisal report for the second enterprise based on the real-time asset appraisal result using a preset asset appraisal report template.
[0124] Embodiment 2:
[0125] like Figure 2As shown, this embodiment provides an asset evaluation system driven by enterprise contract invoices, which is used to implement an asset evaluation method. The system includes a model building unit, a preprocessing unit, an enterprise portrait generation unit, a text recognition unit, a data verification unit, an asset evaluation unit, and a report generation unit connected in sequence;
[0126] A model building unit, for building a text recognition model, a corporate portrait generation model, and an asset evaluation model using an artificial intelligence algorithm based on historical corporate operating data of different first companies and a number of historical corporate contract invoice documents;
[0127] A preprocessing unit is used to collect the real-time enterprise operation data and a plurality of real-time enterprise contract invoice files of the second enterprise, perform preprocessing, and obtain the preprocessed real-time enterprise operation data and a plurality of preprocessed real-time enterprise contract invoice files;
[0128] An enterprise portrait generation unit, configured to generate an enterprise portrait based on the pre-processed real-time enterprise operation data using an enterprise portrait generation model to obtain a real-time enterprise portrait of the second enterprise;
[0129] A text recognition unit, used to perform text recognition based on a number of pre-processed real-time enterprise contract invoice files using a text recognition model to obtain a number of real-time enterprise contract invoice data;
[0130] A data checking unit is used to check a number of real-time enterprise contract invoice data according to an enterprise contract invoice database preset by the second enterprise, and obtain a number of checked real-time enterprise contract invoice data;
[0131] The asset evaluation unit is used to perform asset evaluation based on the real-time enterprise portrait and a number of verified real-time enterprise contract invoice data, using the asset evaluation model to obtain real-time asset evaluation results;
[0132] The report generating unit is used to generate a real-time asset evaluation report of the second enterprise according to the real-time asset evaluation result and using a preset asset evaluation report template.
[0133] The present invention discloses an asset evaluation method and system driven by enterprise contract invoices. By performing image conversion and text recognition, the enterprise contract invoice documents are digitally converted. As important supporting data for asset evaluation, the application effect and data value of the enterprise contract invoice are enhanced, while the intelligence level, evaluation efficiency and practicality of asset evaluation are improved, the manpower cost investment is reduced, and the workload is alleviated. Through the enterprise portrait generation model, asset evaluation is performed in combination with the enterprise portrait, which improves the authenticity and comprehensiveness. Through the asset evaluation model constructed by the deep learning algorithm, the deep relationship between the enterprise portrait, the enterprise contract invoice data and the enterprise assets is excavated, which improves the accuracy of the evaluation. In addition, through the generated asset evaluation report, the asset evaluation results of the enterprise are intuitively displayed, which improves the convenience.
[0134] The present invention is not limited to the above optional implementations, and anyone can derive other various forms of products under the enlightenment of the present invention. The above specific implementations should not be understood as limiting the scope of protection of the present invention. The scope of protection of the present invention should be based on the definition in the claims, and the description can be used to interpret the claims.
Claims
1. An asset valuation method based on enterprise contract invoice drive, characterized by: The steps include: Based on the historical business operation data of different first-tier enterprises and several historical enterprise contract and invoice documents, artificial intelligence algorithms are used to build text recognition models, enterprise portrait generation models, and asset evaluation models; Collecting the real-time business operation data and several real-time enterprise contract invoice files of the second enterprise, and preprocessing them to obtain the preprocessed real-time business operation data and several preprocessed real-time enterprise contract invoice files; According to the pre-processed real-time enterprise operation data, the enterprise portrait generation model is used to generate the enterprise portrait, thereby obtaining the real-time enterprise portrait of the second enterprise; Based on a number of pre-processed real-time enterprise contract invoice files, a text recognition model is used to perform text recognition to obtain a number of real-time enterprise contract invoice data; According to the enterprise contract invoice database preset by the second enterprise, data verification is performed on a number of real-time enterprise contract invoice data to obtain a number of verified real-time enterprise contract invoice data; Based on the real-time enterprise portrait and some verified real-time enterprise contract invoice data, use the asset valuation model to conduct asset valuation and obtain real-time asset valuation results; According to the real-time asset evaluation results, a real-time asset evaluation report of the second enterprise is generated using a preset asset evaluation report template.
