CN117575601A - Transaction behavior identification method and device, electronic equipment and storage medium - Google Patents

Transaction behavior identification method and device, electronic equipment and storage medium Download PDF

Info

Publication number
CN117575601A
CN117575601A CN202311554691.5A CN202311554691A CN117575601A CN 117575601 A CN117575601 A CN 117575601A CN 202311554691 A CN202311554691 A CN 202311554691A CN 117575601 A CN117575601 A CN 117575601A
Authority
CN
China
Prior art keywords
target
prompt
large model
instruction
flow
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202311554691.5A
Other languages
Chinese (zh)
Inventor
李犇
张�杰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Zhongguancun Kejin Technology Co Ltd
Original Assignee
Beijing Zhongguancun Kejin Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Zhongguancun Kejin Technology Co Ltd filed Critical Beijing Zhongguancun Kejin Technology Co Ltd
Priority to CN202311554691.5A priority Critical patent/CN117575601A/en
Publication of CN117575601A publication Critical patent/CN117575601A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q20/00—Payment architectures, schemes or protocols
    • G06Q20/38—Payment protocols; Details thereof
    • G06Q20/40—Authorisation, e.g. identification of payer or payee, verification of customer or shop credentials; Review and approval of payers, e.g. check credit lines or negative lists
    • G06Q20/401—Transaction verification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/35—Clustering; Classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/04—Trading; Exchange, e.g. stocks, commodities, derivatives or currency exchange
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10—Character recognition
    • G06V30/19—Recognition using electronic means
    • G06V30/191—Design or setup of recognition systems or techniques; Extraction of features in feature space; Clustering techniques; Blind source separation
    • G06V30/19173—Classification techniques
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Accounting & Taxation (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Finance (AREA)
  • Strategic Management (AREA)
  • General Business, Economics & Management (AREA)
  • Data Mining & Analysis (AREA)
  • Computer Security & Cryptography (AREA)
  • Databases & Information Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Multimedia (AREA)
  • Development Economics (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Technology Law (AREA)
  • Machine Translation (AREA)

Abstract

The application provides a transaction behavior identification method, a device, electronic equipment and a storage medium, wherein the method comprises the following steps: generating a record text based on the abnormal transaction behavior; generating a target prompt instruction according to the recorded text; performing instruction adjustment on the pre-trained large model according to the target prompt instruction to obtain a first target large model; inputting the recorded text into a first target large model to obtain a first recognition result; and inputting the first recognition result into a second target large model to obtain a second recognition result. According to the method and the device, the target prompt instruction is determined through the recorded text of the abnormal transaction behavior, the pre-trained large model is subjected to instruction adjustment according to the target prompt instruction, and the first target large model is obtained, so that the recorded text is input into the first target large model, the first recognition result is obtained, the second recognition result is finally obtained by inputting the recorded text into the first target large model, and the recognition efficiency of the abnormal transaction behavior is improved.

Description

Transaction behavior identification method and device, electronic equipment and storage medium
Technical Field
The present disclosure relates to the field of large model technologies, and in particular, to a method and apparatus for identifying transaction behavior, an electronic device, and a storage medium.
Background
With the development of communication technology, more and more users use online payment, and the online payment is prone to abnormal transaction behaviors, such as telecommunication fraud. At present, the case situation of abnormal transaction behaviors is generally identified and analyzed manually, so that the problem of low transaction behavior identification efficiency is caused.
Disclosure of Invention
The embodiment of the application provides a transaction behavior identification method, a device, electronic equipment and a storage medium, which are used for solving the problem of low transaction behavior identification efficiency in the prior art.
In order to solve the technical problems, the application is realized as follows:
in a first aspect, an embodiment of the present application provides a method for identifying transaction behavior. The method comprises the following steps:
generating a record text based on the abnormal transaction behavior, wherein the record text comprises an event flow of the abnormal transaction behavior;
generating a target prompt instruction according to the recorded text;
performing instruction adjustment on the pre-trained large model according to the target prompt instruction to obtain a first target large model;
inputting the recorded text into a first target large model to obtain a first recognition result;
and inputting the first identification result into a second target large model to obtain a second identification result, wherein the second identification result is used for indicating whether the flow information is matched with the historical flow information corresponding to the historical abnormal transaction behavior.
