An investment project duplicate checking method, system, computer and storage medium

Abstract

The application provides an investment project duplicate checking method and system, a computer and a storage medium, which comprises the following steps: collecting declaration data of an investment project to generate a standardized data set; performing semantic analysis on the standardized data set to extract a semantic feature vector, construct a knowledge graph related to the declaration data and generate an associated feature vector, and combine to obtain a fusion feature vector; calculating a semantic deviation compensation amount based on a compensation model according to the mean value of the fusion feature vector and a historical feature vector; calculating a target feature vector, and calculating the similarity between the target feature vector and a historical project feature vector to screen duplicate declaration data with a similarity greater than a preset threshold. Through multi-modal feature extraction, the integration of semantics and knowledge graph association is achieved, the semantic understanding deficiency of text matching is overcome, the feature vector is adjusted by dynamic semantic compensation, and finally efficient decision-making is realized through the similarity matching step, thereby improving the duplicate checking accuracy and efficiency as a whole.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method, system, computer, and storage medium for detecting duplicates in investment projects. Background Technology

[0002] In the management of fixed asset investment projects, the duplicate checking process during the review stage is crucial to ensuring the rationality and effectiveness of investment projects. Duplication of investment and waste of resources not only lead to the ineffective use of funds but may also trigger vicious competition between projects, hindering healthy economic development. Furthermore, subcontracting and splitting project applications can lead to chaos in project management, increase the difficulty of supervision, and reduce investment efficiency.

[0003] Currently, plagiarism detection in the fixed asset investment project review stage mainly relies on manual review and simple text matching technology. Manual review is inefficient, struggles to handle the large volume of applications, and is easily influenced by human factors, leading to inaccurate results. While simple text matching technology can quickly locate some duplicate applications, it suffers from insufficient semantic understanding, easily resulting in missed or incorrect checks. It also fails to effectively identify semantically similar but differently worded applications, resulting in poor plagiarism detection performance. Summary of the Invention

[0004] In view of the shortcomings of the prior art, the purpose of this invention is to provide a method, system, computer and storage medium for checking investment projects for plagiarism, in order to solve the technical problems of low efficiency and poor effect of plagiarism checking in the prior art.

[0005] To achieve the above objectives, in a first aspect, the present invention provides: a method for detecting duplicate investment projects, comprising the following steps:

[0006] Collect application data for investment projects, clean and standardize them, and generate a standardized dataset.

[0007] A pre-trained deep learning model is used to perform semantic analysis on a standardized dataset to extract semantic feature vectors, construct a knowledge graph related to the declaration data and generate associated feature vectors, and calculate and fuse feature vectors based on the semantic feature vectors and the associated feature vectors through dynamic weights.

[0008] Based on the mean of the fused feature vector and the historical feature vector, the semantic deviation compensation amount is calculated based on the compensation model;

[0009] The target feature vector is calculated based on the fused feature vector and the deviation compensation amount. The similarity between the target feature vector and the historical project feature vector is calculated using the cosine similarity function to filter out duplicate application data with similarity greater than a preset threshold.

[0010] According to one aspect of the above technical solution, the steps of constructing a knowledge graph related to the declaration data and generating associated feature vectors specifically include:

[0011] A knowledge graph is constructed based on the extracted entities, defining the types of relationships between entities, including holding relationships, geographical proximity relationships, and project type similarity relationships, and stored using a graph database;

[0012] The knowledge graph is queried to retrieve related information based on the investment entity and construction location fields of the current application project, including directly related projects and indirectly related projects, and a relationship strength threshold is set to filter low-relevance data;

[0013] The graph embedding technique is used to transform the association information into numerical feature vectors, including applying the Node2Vec algorithm to vectorize entities and synthesizing the final association feature vectors through weighted averaging.

[0014] According to one aspect of the above technical solution, the calculation expression of the compensation model is:

[0015] ;

[0016] ;

[0017] In the formula, This is the deviation compensation amount. To fuse feature vectors, The mean of the historical feature vectors. For semantic compensation coefficients, This is the scale compensation coefficient. This represents the deviation in investment scale. The current investment scale of the projects being submitted. This represents the average investment size of historical projects. This is the largest normalization factor.

