String tag behavior identification method, device, equipment, storage medium and product

Abstract

The application discloses a serial label behavior identification method and device, equipment, storage medium and product, relates to the technical field of artificial intelligence, and the serial label behavior identification method comprises: obtaining current bidding information and historical bidding information of a bidder; performing anti-serial label behavior identification on the current bidding information, the historical bidding information and the network behavior of the bidder based on a large language model to obtain an anti-serial label behavior identification result; and performing serial label behavior early warning according to the anti-serial label behavior identification result. Since the application performs anti-serial label behavior identification on the current bidding information, the historical bidding information and the network behavior of the bidder based on a large language model, and performs serial label behavior early warning according to the anti-serial label behavior identification result, compared with the existing way of manually checking serial label behavior in the bidding process, the above-mentioned way of the application can intelligently and comprehensively and accurately identify serial label behavior.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to methods, apparatus, devices, storage media and products for identifying tag-scratching behavior. Background Technology

[0002] In the bidding and tendering field, with the widespread application of electronic bidding and tendering systems, bid rigging and collusion are commonplace, especially in bidding projects with high bid amounts and broad scopes. These violations seriously disrupt the procurement order, not only undermining the principle of fair competition but also harming the interests of enterprises and increasing the risks of supplier selection. In the long run, this will lead to the unhealthy development of the bidding and tendering market and may even result in asset loss and waste of social resources. Traditional anti-collusion measures mainly rely on manual review and rule setting, lacking real-time and intelligent means, making it difficult to comprehensively and accurately identify bid rigging and collusion. Furthermore, these methods are inefficient and easily circumvented. Therefore, how to efficiently identify bid rigging has become an urgent technical problem to be solved. Summary of the Invention

[0003] The main purpose of this application is to provide a method, apparatus, device, storage medium and product for identifying tag-scraping behavior, which aims to solve the technical problem that existing tag-scraping behavior identification relies on manual review and has low identification efficiency.

[0004] To achieve the above objectives, this application proposes a method for identifying cross-trading behavior, the method comprising:

[0005] Obtain the current and historical bidding information of bidders;

[0006] Based on a large language model, anti-collusion behavior identification is performed on the current bidding information, the historical bidding information, and the network behavior of the bidders to obtain anti-collusion behavior identification results.

[0007] Based on the anti-scraping behavior identification results, a warning of scrambling behavior is issued.

[0008] Optionally, the large language model includes a general large language model and a preset bid-rigging behavior identification model. The step of identifying bid-rigging behavior based on the large language model of the current bidding information, the historical bidding information, and the bidder's network behavior to obtain the bid-rigging behavior identification result includes:

[0009] Based on the preset bid-rigging behavior identification model, the current bidding information and the bidder's network behavior are used to identify anti-bid-rigging behavior, and a first anti-bid-rigging behavior identification result is obtained.

[0010] The historical bidding information is input into the general large language model to obtain the second anti-collusion behavior identification result output by the general large language model;

[0011] The anti-scraping behavior identification result is determined based on the first anti-scraping behavior identification result and the second anti-scraping behavior identification result.

[0012] Optionally, the step of identifying anti-collusion behavior based on the preset collusion behavior identification model for the current bidding information and the bidder's network behavior to obtain a first anti-collusion behavior identification result includes:

[0013] Based on the current bidding information, determine the bid document information and the bidder's basic information;

[0014] Based on the preset bid-rigging behavior recognition model, information analysis is performed on the basic information and the bid document information to obtain information recognition results;

[0015] The bidder's network behavior is input into the preset anti-collusion behavior identification model to identify anti-collusion behavior and obtain the network behavior identification result.

[0016] The first anti-scraping behavior identification result is determined based on the network behavior identification result and the information identification result.

[0017] Optionally, the step of performing information analysis on the basic information and the tender document information based on the preset bid-rigging behavior identification model to obtain the information identification result includes:

[0018] Based on the preset bid-rigging behavior identification model, the basic information is analyzed to determine the relationship between the bidders.

[0019] The bid document information is feature extracted according to the preset bid-rigging behavior recognition model, and the bid document similarity between bid documents is determined according to the feature extraction results;

[0020] The information identification result is determined based on the aforementioned relationship and the similarity of the tender documents.

[0021] Optionally, the step of inputting the bidder's network behavior into the preset bid-rigging behavior identification model for anti-bid-rigging behavior identification and obtaining network behavior identification results includes:

[0022] Based on the preset bid-rigging behavior identification model, the network behavior features of the bidder are extracted to obtain network behavior features;

[0023] Based on the network behavior characteristics, target network behaviors that meet preset abnormal conditions are determined.

[0024] The network behavior identification result is determined based on the target network behavior.

[0025] Optionally, before the step of obtaining the bidder's current bidding information and historical bidding information, the method further includes:

[0026] Collect the bidders' facial information;

[0027] Based on the facial information, determine whether the bidder is a legitimate user;

[0028] If the bidder is not a legitimate user, a preset warning message will be sent.

[0029] Optionally, the step of determining whether the bidder is a legitimate user based on the facial information includes:

[0030] The target facial features are determined based on the facial information;

[0031] The target facial features are matched against the user information in the bidding application to obtain the matching result;

[0032] The matching results will be used to determine whether the bidder is a legitimate user.

[0033] Optionally, after the step of identifying anti-collusion behavior based on the current bidding information, the historical bidding information, and the bidder's network behavior using a large language model to obtain the anti-collusion behavior identification result, the method further includes:

[0034] When the anti-spoofing behavior identification result is that there is no spoofing behavior, a preset encryption algorithm is obtained;

[0035] The current bidding information is encrypted based on the preset encryption algorithm to obtain encrypted bidding information;

[0036] The encrypted bidding information is sent to the preset bidding terminal.

[0037] Optionally, the step of issuing a warning about cross-selling behavior based on the anti-cross-selling behavior identification result includes:

[0038] When the anti-scraping behavior identification result indicates the presence of scrambling behavior, a preset early warning prompt strategy is obtained;

[0039] Warnings are issued based on the preset warning strategy.

[0040] Optionally, after the step of identifying anti-collusion behavior based on the current bidding information, the historical bidding information, and the bidder's network behavior using a large language model to obtain the anti-collusion behavior identification result, the method further includes:

[0041] Sample data is determined based on the anti-collusion behavior identification results, the current bidding information, the historical bidding information, and the bidder's network behavior;

[0042] The large language model is iteratively optimized based on the sample data to obtain a new large language model, and the new large language model is used to replace the old large language model.