Optionally, the target prompt instruction includes an element prompt instruction and a flow prompt instruction, and the generating the target prompt instruction according to the recorded text includes:
obtaining target prompt words, wherein the target prompt words comprise element prompt words and flow prompt words;
generating the element prompt instruction according to the element prompt word;
and generating the flow prompt instruction according to the flow prompt word.
Optionally, after the target prompt word is obtained, the method includes:
acquiring a prompt instruction template;
the generating the element prompt instruction according to the element prompt word comprises the following steps:
filling the element prompt words into the prompt instruction template to generate the element prompt instruction;
the generating a flow prompt instruction according to the flow prompt word comprises the following steps:
and filling the flow prompt word into the prompt instruction template to generate the flow prompt instruction.
Optionally, before the instruction adjustment is performed on the pre-trained large model according to the target prompt instruction to obtain the first target large model, the method further includes:
acquiring a domain corpus, wherein the domain corpus comprises text data of historical abnormal transaction behaviors;
and performing domain adjustment on the first large model according to the domain corpus to obtain the pre-trained large model, wherein the first large model is a large language model.
Optionally, the performing instruction adjustment on the pre-trained large model according to the target prompt instruction to obtain a first target large model includes:
generating a domain prompt instruction corpus according to the element prompt instruction and the flow prompt instruction;
and carrying out instruction adjustment on the pre-trained large model based on the domain prompt instruction corpus to obtain the first target large model.
Optionally, after the first recognition result is input into the second target large model and the second recognition result is obtained, the method further includes:
acquiring data information of historical abnormal transaction behaviors, wherein the data information is used for indicating the types, flows and information of the historical abnormal transaction behaviors;
and generating an analysis result based on the first identification result, the second identification result and the data information of the historical abnormal transaction behavior, wherein the analysis result is used for indicating the development trend of the abnormal transaction behavior.
Optionally, the generating the record text based on the abnormal transaction behavior includes:
acquiring record information corresponding to abnormal transaction behaviors;
and inputting the recorded information into an optical character recognition OCR model for format conversion to obtain the recorded text.
In a second aspect, an embodiment of the present application further provides a transaction behavior identification device. The transaction behavior recognition device comprises:
the first generation module is used for generating a record text based on the abnormal transaction behavior, wherein the record text comprises an event flow of the abnormal transaction behavior;
the second generation module is used for generating a target prompt instruction according to the recorded text;
the adjustment module is used for carrying out instruction adjustment on the pre-trained large model according to the target prompt instruction to obtain a first target large model;
the first input module is used for inputting the recorded text into a first target large model to obtain a first identification result;
the second input module is used for inputting the first identification result into a second target large model to obtain a second identification result, and the second identification result is used for indicating whether the flow information is matched with the historical flow information corresponding to the historical abnormal transaction behavior.
In a third aspect, an embodiment of the present application further provides an electronic device, including a processor, a memory, and a computer program stored on the memory and executable on the processor, where the computer program when executed by the processor implements the steps of the method for identifying transaction behavior described above.
In a fourth aspect, embodiments of the present application further provide a computer readable storage medium having a computer program stored thereon, where the computer program when executed by a processor implements the steps of the above-described transaction behavior identification method.
The application provides a transaction behavior identification method, a device, electronic equipment and a storage medium, wherein the method comprises the following steps: generating a record text based on the abnormal transaction behavior, wherein the record text comprises an event flow of the abnormal transaction behavior; generating a target prompt instruction according to the recorded text; performing instruction adjustment on the pre-trained large model according to the target prompt instruction to obtain a first target large model; inputting the recorded text into a first target large model to obtain a first recognition result; and inputting the first identification result into a second target large model to obtain a second identification result, wherein the second identification result is used for indicating whether the flow information is matched with the historical flow information corresponding to the historical abnormal transaction behavior. According to the method and the device, the target prompt instruction is determined through the recorded text of the abnormal transaction behavior, the pre-trained large model is subjected to instruction adjustment according to the target prompt instruction, and the first target large model is obtained, so that the recorded text is input into the first target large model, the first recognition result is obtained, the second recognition result is finally obtained by inputting the recorded text into the first target large model, and the recognition efficiency of the abnormal transaction behavior is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the description of the embodiments of the present application will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a method for identifying transaction behavior provided by an embodiment of the present application;
FIG. 2 is a schematic diagram of a model training architecture provided in an embodiment of the present application;
FIG. 3 is a block diagram of a transaction behavior recognition device provided in an embodiment of the present application;
fig. 4 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
The following description of the embodiments of the present application will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all, of the embodiments of the present application. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments herein without making any inventive effort, are intended to be within the scope of the present application.