[0018] According to one aspect of the above technical solution, the calculation expression for the fused feature vector is:

[0019] ;

[0020] In the formula, For semantic feature vectors, These are the semantic fusion weight coefficients. For associated feature vectors, This refers to the correlation and fusion weighting coefficient.

[0021] According to one aspect of the above technical solution, the update mechanism for the semantic fusion weight coefficients is as follows:

[0022] Calculate the complexity parameter of the declared data, and then calculate the weight coefficient based on the complexity parameter using the Sigmoid function:

[0023] ;

[0024] ;

[0025] In the formula, Let L be the complexity parameter, and L be the text length. Where F is the historical maximum text length, and F is the number of key fields in the declared data. This represents the maximum number of fields in history.

[0026] According to one aspect of the above technical solution, the data cleaning of the declared data includes the following steps:

[0027] Use regular expressions to match key fields in the declaration data, including project name, investment scale, and declaration time, to extract structured information;

[0028] The extracted investment amounts are standardized in terms of unit and outliers are removed.

[0029] According to one aspect of the above technical solution, the calculation expression for the cosine similarity function is:

[0030] ;

[0031] In the formula, The similarity value. The target feature vector, For historical project feature vectors, This is the vector modulus operator.

[0032] Secondly, this invention provides an investment project plagiarism detection system, comprising:

[0033] The data processing module is used to collect application data for investment projects, clean and standardize it, and generate standardized datasets.

[0034] The fusion module is used to perform semantic analysis on the standardized dataset using a pre-trained deep learning model to extract semantic feature vectors, construct a knowledge graph related to the declaration data and generate associated feature vectors, and calculate the fused feature vector based on the semantic feature vectors and the associated feature vectors through dynamic weights.

[0035] The compensation module is used to calculate the semantic deviation compensation amount based on the mean of the fused feature vector and the historical feature vector, according to the compensation model.

[0036] The filtering module is used to calculate the target feature vector based on the fused feature vector and the deviation compensation amount, and to calculate the similarity between the target feature vector and the historical project feature vector through the cosine similarity function, so as to filter duplicate application data with similarity greater than a preset threshold.

[0037] Compared with existing technologies, the beneficial effects of this invention are as follows: by automatically cleaning and standardizing data through data preprocessing and verification steps, the subjectivity and inefficiency of manual review are eliminated; by integrating semantics and knowledge graph associations through multimodal feature extraction, the semantic understanding of text matching is overcome; by adjusting feature vectors through dynamic semantic compensation; and finally, by achieving efficient decision-making through similarity matching steps, the overall accuracy and efficiency of plagiarism detection are improved, effectively preventing subcontracting and splitting of applications and reducing the regulatory burden. Attached Figure Description

[0038] Figure 1 This is a flowchart illustrating the investment project duplication detection method in the first embodiment of the present invention;

[0039] Figure 2 This is a structural block diagram of the investment project duplication detection system in the second embodiment of the present invention;

[0040] Figure 3 This is a schematic diagram of the hardware structure of the computer in the third embodiment of the present invention;

[0041] The following detailed description, in conjunction with the accompanying drawings, will further illustrate the present invention. Detailed Implementation

[0042] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Several embodiments of the invention are illustrated in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0043] It should be noted that when a component is said to be "fixed to" another component, it can be directly on the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0044] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0045] Example 1

[0046] Please see Figure 1 The figure shows a flowchart of the investment project deduplication method in the first embodiment of the present invention. As shown in the figure, the method includes the following steps:

[0047] Step S100: Collect the application data of investment projects, clean and standardize them to generate a standardized dataset.

[0048] The application data includes key fields such as project name, investment entity, construction location, investment scale, and project content. Data cleaning aims to remove noise, errors, and inconsistencies from the application data to improve data quality. Standardization processing converts the cleaned data into a unified format to ensure consistency and comparability. This includes formatting text fields such as project name and construction location, such as using standardized naming conventions (capitalizing the first letter of the project name and removing extra spaces), and converting locations to standard latitude and longitude coordinates. The document's example mentions "standardizing the naming conventions for project names," and also includes unit conversion for numerical fields such as investment scale, such as converting all to "ten thousand yuan" units, and standardizing the format of date fields.

[0049] Preferably, the data cleaning of the declared data includes the following steps:

[0050] Use regular expressions to match key fields in the declaration data, including project name, investment scale, and declaration time, to extract structured information;

[0051] The extracted investment amounts are standardized in terms of unit and outliers are removed.