[0043] Optionally, after the step of issuing a warning about cross-selling behavior based on the anti-cross-selling behavior identification result, the method further includes:

[0044] Receive feedback information from the bidder;

[0045] The large language model is optimized based on the feedback information.

[0046] Furthermore, to achieve the above objectives, this application also proposes a smuggling behavior identification device, which includes:

[0047] The acquisition module is used to acquire the bidder's current bidding information and historical bidding information;

[0048] The behavior recognition module is used to identify anti-collusion behavior based on the current bidding information, the historical bidding information, and the network behavior of the bidders, and to obtain the anti-collusion behavior recognition result.

[0049] The early warning module is used to issue early warnings about cross-selling behavior based on the anti-cross-selling behavior identification results.

[0050] Optionally, the large language model includes a general large language model and a preset anti-scraping behavior identification model. The behavior identification module is further used to identify anti-scraping behavior based on the preset anti-scraping behavior identification model for the current bidding information and the bidder's network behavior, and obtain a first anti-scraping behavior identification result.

[0051] The historical bidding information is input into the general large language model to obtain the second anti-collusion behavior identification result output by the general large language model;

[0052] The anti-scraping behavior identification result is determined based on the first anti-scraping behavior identification result and the second anti-scraping behavior identification result.

[0053] Optionally, the behavior recognition module is further configured to determine the bid document information and the bidder's basic information based on the current bidding information;

[0054] Based on the preset bid-rigging behavior recognition model, information analysis is performed on the basic information and the bid document information to obtain information recognition results;

[0055] The bidder's network behavior is input into the preset anti-collusion behavior identification model to identify anti-collusion behavior and obtain the network behavior identification result.

[0056] The first anti-scraping behavior identification result is determined based on the network behavior identification result and the information identification result.

[0057] Optionally, the behavior recognition module is further configured to perform bidder association information analysis on the basic information based on the preset bid-rigging behavior recognition model, and determine the association relationship between the bidders;

[0058] The bid document information is feature extracted according to the preset bid-rigging behavior recognition model, and the bid document similarity between bid documents is determined according to the feature extraction results;

[0059] The information identification result is determined based on the aforementioned relationship and the similarity of the tender documents.

[0060] Optionally, the behavior recognition module is further configured to extract features of the bidder's network behavior based on the preset bid-rigging behavior recognition model to obtain network behavior features;

[0061] Based on the network behavior characteristics, target network behaviors that meet preset abnormal conditions are determined.

[0062] The network behavior identification result is determined based on the target network behavior.

[0063] Optionally, the acquisition module is also used to collect the bidder's facial information;

[0064] Based on the facial information, determine whether the bidder is a legitimate user;

[0065] If the bidder is not a legitimate user, a preset warning message will be sent.

[0066] Optionally, the acquisition module is further configured to determine the target facial features based on the facial information;

[0067] The target facial features are matched against the user information in the bidding application to obtain the matching result;

[0068] The matching results will be used to determine whether the bidder is a legitimate user.

[0069] Optionally, the behavior recognition module is further configured to obtain a preset encryption algorithm when the anti-scraping behavior recognition result is that there is no scrambling behavior;

[0070] The current bidding information is encrypted based on the preset encryption algorithm to obtain encrypted bidding information;

[0071] The encrypted bidding information is sent to the preset bidding terminal.

[0072] Optionally, the early warning module is further configured to obtain a preset early warning prompt strategy when the anti-scraping behavior identification result indicates the existence of scrambling behavior;

[0073] Warnings are issued based on the preset warning strategy.

[0074] Optionally, the behavior recognition module is further configured to determine sample data based on the anti-collusion behavior recognition result, the current bidding information, the historical bidding information, and the bidder's network behavior;

[0075] The large language model is iteratively optimized based on the sample data to obtain a new large language model, and the new large language model is used to replace the old large language model.

[0076] Optionally, the early warning module is also used to receive feedback information from the bidder;

[0077] The large language model is optimized based on the feedback information.

[0078] In addition, to achieve the above objectives, this application also proposes a tag-scratching behavior identification device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the tag-scratching behavior identification method as described above.

[0079] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the tag-scratching behavior identification method described above.

[0080] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the tag-scratching behavior identification method described above.

[0081] This application obtains the current and historical bidding information of bidders; it uses a large language model to identify anti-collusion behavior based on the current bidding information, the historical bidding information, and the bidders' online behavior, obtaining anti-collusion behavior identification results; and it issues a collusion behavior warning based on the anti-collusion behavior identification results. Because this application uses a large language model to identify anti-collusion behavior based on the current and historical bidding information and the bidders' online behavior, and issues a collusion behavior warning based on the anti-collusion behavior identification results, compared to existing methods of manually reviewing bids for collusion, the above method of this application can comprehensively and accurately identify collusion behavior using intelligent means. Attached Figure Description

[0082] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0083] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0084] Figure 1 This is a flowchart illustrating an embodiment of the tag-scratching behavior identification method of this application.

[0085] Figure 2 This is a flowchart illustrating Embodiment 2 of the tag-scraping behavior identification method of this application.

[0086] Figure 3 This is a flowchart illustrating Embodiment 3 of the tag-scraping behavior identification method of this application.

[0087] Figure 4 This is a schematic diagram of the AI ​​browser module provided in Embodiment 3 of the tag-scraping behavior identification method of this application;

[0088] Figure 5 This is a schematic diagram of the module structure of the tag-scratching behavior identification device according to an embodiment of this application;

[0089] Figure 6 This is a schematic diagram of the device structure of the hardware operating environment involved in the tag-scratching behavior identification method in the embodiments of this application.

[0090] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0091] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0092] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0093] The main solution of this application embodiment is as follows: Obtain the bidder's current bidding information and historical bidding information; perform anti-collusion behavior identification on the current bidding information, historical bidding information, and the bidder's network behavior based on a large language model, and obtain the anti-collusion behavior identification result; and issue a collusion behavior warning based on the anti-collusion behavior identification result. Since this application uses a large language model to identify anti-collusion behavior based on the current bidding information, historical bidding information, and the bidder's network behavior, and issues a collusion behavior warning based on the anti-collusion behavior identification result, compared to existing methods of manually reviewing bid-rigging processes, the above method of this application can comprehensively and accurately identify collusion behavior using intelligent means.