The embodiment of the application provides a transaction behavior identification method. Referring to fig. 1, fig. 1 is a flowchart of a method for identifying transaction behavior according to an embodiment of the present application, as shown in fig. 1, including the following steps:
step 101, generating a record text based on abnormal transaction behaviors, wherein the record text comprises event flows of the abnormal transaction behaviors.
In this embodiment, the abnormal transaction behavior is generally actions such as telecommunication fraud, and in this embodiment, after the abnormal transaction behavior occurs to the user, the transcript content is generally generated, so that the staff can analyze the type, flow and information of the abnormal transaction behavior conveniently. Among these types are phishing, software fraud, linking fraud, picture fraud, etc. The process is a process that the user is deceived, for example, the user clicks a link sent by a fraud party, thereby causing the mobile phone of the user to be poisoned, and further causing the bank card of the user to be stolen and swiped. The information is relevant characteristics of the user, such as the amount of fraud, the age of the user, the occupation of the user, the income of the user, etc.
A recorded text is generated by the abnormal transaction behavior, and specifically, the recorded text details the event flow of the abnormal transaction behavior.
And 102, generating a target prompt instruction according to the recorded text.
In this embodiment, the target prompt instruction is a prompt instruction in a large language model (Large Language Models, LLM), and the prompt instruction can help the large language model (Large Language Models, LLM) to understand the intention of the user better and output more accurate and meaningful content, so that the large model meets the requirement of the user better.
Specifically, one or more prompt instructions (prompt) are generated according to the recorded text, and the one or more prompt instructions conduct instruction fine adjustment on the large model, so that the large model can have a better recognition effect on the recorded text.
And 103, performing instruction adjustment on the pre-trained large model according to the target prompt instruction to obtain a first target large model.
In this embodiment, prompt instructions of different tasks are input to a pre-trained large model for instruction fine adjustment, where the pre-trained large model is a large model that is primarily trained, and has a certain recognition function. And carrying out iterative training on the pre-trained large model in a command fine-tuning mode until the large model converges to a final effect, thereby finally obtaining the target large model.
And 104, inputting the recorded text into a first target large model to obtain a first recognition result.
In this embodiment, by inputting the recorded text into the first target large model, a first recognition result is output, where the first recognition result is used to indicate a fraud flow and fraud elements corresponding to the recorded text.
Step 105, inputting the first recognition result into a second target big model to obtain a second recognition result, wherein the second recognition result is used for indicating whether the flow information is matched with the historical flow information corresponding to the historical abnormal transaction behavior.
In this embodiment, the first recognition result is input into the second target large model, and the second recognition result is output, so as to determine whether the current abnormal transaction behavior is matched with the historical transaction behavior, and when the current abnormal transaction behavior is matched with the historical transaction behavior, the current abnormal transaction behavior is indicated to be a telecommunication fraud means, and when the current abnormal transaction behavior is not matched with the historical transaction behavior, the current abnormal transaction behavior is indicated to be a novel fraud means, and recording can be performed and early warning is performed to other users. The second target large model is a field large language model, comprises a plurality of historical fraud flows, accompanies the second identification result, outputs the score of the case, and judges that the case is a novel fraud method if the score is larger than a threshold value. And early warning is carried out.