[0052] Step S200: Use a pre-trained deep learning model to perform semantic analysis on the standardized dataset to extract semantic feature vectors, construct a knowledge graph related to the declaration data and generate associated feature vectors, and calculate and fuse feature vectors based on the semantic feature vectors and the associated feature vectors through dynamic weights.

[0053] Specifically, in this embodiment, the pre-trained deep learning model is preferably BERT, which is used to perform deep processing on the standardized dataset and extract semantic feature vectors based on the output layer of the model to quantify the semantic information of the text; the knowledge graph construction aims to capture entity relationships in the declaration data and generate associated feature vectors to supplement semantic analysis.

[0054] Preferably, in this embodiment, the step of constructing a knowledge graph related to the declaration data and generating associated feature vectors specifically includes:

[0055] A knowledge graph is constructed based on the extracted entities, defining the types of relationships between entities, including controlling stake relationships, geographical proximity relationships, and project type similarity relationships, and stored using a graph database. Specifically, a knowledge graph is constructed using a graph database (such as Neo4j) based on entities in a standardized dataset (such as investment entities and construction locations). Relationships between entities include controlling stake relationships, geographical proximity (such as distance less than 5 kilometers), or project type similarity.

[0056] The knowledge graph is queried to retrieve related information based on the investment entity and construction location fields of the current application project, including directly and indirectly related projects. A relationship strength threshold is set to filter out low-relevance data. Specifically, for the current application project, the knowledge graph is queried to retrieve related entities, such as historical projects of the same investment entity. The knowledge graph query uses the Cypher graph query language and filters are set based on time range or relationship strength.

[0057] Graph embedding techniques are used to transform association information into numerical feature vectors. This includes applying the Node2Vec algorithm for entity vectorization and synthesizing the final association feature vector through weighted averaging. Specifically, the Node2Vec algorithm maps the queried association entities into low-dimensional vectors. These vectors are then used to synthesize the association feature vector through weighted averaging or attention mechanisms. .

[0058] Preferably, the calculation expression for the above-mentioned fused feature vector is:

[0059] ;

[0060] In the formula, For semantic feature vectors, These are the semantic fusion weight coefficients. For associated feature vectors, This refers to the correlation and fusion weighting coefficient.

[0061] The update mechanism for the semantic fusion weight coefficients is as follows:

[0062] Calculate the complexity parameter of the declared data, and then calculate the weight coefficient based on the complexity parameter using the Sigmoid function:

[0063] ;

[0064] ;

[0065] In the formula, Let L be the complexity parameter, and L be the text length. Where F is the historical maximum text length, and F is the number of key fields in the declared data. This represents the maximum number of fields in history.

[0066] Furthermore, the sum of the semantic fusion weight coefficient and the association fusion weight coefficient is 1. The weight calculation is dynamically adjusted based on the complexity of the application data. Increasing the semantic fusion weight coefficient makes the fused features more focused on semantic analysis, while increasing the association fusion weight coefficient makes the fused features more focused on association features. Automated weight calculation reduces the need for manual parameter tuning and improves system processing efficiency.

[0067] Step S300: Calculate the semantic deviation compensation amount based on the mean of the fused feature vector and the historical feature vector using the compensation model.

[0068] Specifically, in this embodiment, the calculation expression for the compensation model is:

[0069] ;

[0070] ;

[0071] In the formula, This is the deviation compensation amount. To fuse feature vectors, The mean of the historical feature vectors. For semantic compensation coefficients, This is the scale compensation coefficient. This represents the deviation in investment scale. The current investment scale of the projects being submitted. This represents the average investment size of historical projects. This is the maximum normalization factor. The core of the compensation model is based on deviation analysis of eigenvectors, dynamically calculating the compensation amount. The average level is calculated from historical data and serves as a benchmark. The semantic compensation coefficient and the scale compensation coefficient can be dynamically adjusted based on data characteristics, enabling the system to adapt to changes in different project types or application patterns. For example, for text-intensive projects, the model automatically strengthens semantic compensation; for scale-sensitive projects, it focuses on scale compensation.

[0072] Step S400: Calculate the target feature vector based on the fused feature vector and the deviation compensation amount, and calculate the similarity between the target feature vector and the historical project feature vector using the cosine similarity function to filter out duplicate application data with similarity greater than a preset threshold.