[0094] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device or AI browser capable of performing the above functions. The following description uses the AI ​​browser as an example to illustrate this embodiment and the subsequent embodiments.

[0095] Based on this, embodiments of this application provide a method for identifying tag-scratching behavior, referring to... Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of the tag-scraping behavior identification method of this application.

[0096] In this embodiment, the method for identifying trolling behavior includes the following steps:

[0097] Step S10: Obtain the bidder's current bidding information and historical bidding information;

[0098] It should be noted that the bidder can be a legal person, other organization, or individual that responds to a tender and participates in the competition with the aim of winning the bid. The bidder's current bidding information may include the bidder's basic information and bid documents. The basic information may include the bidder's name, contact information, company name, registered address, and legal representative. The bid documents may be responsive documents prepared by the bidder in accordance with the requirements of the tender documents, and may consist of commercial documents, technical documents, price quotations, and other parts. The historical bidding information may include the bidder's historical bid documents in previous bidding processes, performance information in bidding projects, and information on whether there has been any abnormal behavior such as bid rigging.

[0099] Step S20: Based on the large language model, identify anti-collusion behavior of the current bidding information, the historical bidding information, and the network behavior of the bidder to obtain the anti-collusion behavior identification result;

[0100] It should be noted that the Large Language Model (LLM) is a deep learning model trained using a large amount of text data. It can generate natural language text or understand the meaning of language text. The LLM can handle various natural language tasks, such as text classification, question answering, and dialogue. The anti-collusion behavior identification based on the LLM for the current bidding information, historical bidding information, and the bidder's online behavior can be achieved by using the LLM to analyze user behavior data (including the current bidding information, historical bidding information, and the bidder's online behavior) in real time to identify potential collusion behaviors. Multimodal fusion analysis, combining text, images, and online behavior data, improves the accuracy of the analysis and detection. The bidder's online behavior may include information collected by the AI ​​browser, such as the CPU serial number, memory serial number, hard disk serial number, MAC address, IP address, and operating system version of the terminal used by the bidder.

[0101] In practice, when applying for bidding, bidders must first register and log in to the bidding platform. Registration requires uploading basic information, including facial recognition data. Subsequent bidding processes, such as uploading bid documents, must be performed by the registered bidder. After successful registration and login, the bidding platform will instruct the bidder to install a dedicated AI browser. Bidders can only participate in the bidding process by accessing the bidding system through this installed AI browser. The AI ​​browser can obtain information about the bidder's currently used terminal, including CPU serial number, memory serial number, hard drive serial number, MAC address, IP address, and operating system version.

[0102] Furthermore, in order to improve the prediction accuracy of the large language model, after step S20, the method further includes: determining sample data based on the anti-collusion behavior identification result, the current bidding information, the historical bidding information, and the bidder's network behavior;

[0103] The large language model is iteratively optimized based on the sample data to obtain a new large language model, and the new large language model is used to replace the old large language model.

[0104] It should be noted that, in this embodiment, after predicting the anti-collusion behavior identification result through the large language model, it can be further confirmed by relevant personnel. After obtaining the confirmation result from relevant personnel, the input and output data (anti-collusion behavior identification result, current bidding information, historical bidding information, and bidders' network behavior) can be used as sample data for iterative optimization of the large language model. It can also receive suggestions from various bidders based on their operation and use of the AI ​​browser, and further optimize relevant parameters and pre-set judgment rules in the large language model based on these suggestions.

[0105] Step S30: Issue a warning about cross-selling behavior based on the anti-cross-selling behavior identification results.

[0106] It should be noted that the anti-collusion behavior identification result can be used to issue an early warning based on a preset early warning strategy when the anti-collusion behavior identification result indicates that collusion behavior may exist. The preset early warning strategy is a pre-set early warning method, which may include strategies such as automatically sending relevant warning information to relevant personnel, reminding them to pay attention to preventing collusion behavior, canceling the bidder's bidding qualification, and adding the bidder to the blacklist.

[0107] Furthermore, in order to ensure the data security of bidders and prevent data leakage, after step S20, the method further includes: when the anti-collusion behavior identification result is that there is no collusion behavior, obtaining a preset encryption algorithm;

[0108] The current bidding information is encrypted based on the preset encryption algorithm to obtain encrypted bidding information;

[0109] The encrypted bidding information is sent to the preset bidding terminal.

[0110] It should be noted that the preset encryption algorithm can be a national cryptographic algorithm, including SM1, SM2, SM3, SM4, etc. The preset bidding terminal can be a bidding system.

[0111] In practice, all data exchanges between the AI ​​browser and the bidding system are encrypted using a pre-set encryption algorithm to prevent interception and ensure the security of data transmission and storage. Furthermore, both the AI ​​browser and the bidding system undergo regular security checks and vulnerability patching to guarantee their security.

[0112] This embodiment obtains the bidder's current and historical bidding information; it uses a large language model to identify anti-collusion behavior based on the current bidding information, historical bidding information, and the bidder's online behavior, obtaining anti-collusion behavior identification results; and it issues a collusion behavior warning based on the anti-collusion behavior identification results. Because this embodiment uses a large language model to identify anti-collusion behavior based on the current bidding information, historical bidding information, and the bidder's online behavior, and issues a collusion behavior warning based on the anti-collusion behavior identification results, compared to existing methods of manually reviewing bids for collusion, this embodiment can comprehensively and accurately identify collusion behavior using intelligent means.

[0113] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 2 , Figure 2 This is a flowchart illustrating a second embodiment of the tag-scraping behavior recognition method of this application. In step S20, the large language model includes a general large language model and a preset tag-scraping behavior recognition model. Step S20 also includes the following steps:

[0114] Step S201: Based on the preset bid-rigging behavior identification model, perform anti-bid-rigging behavior identification on the current bidding information and the bidder's network behavior to obtain the first anti-bid-rigging behavior identification result;

[0115] It should be noted that the preset bid-rigging behavior identification model can be a large language model trained on sample data for identifying anti-bid-rigging behavior in the current bidding information and the bidder's network behavior. The sample data may include pre-collected information such as the current bidding information, historical bidding information, the bidder's network behavior, and indicators of whether bid-rigging behavior exists. The step of identifying anti-bid-rigging behavior based on the preset bid-rigging behavior identification model to obtain a first anti-bid-rigging behavior identification result can be achieved by inputting the current bidding information and the bidder's network behavior into the preset bid-rigging behavior identification model and obtaining the first anti-bid-rigging behavior identification result output by the preset bid-rigging behavior identification model.