The application provides a transaction behavior identification method, which comprises the following steps: generating a record text based on the abnormal transaction behavior, wherein the record text comprises an event flow of the abnormal transaction behavior; generating a target prompt instruction according to the recorded text; performing instruction adjustment on the pre-trained large model according to the target prompt instruction to obtain a first target large model; inputting the recorded text into a first target large model to obtain a first recognition result; and inputting the first identification result into a second target large model to obtain a second identification result, wherein the second identification result is used for indicating whether the flow information is matched with the historical flow information corresponding to the historical abnormal transaction behavior. According to the method and the device, the target prompt instruction is determined through the recorded text of the abnormal transaction behavior, the pre-trained large model is subjected to instruction adjustment according to the target prompt instruction, and the first target large model is obtained, so that the recorded text is input into the first target large model, the first recognition result is obtained, the second recognition result is finally obtained by inputting the recorded text into the first target large model, and the recognition efficiency of the abnormal transaction behavior is improved.
In some possible implementations, optionally, the target prompt instruction includes an element prompt instruction and a flow prompt instruction, and the generating the target prompt instruction according to the recorded text includes:
obtaining target prompt words, wherein the target prompt words comprise element prompt words and flow prompt words;
generating the element prompt instruction according to the element prompt word;
and generating the flow prompt instruction according to the flow prompt word.
In this embodiment, an element prompt instruction and a flow prompt instruction are generated according to the recorded text, where in this embodiment, instruction fine adjustment is performed on the pre-trained large model by using one or more prompt instructions, so that the pre-trained large model achieves an effect, and it needs to be described that, under the condition that the more prompt instructions are, the pre-trained large model meets the requirements, and in this embodiment, two are taken as examples to be described.
Illustratively, the recorded text is as follows: "record": "202X year XX month XX last year, who pulled me to a micro-letter group, the information sent by a person in the group is that the job information of a person is that the job of ZZ software attention is taken in APP of a name YY, then the person clicks APP of the group to register the job of ZZ software attention in APP of a name public welfare 68 group, then the person paying attention to the appointed account number according to the requirement of the group sends screenshot to the group, then the person seeking a name assistant with a fragrance in APP sends the micro-letter to the code him, then the my micro-letter receives a commission of 3.8 units, then me takes about ten commissions sequentially, the price of the commission is also increased from 3,8 units to 7.8 units, then the assistant-fragrance tells me to obtain higher commission, I ask him how to operate, she say that I add a job calling a sender to spend for her to do proxy tasks, I add sender-spend after she is first to chat with I normally, then tell I need to do two power-assisted tasks to upgrade to proxy, then I just recharge a 200 yuan and a 300 yuan of principal tasks according to her say, then the two tasks are returned to me 60 yuan of commission respectively, I wait for me to spend after doing the two tasks, and then tell I say that I send a 589 in the group, I don't pay attention to send 598 at that time, then send sender-spend to tell me because of me cause system error to recharge me 10288 yuan, I do not walk to recharge me 400 yuan of repair, I say me more money, then recharge me 10000 first, after the self-charging is finished, the loan of the self is charged with 30000 units, and then the self feels that the self is deceived, so that the public security office alarms. "
For the converted transcript text, generating a fraud element extraction instruction (prompt), wherein the instruction is constructed as follows:
{ "Record": < transcript text data >, "sample": please you are as a policeman to handle a case, and the key information elements in the case are extracted according to the records of the telecom fraud victim: victim information, suspect information, time, place, micro-signals, weChat name, app name, etc., output according to json format: "}.
And inputting a fraud element extraction instruction (prompt) into a domain Large Language Model (LLM) to obtain fraud case key information element data of the output of the model.
Generating a fraud flow extraction instruction (prompt), the instruction being structured as follows:
{ "Record": < text data >, "sample": "please you as a policeman transacting case, extract the suspicious person to implement the fraudulent process according to the Record of the victim of telecommunication fraud, and data according to json format: "}.
And inputting a fraud flow extraction instruction (prompt) into a domain Large Language Model (LLM) to obtain fraud case flow data output by the model.
Optionally, after the target prompt word is obtained, the method includes:
acquiring a prompt instruction template;
the generating the element prompt instruction according to the element prompt word comprises the following steps:
filling the element prompt words into the prompt instruction template to generate the element prompt instruction;
the generating a flow prompt instruction according to the flow prompt word comprises the following steps:
and filling the flow prompt word into the prompt instruction template to generate the flow prompt instruction.