[0073] Specifically, the cosine similarity function is calculated as follows:

[0074] ;

[0075] In the formula, The similarity value. The target feature vector, For historical project feature vectors, This is a vector magnitude operator. Preferably, the preset threshold is 0.8.

[0076] In summary, the investment project duplication detection method in the above embodiments of the present invention automatically cleans and standardizes data through data preprocessing and verification steps, eliminating the subjectivity and inefficiency of manual review. Furthermore, it overcomes the insufficient semantic understanding of text matching by integrating semantics and knowledge graph associations through multimodal feature extraction. It adjusts feature vectors using dynamic semantic compensation and finally achieves efficient decision-making through similarity matching steps. As a result, the overall accuracy and efficiency of duplication detection are improved, effectively preventing subcontracting and splitting of applications and reducing the regulatory burden.

[0077] Example 2

[0078] The second embodiment of this application also provides an investment project duplication detection system, which is used to implement the embodiments and preferred embodiments described above, and will not be repeated hereafter. As used below, the terms "module," "unit," "subunit," etc., can refer to a combination of software and / or hardware that performs a predetermined function. Although the system described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0079] like Figure 2 As shown, the system includes: a data processing module 100, a fusion module 200, a compensation module 300, and a filtering module 400.

[0080] Data processing module 100 is used to collect application data of investment projects, clean and standardize it, and generate a standardized dataset.

[0081] The fusion module 200 is used to perform semantic analysis on the standardized dataset using a pre-trained deep learning model to extract semantic feature vectors, construct a knowledge graph related to the declaration data and generate associated feature vectors, and calculate the fusion feature vector based on the semantic feature vectors and the associated feature vectors through dynamic weights.

[0082] The compensation module 300 is used to calculate the semantic deviation compensation amount based on the mean of the fused feature vector and the historical feature vector, according to the compensation model.

[0083] The filtering module 400 is used to calculate the target feature vector based on the fused feature vector and the deviation compensation amount, and to calculate the similarity between the target feature vector and the historical project feature vector through the cosine similarity function, so as to filter duplicate application data with similarity greater than a preset threshold.

[0084] It should be noted that the modules can be functional modules or program modules, and can be implemented in software or hardware. For modules implemented in hardware, the modules can reside in the same processor; or the modules can be located in different processors in any combination.

[0085] Example 3

[0086] A third embodiment of this application provides a computer that may include a processor 81 and a memory 82 storing computer program commands.

[0087] Specifically, the processor 81 may include a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0088] The memory 82 may include a large-capacity storage device for data or commands. For example, and not limitingly, the memory 82 may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), flash memory, an optical disk drive, a magneto-optical disk drive, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 82 may include removable or non-removable (or fixed) media. Where appropriate, the memory 82 may be internal or external to a data processing device. In a particular embodiment, the memory 82 is non-volatile memory. In a particular embodiment, the memory 82 includes read-only memory (ROM) and random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), an electrically alterable read-only memory (EAROM), or flash memory, or a combination of two or more of these. Where appropriate, the RAM can be Static Random-Access Memory (SRAM) or Dynamic Random-Access Memory (DRAM). DRAM can be Fast Page Mode Dynamic Random-Access Memory (FPMDRAM), Extended Data Out Dynamic Random-Access Memory (EDODRAM), Synchronous Dynamic Random-Access Memory (SDRAM), etc.

[0089] The memory 82 can be used to store or cache various data files that need to be processed and / or communicated, as well as possible computer program commands executed by the processor 81.

[0090] The processor 81 reads and executes computer program commands stored in the memory 82 to implement any of the investment project duplication methods in the above embodiments.

[0091] In some embodiments, the computer may further include a communication interface 83 and a bus 80. For example, Figure 3 As shown, the processor 81, memory 82, and communication interface 83 are connected through bus 80 and complete communication with each other.

[0092] The communication interface 83 is used to enable communication between the various modules, devices, units, and / or equipment in the embodiments of this application. The communication interface 83 can also enable data communication with other components such as external devices, image / data acquisition devices, databases, external storage, and image / data processing workstations.