[0116] Furthermore, in order to improve the efficiency of identifying bid-rigging behavior, step S201 may include: determining the bid document information and the bidder's basic information based on the current bidding information;

[0117] Based on the preset bid-rigging behavior recognition model, information analysis is performed on the basic information and the bid document information to obtain information recognition results;

[0118] The bidder's network behavior is input into the preset anti-collusion behavior identification model to identify anti-collusion behavior and obtain the network behavior identification result.

[0119] The first anti-scraping behavior identification result is determined based on the network behavior identification result and the information identification result.

[0120] It should be noted that the tender document information can be technical, pricing, and other responsive documents uploaded by the bidder during the bidding process. The basic information may include the bidder's company name, registered address, legal representative, etc. Based on the preset bid-rigging behavior identification model, information analysis is performed on the basic information and the tender document information to obtain the information identification result. This can be based on bidder information analysis and tender document content analysis using the preset bid-rigging behavior identification model. Specifically, bidder information analysis includes extracting the bidder's basic information and analyzing whether there are any related relationships. If there are related relationships among multiple bidders, it is determined that bid-rigging behavior may exist. Tender document content analysis includes extracting key content of the tender documents (such as technical solutions, pricing, commitments, etc.), analyzing similarity, and determining whether there is a risk of bid-rigging based on the similarity and a preset similarity threshold. The process of inputting the bidder's network behavior into the preset bid-rigging behavior identification model for anti-bid-rigging behavior identification can obtain network behavior identification results. This can include using the preset bid-rigging behavior identification model to monitor the upload time, IP address, and network of the tender documents, analyzing whether multiple tender documents originate from the same IP address or network within the same time period, and obtaining network behavior identification results. The determination of the first anti-collusion behavior identification result based on the network behavior identification result and the information identification result can be based on the collusion risk information identified in the network behavior identification result and the information identification result and a pre-set judgment rule. The pre-set judgment rule can be a pre-set condition under which a collusion risk is determined and an early warning is required. For example, if two bid documents are uploaded through the same IP address within a preset time interval, it is determined that collusion behavior may exist.

[0121] Furthermore, in order to accurately identify whether bidders are engaging in bid rigging, the preset bid rigging behavior identification model performs information analysis on the basic information and the bid document information to obtain the information identification result, including the following steps:

[0122] Based on the preset bid-rigging behavior identification model, the basic information is analyzed to determine the relationship between the bidders.

[0123] The bid document information is feature extracted according to the preset bid-rigging behavior recognition model, and the bid document similarity between bid documents is determined according to the feature extraction results;

[0124] The information identification result is determined based on the aforementioned relationship and the similarity of the tender documents.

[0125] It should be noted that the step of analyzing the bidder association information based on the preset bid-rigging behavior identification model to determine the relationships between bidders can be achieved by using the preset bid-rigging behavior identification model to determine whether there are relationships between each bidder and its corresponding company based on the basic information, such as whether there are business dealings, whether they are the same legal entity, whether their registered addresses are the same, or whether there is overlap in investor information. The step of extracting features from the bid document information according to the preset bid-rigging behavior identification model and determining the bid document similarity based on the feature extraction results can be achieved by extracting features from each bidder's bid documents, calculating the similarity of each bid document based on the feature extraction results, including the similarity of technical solutions, quotations, commitments, etc., and then calculating the final bid document similarity based on the pre-set weights corresponding to the technical solutions, quotations, commitments, etc. The step of determining the information identification result based on the association and the bid document similarity can be achieved by the preset bid-rigging behavior identification model predicting the information identification result based on the association and the bid document similarity. Alternatively, a pre-set bid-rigging behavior identification model can be used to output only the association relationship and the similarity of the bid documents, and the information identification result can be determined based on the association relationship and the similarity of the bid documents using pre-set judgment rules. The pre-set judgment rules may include determining that bid-rigging exists if there is an association relationship between the bidders; or determining that bid-rigging exists if the similarity of the bid documents is greater than a preset similarity threshold.

[0126] Furthermore, in order to accurately identify whether bidders are engaging in bid rigging, the step of inputting the bidders' online behavior into the preset bid rigging behavior identification model for anti-bid rigging behavior identification and obtaining the network behavior identification result includes:

[0127] Based on the preset bid-rigging behavior identification model, the network behavior features of the bidder are extracted to obtain network behavior features;

[0128] Based on the network behavior characteristics, target network behaviors that meet preset abnormal conditions are determined.

[0129] The network behavior identification result is determined based on the target network behavior.

[0130] It should be noted that the feature extraction of the bidder's network behavior can yield network behavior features such as the attributes of the uploaded documents, the path of the bid documents, the directory of the bid documents, the upload time of the bid documents, the IP address, and the network connected. The preset anomaly conditions can include identical or similar paths for bid documents, identical or similar directories for bid documents, identical or similar names for bid documents, and identical IP addresses or networks used to upload bid documents. Determining target network behaviors that meet the preset anomaly conditions based on these network behavior features can involve determining whether each bidder's network behavior meets the aforementioned preset anomaly conditions. If it does, the network behaviors that meet the preset anomaly conditions and the corresponding bidders are identified, and a network behavior identification result is generated.

[0131] Step S202: Input the historical bidding information into the general large language model to obtain the second anti-collusion behavior identification result output by the general large language model;

[0132] It should be noted that the step of inputting the historical bidding information into the general large language model to obtain the second anti-collusion behavior identification result output by the general large language model can be achieved by using the general large language model to analyze the bidding data and behavior of each bidder in the historical bidding information, to obtain the bidder's performance in different projects, to identify the bidder's abnormal behavior, and to obtain the second anti-collusion behavior identification result.

[0133] Step S203: Determine the anti-scraping behavior identification result based on the first anti-scraping behavior identification result and the second anti-scraping behavior identification result.

[0134] It should be noted that, when determining the anti-collusion behavior identification result based on the first and second anti-collusion behavior identification results, the anti-collusion behavior identification result indicates that the bidder has a risk of collusion. Conversely, if both the first and second anti-collusion behavior identification results indicate no risk of collusion, the anti-collusion behavior identification result indicates that the bidder does not have a risk of collusion.