In this embodiment, as shown in fig. 2, after relevant instructions are filled in a prompt instruction template, different prompt instructions may be generated, for example, the element prompt words and the record text are filled in the prompt instruction template, the element prompt instructions are generated, and the flow prompt words and the record text are filled in the prompt instruction template, so as to generate the flow prompt instructions.
Optionally, before the instruction adjustment is performed on the pre-trained large model according to the target prompt instruction to obtain the first target large model, the method further includes:
acquiring a domain corpus, wherein the domain corpus comprises text data of historical abnormal transaction behaviors;
and performing domain adjustment on the first large model according to the domain corpus to obtain the pre-trained large model, wherein the first large model is a large language model.
In this embodiment, the domain corpus includes language+resource, where language refers to a defined domain of resource, resource=resource+source, which is a source or summary of the data, and together, such a definition is formed: any collection of language units may be referred to as a language resource. The language resource is an indispensable component part in the natural language processing task, on one hand, the language resource is the support of the related language processing task, the prior knowledge is provided for the language processing task to assist, on the other hand, the language processing task also provides requirements for the language resource, and the technical support function can be achieved for the construction and expansion of the language resource.
The domain corpus in the embodiment includes text data of historical abnormal transaction behaviors, wherein the historical abnormal transaction behaviors can be texts generated by multiple abnormal transaction behaviors, domain fine adjustment is performed on the first large model through the domain corpus, so that the first large model can clearly know relevant terms, contents, information and the like in the abnormal transaction domain, and a pre-trained large model can be obtained, wherein the pre-trained large model can accurately identify relevant information in the domain.
Optionally, the performing instruction adjustment on the pre-trained large model according to the target prompt instruction to obtain a first target large model includes:
generating a domain prompt instruction corpus according to the element prompt instruction and the flow prompt instruction;
and carrying out instruction adjustment on the pre-trained large model based on the domain prompt instruction corpus to obtain the first target large model.
In this embodiment, the element prompt instruction and the flow prompt instruction are combined to generate a domain prompt instruction corpus, and instruction adjustment is performed on the pre-trained large model according to the domain prompt instruction corpus, so as to obtain a target large model.
In the above embodiment, the first large model is subjected to the domain fine adjustment by the domain corpus, but the present embodiment can perform the instruction adjustment on the pre-trained large model by the domain prompt instruction corpus, and the difference between the domain fine adjustment and the instruction fine adjustment is that. Instruction trimming differs from trimming in domain-specific data in that instruction trimming is typically performed in a base model, i.e., some parameters are trimmed to accommodate a task, whereas trimming in domain-specific data focuses more on finding an appropriate model in a domain to improve the accuracy of the data.
Optionally, after the first recognition result is input into the second target large model and the second recognition result is obtained, the method further includes:
acquiring data information of historical abnormal transaction behaviors, wherein the data information is used for indicating the types, flows and information of the historical abnormal transaction behaviors;
and generating an analysis result based on the first identification result, the second identification result and the data information of the historical abnormal transaction behavior, wherein the analysis result is used for indicating the development trend of the abnormal transaction behavior.
In this embodiment, the data information of the historical abnormal transaction behavior and the recognition result of the current abnormal transaction behavior are analyzed, so as to analyze and judge trends of fraud cases, for example, to judge whether the crime rate of the current telecommunication fraud is increased or whether the current telecommunication fraud is of a novel fraud type, etc.
Optionally, the generating the record text based on the abnormal transaction behavior includes:
acquiring record information corresponding to abnormal transaction behaviors;
and inputting the recorded information into an optical character recognition OCR model for format conversion to obtain the recorded text.
In this embodiment, the record information corresponding to the abnormal transaction behavior is a report form of a typical victim. An optical character recognition (Optical Character Recognition, OCR) model refers to a technique of searching, extracting, and recognizing characters in a picture, determining the shape thereof by detecting dark and bright patterns, and then translating the shape into computer characters by a character recognition method. The record text is obtained by converting the record into a record text by using an OCR model.