2. The asset valuation method based on enterprise contract invoice drive according to claim 1 is characterized by: Based on the historical business operation data of different first-class enterprises and several historical enterprise contract invoice documents, artificial intelligence algorithms are used to build text recognition models, enterprise portrait generation models and asset evaluation models, including the following steps: Collecting historical business operation data and several historical enterprise contract invoice files of different first enterprises, and preprocessing them to obtain several preprocessed historical business operation data and several preprocessed historical enterprise contract invoice files; Based on several pre-processed historical enterprise contract invoice documents, an image text recognition algorithm is used to build a text recognition model and generate several historical enterprise contract invoice data; Based on some pre-processed historical business data, a deep learning algorithm is used to build a business portrait generation model and generate some historical business portraits; Based on the historical corporate portraits of different first-tier companies and the corresponding historical corporate contract invoice data, an asset valuation model is constructed using a deep learning algorithm.
3. The asset valuation method based on enterprise contract invoice drive according to claim 2 is characterized by: The preprocessing includes standardization processing, data dimension reduction and enterprise portrait label adding of historical enterprise operation data, and data cleaning, format processing, image processing and label adding of historical enterprise contract invoice documents.
4. The asset valuation method based on enterprise contract invoice drive according to claim 2 is characterized by: The text recognition model is constructed based on the CTPN-CRNN algorithm.
5. The asset valuation method based on enterprise contract invoice drive according to claim 2 is characterized by: The enterprise portrait generation model is constructed based on the N-GAN-MLP algorithm, where N is the preset enterprise portrait focus dimension.
6. The asset valuation method based on enterprise contract invoice drive according to claim 2 is characterized by: The asset valuation model is constructed based on the CNN-RF-MLP algorithm.
7. The asset valuation method based on enterprise contract invoice drive according to claim 5 is characterized by: According to the pre-processed real-time enterprise operation data, the enterprise portrait generation model is used to generate the enterprise portrait to obtain the real-time enterprise portrait of the second enterprise, including the following steps: Inputting the preprocessed real-time enterprise operation data into the enterprise portrait generation model, and using the enterprise portrait generation model to extract the real-time enterprise operation data focus feature group of the preprocessed real-time enterprise operation data; Perform feature fusion on N real-time enterprise management data focus features in the real-time enterprise management data focus feature group to obtain the real-time enterprise management data focus fusion feature; Based on the real-time enterprise operation data, the enterprise portrait is generated to obtain the real-time enterprise portrait of the second enterprise.
8. The asset valuation method based on enterprise contract invoice drive according to claim 4 is characterized by: According to a number of pre-processed real-time enterprise contract invoice files, a text recognition model is used to perform text recognition to obtain a number of real-time enterprise contract invoice data, including the following steps: Input the pre-processed real-time enterprise contract invoice file into the text recognition model to locate the text, and obtain a number of real-time text location areas and corresponding real-time text location positions; According to the real-time text positioning position, a plurality of real-time text positioning areas of the pre-processed real-time enterprise contract invoice file are cropped to obtain a plurality of real-time text positioning area images; According to a number of real-time text positioning area images, text recognition is performed, and according to the corresponding real-time text positioning positions, text reorganization is performed to obtain the corresponding real-time enterprise contract invoice data; All pre-processed real-time enterprise contract invoice files are traversed to obtain a number of real-time enterprise contract invoice data.
9. The asset valuation method based on enterprise contract invoice drive according to claim 6 is characterized by: Based on the real-time enterprise portrait and some verified real-time enterprise contract invoice data, use the asset valuation model to conduct asset valuation and obtain real-time asset valuation results, including the following steps: Input the real-time enterprise portrait and a number of verified real-time enterprise contract invoice data into the asset valuation model, and use the asset valuation model to extract the real-time enterprise portrait features of the real-time enterprise portrait; Performing matrix conversion on a plurality of verified real-time enterprise contract invoice data to obtain a corresponding real-time enterprise contract invoice data matrix, and extracting a real-time enterprise contract invoice data feature group of the real-time enterprise contract invoice data matrix; Perform feature fusion on the real-time enterprise portrait feature and the real-time enterprise contract invoice data feature group to obtain the real-time enterprise portrait contract invoice fusion feature; Based on the fusion features of real-time enterprise portrait contract and invoice, asset appraisal is carried out to obtain real-time asset appraisal results.
10. An asset evaluation system driven by enterprise contract invoices, used to implement the asset evaluation method according to any one of claims 1 to 9, characterized in that: The system comprises a model building unit, a preprocessing unit, an enterprise portrait generating unit, a text recognition unit, a data checking unit, an asset evaluation unit and a report generating unit which are connected in sequence.
Citation Information
Patent Citations
Asset assessment method and device based on image recognition and computer storage medium
CN110991441A
Enterprise data processing method and device, computer equipment and storage medium
CN111340365A
Business dealer value information evaluation method and system, storage medium and computer equipment
CN111985937A
Intelligent tax management system and method based on big data
CN113901779A
Enterprise portrait label intelligent generation method
CN115952259A