[0135] This embodiment uses the preset bid-rigging behavior identification model to identify anti-bid-rigging behaviors in the current bidding information and the bidder's network behavior, obtaining a first anti-bid-rigging behavior identification result. The historical bidding information is then input into the general large language model to obtain a second anti-bid-rigging behavior identification result output by the general large language model. The final anti-bid-rigging behavior identification result is determined based on the first and second results. This embodiment combines the general large language model and the preset bid-rigging behavior identification model to identify bidders' bid-rigging behavior, effectively preventing and detecting such behavior. This significantly improves the accuracy of identifying bid-rigging and collusion, reduces the probability of false positives and false negatives, and maintains a fair and competitive market environment.

[0136] Based on the above embodiments of this application, in the third embodiment of this application, the same or similar content as the above embodiments can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 3 , Figure 3 This is a flowchart illustrating an embodiment of the tag-scratching behavior identification method of this application. Before step S10, the method further includes the following steps:

[0137] Step S001: Collect the bidder's facial information;

[0138] It should be noted that in this embodiment, when registering on the bidding platform, bidders need to upload their basic information, including facial information. During subsequent bidding processes, such as uploading bid documents, the registered bidder must perform the operations themselves. When the bidder performs bidding operations using the AI ​​browser, the AI ​​browser will collect the bidder's facial information in real time.

[0139] Step S002: Determine whether the bidder is a legitimate user based on the facial information;

[0140] It should be noted that the step of determining whether the bidder is a legitimate user based on the facial information can be achieved by comparing the facial information with the facial information of the bidder when registering on the bidding platform. If the comparison is successful, the bidder is determined to be a legitimate user.

[0141] Furthermore, in order to reduce the risk of collusion, step S002 may include: determining the target facial features based on the facial information;

[0142] The target facial features are matched against the user information in the bidding application to obtain the matching result;

[0143] The matching results will be used to determine whether the bidder is a legitimate user.

[0144] It should be noted that determining the target facial features based on the facial information can be achieved by using a facial feature extraction algorithm to extract facial features from the facial information. The facial feature extraction algorithm may include DeepFace, Viola-Jones, etc. The user information for applying for bidding may include the facial feature information of all bidders applying for bidding. After determining other basic information of the bidders, the facial feature information of the bidders during registration can also be directly obtained based on the bidders' basic information. The target facial features are then matched with the facial feature information of the bidders during registration to obtain a matching result. If the match is successful, the bidder is determined to be a legitimate user; if the match is unsuccessful, the bidder is determined to be an illegitimate user, the bidder's bidding operations are restricted, and a warning message is issued to the bidding platform or relevant personnel.

[0145] Step S003: When the bidder is not a legitimate user, send a preset warning message.

[0146] It should be noted that the bidder in this embodiment must be the same as the bidder registered on the bidding platform. If they are not the same, the bidder is deemed not to meet the bidding conditions, a warning message is sent to the relevant personnel, and their bidding operations are restricted.

[0147] In specific implementation, it can be referred to Figure 4 , Figure 4 This diagram illustrates the AI ​​browser module provided in Embodiment 3 of the tag-scraping behavior identification method of this application. The AI ​​browser in this embodiment includes a computer information module, a face recognition module, a data acquisition module, an anti-tag-scraping analysis module, an early warning module, a data encryption module, a backend service module, and a user feedback module. The AI ​​browser's interface is responsible for user interface display and user interaction.

[0148] The computer information module can obtain relevant information about the current computer device, including CPU serial number, memory serial number, hard disk serial number, MAC address, IP address, operating system version, etc., providing basic data for the anti-spoofing analysis module.

[0149] Face recognition module: Integrates advanced facial recognition algorithms to ensure accurate face detection even in low-light and occluded conditions, and identifies the user through the facial recognition algorithm. The browser analyzes the captured facial information in real time to determine if the user is legitimate.

[0150] Data acquisition module: Collects user actions and uploaded document attributes during the bidding process, such as the path and directory of the bid documents, and the concentration of bidding time.

[0151] Anti-collusion analysis module: Utilizing AI algorithms (i.e., a general-purpose large language model) and pre-set collusion behavior models (i.e., pre-set collusion behavior recognition models), it analyzes user behavior data in real time to identify potential collusion behaviors. Through multi-modal fusion, it combines multiple data sources such as text, images, and network behavior to perform multi-modal fusion analysis, improving the accuracy of analysis and detection. It uses deep learning techniques (such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs)) to train the model, identify complex collusion patterns, and combines expert knowledge and historical cases to formulate a series of rules to assist the model in decision analysis. This may include the following analyses: 1. Bidder basic information analysis: Extracting basic information of bidders (such as company name, registered address, legal representative, etc.) and analyzing whether there are any related relationships. 2. Bid document content analysis: Using natural language processing technology to extract key content of bid documents (such as technical solutions, quotations, commitments, etc.) and analyzing similarity. 3. Historical bidding record analysis: Combining historical bidding data to analyze the performance of bidders in different projects and identify abnormal behavior. 4. Network Behavior Analysis: Monitor the upload time, IP address, MAC address, CPU serial number, and network information of the tender documents to analyze whether there are multiple tender documents from the same IP address or network within the same time period.

[0152] Early warning module: Based on the anti-bid-collusion analysis results, it automatically sends early warnings to relevant personnel, reminding them to be vigilant against bid-collusion. This includes developing a real-time monitoring engine to immediately analyze and evaluate received bid documents. Once potential bid-collusion is detected, it immediately issues a warning to the bidding party and provides a detailed report.

[0153] Data encryption module: Encrypts data transmitted between the AI ​​browser and the bidding system to prevent interception and ensure the security of data transmission and storage. Regular security checks and vulnerability patching are also performed to ensure the security of the AI ​​browser.

[0154] Backend service module: Responsible for processing data analysis and early warning data, providing real-time monitoring and management functions, and dynamically adjusting the cross-selling behavior identification model and early warning strategy according to user behavior and environmental changes to ensure the effectiveness of anti-cross-selling.

[0155] User feedback module: Users can provide feedback on the operation of the AI ​​browser. Based on user feedback, relevant personnel will continuously optimize the AI ​​algorithm and parameter configuration to improve the AI ​​browser's intelligent analysis and decision-making capabilities.

[0156] This embodiment collects the bidder's facial information; determines whether the bidder is a legitimate user based on the facial information; and sends a preset warning message when the bidder is not a legitimate user. This embodiment reduces the probability of bid rigging by restricting bidders to be identical to those registered on the bidding platform and by collecting and comparing the bidder's facial information in real time.