The application provides a transaction behavior identification method, which comprises the following steps: generating a record text based on the abnormal transaction behavior, wherein the record text comprises an event flow of the abnormal transaction behavior; generating a target prompt instruction according to the recorded text; performing instruction adjustment on the pre-trained large model according to the target prompt instruction to obtain a first target large model; inputting the recorded text into a first target large model to obtain a first recognition result; and inputting the first identification result into a second target large model to obtain a second identification result, wherein the second identification result is used for indicating whether the flow information is matched with the historical flow information corresponding to the historical abnormal transaction behavior. According to the method and the device, the target prompt instruction is determined through the recorded text of the abnormal transaction behavior, the pre-trained large model is subjected to instruction adjustment according to the target prompt instruction, and the first target large model is obtained, so that the recorded text is input into the first target large model, the first recognition result is obtained, the second recognition result is finally obtained by inputting the recorded text into the first target large model, and the recognition efficiency of the abnormal transaction behavior is improved.
Referring to fig. 3, fig. 3 is a block diagram of a transaction behavior recognition device according to an embodiment of the present application. As shown in fig. 3, the transaction behavior recognition device 300 includes:
a first generation module 310, configured to generate a record text based on the abnormal transaction behavior, where the record text includes an event flow of the abnormal transaction behavior;
a second generating module 320, configured to generate a target prompt instruction according to the recorded text;
the adjustment module 330 is configured to perform instruction adjustment on the pre-trained large model according to the target prompt instruction, so as to obtain a first target large model;
the first input module 340 is configured to input the recorded text into a first target large model, so as to obtain a first recognition result;
the second input module 350 is configured to input the first recognition result into a second target big model, and obtain a second recognition result, where the second recognition result is used to indicate whether the flow information matches with historical flow information corresponding to a historical abnormal transaction behavior.
Optionally, the target prompt instruction includes an element prompt instruction and a flow prompt instruction, and the second generating module 320 includes:
the first acquisition sub-module is used for acquiring target prompt words, wherein the target prompt words comprise element prompt words and flow prompt words;
the first generation sub-module is used for generating the element prompt instruction according to the element prompt word;
and the second generation sub-module is used for generating the flow prompt instruction according to the flow prompt word.
Optionally, the method further comprises:
the second acquisition sub-module is used for acquiring a prompt instruction template;
a first generation sub-module comprising:
the first generation unit is used for filling the element prompt words into the prompt instruction template to generate the element prompt instruction;
a second generation sub-module, comprising:
the first generation unit is used for filling the flow prompt word into the prompt instruction template to generate the flow prompt instruction.
Optionally, the method further comprises:
the field acquisition module is used for acquiring field corpus, wherein the field corpus comprises text data of historical abnormal transaction behaviors;
the domain adjusting module is used for carrying out domain adjustment on the first large model according to the domain corpus to obtain the pre-trained large model, wherein the first large model is a large language model.
Optionally, the adjusting module 330 includes:
the third generation sub-module is used for generating a domain prompt instruction corpus according to the element prompt instruction and the flow prompt instruction;
and the adjustment sub-module is used for carrying out instruction adjustment on the pre-trained large model based on the domain prompt instruction corpus to obtain the first target large model.
Optionally, the method further comprises:
the historical information acquisition module is used for acquiring data information of historical abnormal transaction behaviors, wherein the data information is used for indicating types, flows and information of the historical abnormal transaction behaviors;
the historical information analysis module is used for generating an analysis result based on the identification result and the data information of the historical abnormal transaction behavior, and the analysis result is used for indicating the development trend of the abnormal transaction behavior.
Optionally, the first generating module 310 includes:
the information acquisition sub-module is used for acquiring record information corresponding to abnormal transaction behaviors;
and the format conversion sub-module is used for inputting the record information into an optical character recognition OCR model to perform format conversion so as to obtain the record text.
According to the method and the device, the target prompt instruction is determined through the recorded text of the abnormal transaction behavior, the pre-trained large model is subjected to instruction adjustment according to the target prompt instruction, and the first target large model is obtained, so that the recorded text is input into the first target large model, the first recognition result is obtained, the second recognition result is finally obtained by inputting the recorded text into the first target large model, and the recognition efficiency of the abnormal transaction behavior is improved.