[0093] Bus 80 includes hardware, software, or both, that couples computer components together. Bus 80 includes, but is not limited to, at least one of the following: data bus, address bus, control bus, expansion bus, and local bus. For example, and not as a limitation, bus 80 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 80 may include one or more buses. Although specific buses are described and illustrated in the embodiments of this application, this application considers any suitable bus or interconnection.

[0094] Example 4

[0095] The fourth embodiment of this application provides a readable storage medium. This readable storage medium stores computer program commands; when executed by a processor, these computer program commands implement any of the investment project deduplication methods described in the above embodiments.

[0096] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0097] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for detecting duplicate investment projects, characterized in that, Includes the following steps: Collect application data for investment projects, clean and standardize them, and generate a standardized dataset. A pre-trained deep learning model is used to perform semantic analysis on a standardized dataset to extract semantic feature vectors, construct a knowledge graph related to the declaration data and generate associated feature vectors, and calculate and fuse feature vectors based on the semantic feature vectors and the associated feature vectors through dynamic weights. Based on the mean of the fused feature vector and the historical feature vector, the semantic deviation compensation amount is calculated based on the compensation model; The target feature vector is calculated based on the fused feature vector and the deviation compensation amount. The similarity between the target feature vector and the historical project feature vector is calculated using the cosine similarity function to filter out duplicate application data with similarity greater than a preset threshold. The calculation expression for the compensation model is as follows: ； ； In the formula, This is the deviation compensation amount. To fuse feature vectors, The mean of the historical feature vectors. For semantic compensation coefficients, This is the scale compensation coefficient. This represents the deviation in investment scale. The current investment scale of the projects being submitted. This represents the average investment size of historical projects. The largest normalization factor; The calculation expression for the fused feature vector is as follows: ； In the formula, For semantic feature vectors, These are the semantic fusion weight coefficients. For associated feature vectors, For correlation and fusion weighting coefficients; The update mechanism for the semantic fusion weight coefficients is as follows: Calculate the complexity parameter of the declared data, and then calculate the weight coefficient based on the complexity parameter using the Sigmoid function: ； ； In the formula, Let L be the complexity parameter, and L be the text length. Where F is the historical maximum text length, and F is the number of key fields in the declared data. This represents the maximum number of fields in history.

2. The investment project duplication detection method according to claim 1, characterized in that, The specific steps for constructing a knowledge graph related to the application data and generating associated feature vectors include: A knowledge graph is constructed based on the extracted entities, defining the types of relationships between entities, including holding relationships, geographical proximity relationships, and project type similarity relationships, and stored using a graph database; The knowledge graph is queried to retrieve related information based on the investment entity and construction location fields of the current application project, including directly related projects and indirectly related projects, and a relationship strength threshold is set to filter low-relevance data; The graph embedding technique is used to transform the association information into numerical feature vectors, including applying the Node2Vec algorithm to vectorize entities and synthesizing the final association feature vectors through weighted averaging.

3. The investment project duplication detection method according to claim 1, characterized in that, The data cleaning process for the submitted data includes the following steps: Use regular expressions to match key fields in the declaration data, including project name, investment scale, and declaration time, to extract structured information; The extracted investment amounts are standardized in terms of unit and outliers are removed.

4. The investment project duplication detection method according to claim 1, characterized in that, The cosine similarity function is calculated as follows: ； In the formula, The similarity value. The target feature vector, For historical project feature vectors, This is the vector modulus operator.

5. An investment project deduplication system based on the investment project deduplication method according to any one of claims 1-4, characterized in that, include: The data processing module is used to collect application data for investment projects, clean and standardize it, and generate standardized datasets. The fusion module is used to perform semantic analysis on the standardized dataset using a pre-trained deep learning model to extract semantic feature vectors, construct a knowledge graph related to the declaration data and generate associated feature vectors, and calculate the fused feature vector based on the semantic feature vectors and the associated feature vectors through dynamic weights. The compensation module is used to calculate the semantic deviation compensation amount based on the mean of the fused feature vector and the historical feature vector, according to the compensation model. The filtering module is used to calculate the target feature vector based on the fused feature vector and the deviation compensation amount, and to calculate the similarity between the target feature vector and the historical project feature vector through the cosine similarity function, so as to filter duplicate application data with similarity greater than a preset threshold.

6. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the investment project deduplication method as described in any one of claims 1-4.

7. A storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the investment project deduplication method as described in any one of claims 1-4.

Images