[0157] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the tag-matching behavior identification method of this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0158] This application also provides a device for identifying tag-scratching behavior; please refer to [reference needed]. Figure 5 The tag-scratching behavior identification device includes:

[0159] Module 10 is used to obtain the current bidding information and historical bidding information of the bidder;

[0160] The behavior recognition module 20 is used to identify anti-collusion behavior based on the current bidding information, the historical bidding information and the network behavior of the bidder, and obtain the anti-collusion behavior recognition result.

[0161] The early warning module 30 is used to issue an early warning for cross-selling behavior based on the anti-cross-selling behavior identification result.

[0162] This embodiment obtains the bidder's current and historical bidding information; it uses a large language model to identify anti-collusion behavior based on the current bidding information, historical bidding information, and the bidder's online behavior, obtaining anti-collusion behavior identification results; and it issues a collusion behavior warning based on the anti-collusion behavior identification results. Because this embodiment uses a large language model to identify anti-collusion behavior based on the current bidding information, historical bidding information, and the bidder's online behavior, and issues a collusion behavior warning based on the anti-collusion behavior identification results, compared to existing methods of manually reviewing bids for collusion, this embodiment can comprehensively and accurately identify collusion behavior using intelligent means.

[0163] Based on the first embodiment of the tag-scratching behavior identification device of the present invention, a second embodiment of the tag-scratching behavior identification device of the present invention is proposed.

[0164] In this embodiment, the large language model includes a general large language model and a preset anti-scraping behavior identification model. The behavior identification module 20 is further used to identify anti-scraping behavior based on the preset anti-scraping behavior identification model for the current bidding information and the bidder's network behavior, and obtain a first anti-scraping behavior identification result.

[0165] The historical bidding information is input into the general large language model to obtain the second anti-collusion behavior identification result output by the general large language model;

[0166] The anti-scraping behavior identification result is determined based on the first anti-scraping behavior identification result and the second anti-scraping behavior identification result.

[0167] Furthermore, the behavior recognition module 20 is also used to determine the bid document information and the bidder's basic information based on the current bidding information;

[0168] Based on the preset bid-rigging behavior recognition model, information analysis is performed on the basic information and the bid document information to obtain information recognition results;

[0169] The bidder's network behavior is input into the preset anti-collusion behavior identification model to identify anti-collusion behavior and obtain the network behavior identification result.

[0170] The first anti-scraping behavior identification result is determined based on the network behavior identification result and the information identification result.

[0171] Furthermore, the behavior recognition module 20 is also used to perform bidder association information analysis on the basic information based on the preset bid-rigging behavior recognition model, and determine the association relationship between the bidders;

[0172] The bid document information is feature extracted according to the preset bid-rigging behavior recognition model, and the bid document similarity between bid documents is determined according to the feature extraction results;

[0173] The information identification result is determined based on the aforementioned relationship and the similarity of the tender documents.

[0174] Furthermore, the behavior recognition module 20 is also used to extract features of the bidder's network behavior based on the preset bid-rigging behavior recognition model to obtain network behavior features;

[0175] Based on the network behavior characteristics, target network behaviors that meet preset abnormal conditions are determined.

[0176] The network behavior identification result is determined based on the target network behavior.

[0177] Furthermore, the acquisition module 10 is also used to collect the bidder's facial information;

[0178] Based on the facial information, determine whether the bidder is a legitimate user;

[0179] If the bidder is not a legitimate user, a preset warning message will be sent.

[0180] Furthermore, the acquisition module 10 is also used to determine the target facial features based on the facial information;

[0181] The target facial features are matched against the user information in the bidding application to obtain the matching result;

[0182] The matching results will be used to determine whether the bidder is a legitimate user.

[0183] Furthermore, the behavior recognition module 20 is also used to obtain a preset encryption algorithm when the anti-scraping behavior recognition result is that there is no scrambling behavior;

[0184] The current bidding information is encrypted based on the preset encryption algorithm to obtain encrypted bidding information;

[0185] The encrypted bidding information is sent to the preset bidding terminal.

[0186] Furthermore, the early warning module 30 is also used to obtain a preset early warning prompt strategy when the anti-scraping behavior identification result indicates that there is scrambling behavior;

[0187] Warnings are issued based on the preset warning strategy.

[0188] Furthermore, the behavior recognition module 20 is also used to determine sample data based on the anti-bid-rigging behavior recognition result, the current bidding information, the historical bidding information, and the bidder's network behavior;

[0189] The large language model is iteratively optimized based on the sample data to obtain a new large language model, and the new large language model is used to replace the old large language model.

[0190] Furthermore, the early warning module 30 is also used to receive feedback information from the bidder;

[0191] The large language model is optimized based on the feedback information.

[0192] The tag-scratching behavior identification device provided in this application, employing the tag-scratching behavior identification method in the above embodiments, can solve the technical problem that existing tag-scratching behavior identification relies on manual review and has low identification efficiency. Compared with the prior art, the beneficial effects of the tag-scratching behavior identification device provided in this application are the same as those of the tag-scratching behavior identification method provided in the above embodiments, and other technical features in the tag-scratching behavior identification device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0193] This application provides a tag-scratching behavior recognition device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the tag-scratching behavior recognition method in the above embodiment 1.

[0194] The following is for reference. Figure 6 The diagram illustrates a structural schematic of a tag-matching behavior recognition device suitable for implementing embodiments of this application. The tag-matching behavior recognition device in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 6 The illustrated tag-trading behavior recognition device is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0195] like Figure 6As shown, the tag behavior recognition device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the tag behavior recognition device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the tag-spoofing behavior recognition device to communicate wirelessly or wiredly with other devices to exchange data. Although tag-spoofing behavior recognition devices with various systems are shown in the figures, it should be understood that it is not required to implement or possess all of the systems shown. More or fewer systems can be implemented alternatively.

[0196] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0197] The tag-scratching behavior identification device provided in this application, employing the tag-scratching behavior identification method in the above embodiments, can solve the technical problem that existing tag-scratching behavior identification relies on manual review and has low identification efficiency. Compared with the prior art, the beneficial effects of the tag-scratching behavior identification device provided in this application are the same as those of the tag-scratching behavior identification method provided in the above embodiments, and other technical features in this tag-scratching behavior identification device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0198] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0199] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0200] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the tag-scratching behavior identification method in the above embodiments.

[0201] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0202] The aforementioned computer-readable storage medium may be included in the serializer behavior recognition device; or it may exist independently and not assembled into the serializer behavior recognition device.