The embodiment of the application also provides electronic equipment. Referring to fig. 4, an electronic device may include a processor 401, a memory 402, and a program 4021 stored on the memory 402 and executable on the processor 401.
The program 4021, when executed by the processor 401, may implement any of the steps in the method embodiment corresponding to fig. 1:
generating a record text based on the abnormal transaction behavior, wherein the record text comprises an event flow of the abnormal transaction behavior;
generating a target prompt instruction according to the recorded text;
performing instruction adjustment on the pre-trained large model according to the target prompt instruction to obtain a first target large model;
inputting the recorded text into a first target large model to obtain a first recognition result;
and inputting the first identification result into a second target large model to obtain a second identification result, wherein the second identification result is used for indicating whether the flow information is matched with the historical flow information corresponding to the historical abnormal transaction behavior.
Optionally, the target prompt instruction includes an element prompt instruction and a flow prompt instruction, and the generating the target prompt instruction according to the recorded text includes:
obtaining target prompt words, wherein the target prompt words comprise element prompt words and flow prompt words;
generating the element prompt instruction according to the element prompt word;
and generating the flow prompt instruction according to the flow prompt word.
Optionally, after the target prompt word is obtained, the method includes:
acquiring a prompt instruction template;
the generating the element prompt instruction according to the element prompt word comprises the following steps:
filling the element prompt words into the prompt instruction template to generate the element prompt instruction;
the generating a flow prompt instruction according to the flow prompt word comprises the following steps:
and filling the flow prompt word into the prompt instruction template to generate the flow prompt instruction.
Optionally, before the instruction adjustment is performed on the pre-trained large model according to the target prompt instruction to obtain the first target large model, the method further includes:
acquiring a domain corpus, wherein the domain corpus comprises text data of historical abnormal transaction behaviors;
and performing domain adjustment on the first large model according to the domain corpus to obtain the pre-trained large model, wherein the first large model is a large language model.
Optionally, the performing instruction adjustment on the pre-trained large model according to the target prompt instruction to obtain a first target large model includes:
generating a domain prompt instruction corpus according to the element prompt instruction and the flow prompt instruction;
and carrying out instruction adjustment on the pre-trained large model based on the domain prompt instruction corpus to obtain the first target large model.
Optionally, after the first recognition result is input into the second target large model and the second recognition result is obtained, the method further includes:
acquiring data information of historical abnormal transaction behaviors, wherein the data information is used for indicating the types, flows and information of the historical abnormal transaction behaviors;
and generating an analysis result based on the first identification result, the second identification result and the data information of the historical abnormal transaction behavior, wherein the analysis result is used for indicating the development trend of the abnormal transaction behavior.
Optionally, the generating the record text based on the abnormal transaction behavior includes:
acquiring record information corresponding to abnormal transaction behaviors;
and inputting the recorded information into an optical character recognition OCR model for format conversion to obtain the recorded text.
According to the method and the device, the target prompt instruction is determined through the recorded text of the abnormal transaction behavior, the pre-trained large model is subjected to instruction adjustment according to the target prompt instruction, and the first target large model is obtained, so that the recorded text is input into the first target large model, the first recognition result is obtained, the second recognition result is finally obtained by inputting the recorded text into the first target large model, and the recognition efficiency of the abnormal transaction behavior is improved.
The embodiment of the application further provides a computer readable storage medium, on which a computer program is stored, which when executed by a processor, implements each process of the above-mentioned transaction behavior identification method embodiment, and can achieve the same technical effects, so that repetition is avoided, and no further description is provided herein. Wherein the computer readable storage medium is selected from Read-Only Memory (ROM), random access Memory (Random Access Memory, RAM), magnetic disk or optical disk.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
From the above description of the embodiments, it will be clear to those skilled in the art that the above-described embodiment method may be implemented by means of software plus a necessary general hardware platform, but of course may also be implemented by means of hardware, but in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art in the form of a software product stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk), including several instructions for causing a terminal (which may be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to perform the method described in the embodiments of the present application.