[0203] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0204] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0205] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0206] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described smuggling behavior identification method. This solves the technical problem that existing smuggling behavior identification methods rely on manual review and have low identification efficiency. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the smuggling behavior identification method provided in the above embodiments, and will not be repeated here.

[0207] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described method for identifying trolling behavior.

[0208] The computer program product provided in this application can solve the technical problem that existing methods for identifying smuggled targets rely on manual review, resulting in low identification efficiency. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the smuggled target identification method provided in the above embodiments, and will not be repeated here.

[0209] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

[0210] This application also discloses A1. A method for identifying cross-trading behavior, the method comprising the following steps:

[0211] Obtain the current and historical bidding information of bidders;

[0212] Based on a large language model, anti-collusion behavior identification is performed on the current bidding information, the historical bidding information, and the network behavior of the bidders to obtain anti-collusion behavior identification results.

[0213] Based on the anti-scraping behavior identification results, a warning of scrambling behavior is issued.

[0214] A2. The method for identifying bid-rigging behavior as described in A1, wherein the large language model includes a general large language model and a preset bid-rigging behavior identification model, and the step of identifying bid-rigging behavior based on the large language model of the current bidding information, the historical bidding information, and the bidder's network behavior to obtain the bid-rigging behavior identification result includes:

[0215] Based on the preset bid-rigging behavior identification model, the current bidding information and the bidder's network behavior are used to identify anti-bid-rigging behavior, and a first anti-bid-rigging behavior identification result is obtained.

[0216] The historical bidding information is input into the general large language model to obtain the second anti-collusion behavior identification result output by the general large language model;

[0217] The anti-scraping behavior identification result is determined based on the first anti-scraping behavior identification result and the second anti-scraping behavior identification result.

[0218] A3. The method for identifying bid-rigging behavior as described in A2, wherein the step of identifying anti-bid-rigging behavior based on the preset bid-rigging behavior identification model of the current bidding information and the bidder's network behavior to obtain a first anti-bid-rigging behavior identification result includes:

[0219] Based on the current bidding information, determine the bid document information and the bidder's basic information;

[0220] Based on the preset bid-rigging behavior recognition model, information analysis is performed on the basic information and the bid document information to obtain information recognition results;

[0221] The bidder's network behavior is input into the preset anti-collusion behavior identification model to identify anti-collusion behavior and obtain the network behavior identification result.

[0222] The first anti-scraping behavior identification result is determined based on the network behavior identification result and the information identification result.

[0223] A4. The method for identifying bid-rigging behavior as described in A3, wherein the step of performing information analysis on the basic information and the bid document information based on the preset bid-rigging behavior identification model to obtain the information identification result includes:

[0224] Based on the preset bid-rigging behavior identification model, the basic information is analyzed to determine the relationship between the bidders.

[0225] The bid document information is feature extracted according to the preset bid-rigging behavior recognition model, and the bid document similarity between bid documents is determined according to the feature extraction results;

[0226] The information identification result is determined based on the aforementioned relationship and the similarity of the tender documents.

[0227] A5. The method for identifying bid-rigging behavior as described in A3, wherein the step of inputting the bidder's network behavior into the preset bid-rigging behavior identification model for anti-bid-rigging behavior identification and obtaining the network behavior identification result includes:

[0228] Based on the preset bid-rigging behavior identification model, the network behavior features of the bidder are extracted to obtain network behavior features;

[0229] Based on the network behavior characteristics, target network behaviors that meet preset abnormal conditions are determined.

[0230] The network behavior identification result is determined based on the target network behavior.

[0231] A6. The method for identifying bid-rigging behavior as described in A1, prior to the step of obtaining the bidder's current bidding information and historical bidding information, further includes:

[0232] Collect the bidders' facial information;

[0233] Based on the facial information, determine whether the bidder is a legitimate user;

[0234] If the bidder is not a legitimate user, a preset warning message will be sent.

[0235] A7. The method for identifying bid-rigging behavior as described in A6, wherein the step of determining whether the bidder is a legitimate user based on the facial information includes:

[0236] The target facial features are determined based on the facial information;

[0237] The target facial features are matched against the user information in the bidding application to obtain the matching result;

[0238] The matching results will be used to determine whether the bidder is a legitimate user.

[0239] A8. The method for identifying bid-rigging behavior as described in any one of A1-A7, after the step of identifying anti-bid-rigging behavior based on a large language model of the current bidding information, the historical bidding information, and the bidder's network behavior to obtain the anti-bid-rigging behavior identification result, further includes:

[0240] When the anti-spoofing behavior identification result is that there is no spoofing behavior, a preset encryption algorithm is obtained;

[0241] The current bidding information is encrypted using the preset encryption algorithm to obtain encrypted bidding information; the encrypted bidding information is then sent to the preset bidding terminal.

[0242] A9. The method for identifying cross-splitting behavior as described in any one of A1-A7, wherein the step of issuing a cross-splitting behavior warning based on the anti-cross-splitting behavior identification result includes:

[0243] When the anti-scraping behavior identification result indicates the presence of scrambling behavior, a preset early warning prompt strategy is obtained;

[0244] Warnings are issued based on the preset warning strategy.

[0245] A10. The method for identifying bid-rigging behavior as described in any one of A1-A7, after the step of identifying anti-bid-rigging behavior based on a large language model of the current bidding information, the historical bidding information, and the bidder's network behavior to obtain the anti-bid-rigging behavior identification result, further includes:

[0246] Sample data is determined based on the anti-collusion behavior identification results, the current bidding information, the historical bidding information, and the bidder's network behavior;

[0247] The large language model is iteratively optimized based on the sample data to obtain a new large language model, and the new large language model is used to replace the old large language model.

[0248] A11. The method for identifying cross-splitting behavior as described in any one of A1-A7, after the step of issuing a cross-splitting behavior warning based on the anti-cross-splitting behavior identification result, further includes:

[0249] Receive feedback information from the bidder;

[0250] The large language model is optimized based on the feedback information.

[0251] This application also discloses B12. A bid-rigging behavior identification device, the bid-rigging behavior identification device comprising: an acquisition module, used to acquire the current bidding information and historical bidding information of the bidder;

[0252] The behavior recognition module is used to identify anti-collusion behavior based on the current bidding information, the historical bidding information, and the network behavior of the bidders, and to obtain the anti-collusion behavior recognition result.

[0253] The early warning module is used to issue early warnings about cross-selling behavior based on the anti-cross-selling behavior identification results.

[0254] B13. The trolling behavior identification device as described in B12, wherein the large language model includes a general large language model and a preset trolling behavior identification model, and the behavior identification module is further configured to perform anti-trolling behavior identification on the current bidding information and the bidder's network behavior based on the preset trolling behavior identification model, and obtain a first anti-trolling behavior identification result;

[0255] The historical bidding information is input into the general large language model to obtain the second anti-collusion behavior identification result output by the general large language model;

[0256] The anti-scraping behavior identification result is determined based on the first anti-scraping behavior identification result and the second anti-scraping behavior identification result.

[0257] B14. The bid-rigging behavior identification device as described in B13, wherein the behavior identification module is further configured to determine the bid document information and the basic information of the bidder based on the current bidding information;

[0258] Based on the preset bid-rigging behavior recognition model, information analysis is performed on the basic information and the bid document information to obtain information recognition results;

[0259] The bidder's network behavior is input into the preset anti-collusion behavior identification model to identify anti-collusion behavior and obtain the network behavior identification result.

[0260] The first anti-scraping behavior identification result is determined based on the network behavior identification result and the information identification result.

[0261] B15. The bid-rigging behavior identification device as described in B14, wherein the behavior identification module is further configured to perform bidder association information analysis on the basic information based on the preset bid-rigging behavior identification model, and determine the association relationship between the bidders;

[0262] The bid document information is feature extracted according to the preset bid-rigging behavior recognition model, and the bid document similarity between bid documents is determined according to the feature extraction results;

[0263] The information identification result is determined based on the aforementioned relationship and the similarity of the tender documents.

[0264] B16. The bid-rigging behavior identification device as described in B14, wherein the behavior identification module is further configured to extract features of the bidder's network behavior based on the preset bid-rigging behavior identification model to obtain network behavior features;

[0265] Based on the network behavior characteristics, target network behaviors that meet preset abnormal conditions are determined.

[0266] The network behavior identification result is determined based on the target network behavior.

[0267] B17. The bid-rigging behavior identification device as described in B12, wherein the acquisition module is further used to collect the bidder's facial information;

[0268] Based on the facial information, determine whether the bidder is a legitimate user;

[0269] If the bidder is not a legitimate user, a preset warning message will be sent.

[0270] This application also discloses C18. A tag-scratching behavior identification device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the tag-scratching behavior identification method as described in any one of A1 to A11.

[0271] This application also discloses D19. A storage medium, which is a computer-readable storage medium, on which a computer program is stored, which, when executed by a processor, implements the steps of the tag-scratching behavior identification method as described in any one of A1 to A11.

[0272] This application also discloses F20. A computer program product comprising a computer program that, when executed by a processor, implements the steps of the tag-scratching behavior identification method as described in any one of A1 to A11.

Claims

1. A method for identifying tag-scratching behavior, characterized in that, The method for identifying cross-signature behavior includes the following steps: Obtain the current and historical bidding information of bidders; Based on a large language model, anti-collusion behavior identification is performed on the current bidding information, the historical bidding information, and the network behavior of the bidders to obtain anti-collusion behavior identification results. Based on the anti-scraping behavior identification results, a warning of scrambling behavior is issued.

2. The method for identifying tagged behavior as described in claim 1, characterized in that, The large language model includes a general large language model and a preset bid-rigging behavior identification model. The step of identifying bid-rigging behavior based on the large language model of the current bidding information, the historical bidding information, and the bidder's network behavior to obtain the bid-rigging behavior identification result includes: Based on the preset bid-rigging behavior identification model, the current bidding information and the bidder's network behavior are used to identify anti-bid-rigging behavior, and a first anti-bid-rigging behavior identification result is obtained. The historical bidding information is input into the general large language model to obtain the second anti-collusion behavior identification result output by the general large language model; The anti-scraping behavior identification result is determined based on the first anti-scraping behavior identification result and the second anti-scraping behavior identification result.

3. The method for identifying tag-scratching behavior as described in claim 2, characterized in that, The step of identifying anti-collusion behavior based on the preset collusion behavior identification model to obtain a first anti-collusion behavior identification result includes: Based on the current bidding information, determine the bid document information and the bidder's basic information; Based on the preset bid-rigging behavior recognition model, information analysis is performed on the basic information and the bid document information to obtain information recognition results; The bidder's network behavior is input into the preset anti-collusion behavior identification model to identify anti-collusion behavior and obtain the network behavior identification result. The first anti-scraping behavior identification result is determined based on the network behavior identification result and the information identification result.

4. The method for identifying tagged behavior as described in claim 3, characterized in that, The step of analyzing the basic information and the bid document information based on the preset bid-rigging behavior recognition model to obtain the information recognition result includes: Based on the preset bid-rigging behavior identification model, the basic information is analyzed to determine the relationship between the bidders. The bid document information is feature extracted according to the preset bid-rigging behavior recognition model, and the bid document similarity between bid documents is determined according to the feature extraction results; The information identification result is determined based on the aforementioned relationship and the similarity of the tender documents.

5. The method for identifying tag-scratching behavior as described in claim 3, characterized in that, The step of inputting the bidder's network behavior into the preset bid-rigging behavior identification model for anti-bid-rigging behavior identification and obtaining the network behavior identification result includes: Based on the preset bid-rigging behavior identification model, the network behavior features of the bidder are extracted to obtain network behavior features; Based on the network behavior characteristics, target network behaviors that meet preset abnormal conditions are determined. The network behavior identification result is determined based on the target network behavior.

6. The method for identifying tag-scratching behavior as described in claim 1, characterized in that, Before the step of obtaining the bidder's current bidding information and historical bidding information, the method further includes: Collect the bidders' facial information; Based on the facial information, determine whether the bidder is a legitimate user; If the bidder is not a legitimate user, a pre-set warning message will be sent.

7. A device for identifying tag-trading behavior, characterized in that, The tag-scraping behavior identification device includes: The acquisition module is used to acquire the bidder's current bidding information and historical bidding information; The behavior recognition module is used to identify anti-collusion behavior based on the current bidding information, the historical bidding information, and the network behavior of the bidders, and to obtain the anti-collusion behavior recognition result. The early warning module is used to issue early warnings about cross-selling behavior based on the anti-cross-selling behavior identification results.

8. A device for identifying tag-trading behavior, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the tag-scratching behavior identification method as described in any one of claims 1 to 6.

9. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the tag-scratching behavior identification method as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the tag-scratching behavior identification method as described in any one of claims 1 to 6.

Images