The embodiments of the present application have been described above with reference to the accompanying drawings, but the present application is not limited to the above-described embodiments, which are merely illustrative and not restrictive, and many forms may be made by those of ordinary skill in the art without departing from the spirit of the present application and the scope of the claims, which are also within the protection of the present application.

Claims (10)

1. A method of identifying transaction activity, the method comprising:
generating a record text based on the abnormal transaction behavior, wherein the record text comprises an event flow of the abnormal transaction behavior;
generating a target prompt instruction according to the recorded text;
performing instruction adjustment on the pre-trained large model according to the target prompt instruction to obtain a first target large model;
inputting the recorded text into a first target large model to obtain a first recognition result;
and inputting the first identification result into a second target large model to obtain a second identification result, wherein the second identification result is used for indicating whether the flow information is matched with the historical flow information corresponding to the historical abnormal transaction behavior.
2. The method of claim 1, wherein the target cue instructions comprise element cue instructions and flow cue instructions, the generating target cue instructions from the recorded text comprising:
obtaining target prompt words, wherein the target prompt words comprise element prompt words and flow prompt words;
generating the element prompt instruction according to the element prompt word;
and generating the flow prompt instruction according to the flow prompt word.
3. The method according to claim 2, wherein after the target prompt word is obtained, the method comprises:
acquiring a prompt instruction template;
the generating the element prompt instruction according to the element prompt word comprises the following steps:
filling the element prompt words into the prompt instruction template to generate the element prompt instruction;
the generating a flow prompt instruction according to the flow prompt word comprises the following steps:
and filling the flow prompt word into the prompt instruction template to generate the flow prompt instruction.
4. A method according to claim 3, wherein before performing instruction adjustment on the pre-trained large model according to the target hint instruction to obtain the first target large model, the method further comprises:
acquiring a domain corpus, wherein the domain corpus comprises text data of historical abnormal transaction behaviors;
and performing domain adjustment on the first large model according to the domain corpus to obtain the pre-trained large model, wherein the first large model is a large language model.
5. The method according to claim 2, wherein the performing instruction adjustment on the pre-trained large model according to the target prompt instruction to obtain a first target large model includes:
generating a domain prompt instruction corpus according to the element prompt instruction and the flow prompt instruction;
and carrying out instruction adjustment on the pre-trained large model based on the domain prompt instruction corpus to obtain the first target large model.
6. The method of claim 1, wherein after inputting the first recognition result into the second target large model to obtain a second recognition result, the method further comprises:
acquiring data information of historical abnormal transaction behaviors, wherein the data information is used for indicating the types, flows and information of the historical abnormal transaction behaviors;
and generating an analysis result based on the first identification result, the second identification result and the data information of the historical abnormal transaction behavior, wherein the analysis result is used for indicating the development trend of the abnormal transaction behavior.
7. The method of claim 1, wherein generating the recorded text based on abnormal transaction behavior comprises:
acquiring record information corresponding to abnormal transaction behaviors;
and inputting the recorded information into an optical character recognition OCR model for format conversion to obtain the recorded text.
8. A transaction behavior recognition device, comprising:
the first generation module is used for generating a record text based on the abnormal transaction behavior, wherein the record text comprises an event flow of the abnormal transaction behavior;
the second generation module is used for generating a target prompt instruction according to the recorded text;
the adjustment module is used for carrying out instruction adjustment on the pre-trained large model according to the target prompt instruction to obtain a first target large model;
the first input module is used for inputting the recorded text into a first target large model to obtain a first identification result;
the second input module is used for inputting the first identification result into a second target large model to obtain a second identification result, and the second identification result is used for indicating whether the flow information is matched with the historical flow information corresponding to the historical abnormal transaction behavior.
9. An electronic device comprising a processor, a memory and a computer program stored on the memory and executable on the processor, the computer program implementing the steps of the transaction behavior identification method according to any one of claims 1 to 7 when executed by the processor.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon a computer program which, when executed by a processor, implements the steps of the transaction behavior recognition method according to any of claims 1 to 7.
CN202311554691.5A 2023-11-21 2023-11-21 Transaction behavior identification method and device, electronic equipment and storage medium Pending CN117575601A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311554691.5A CN117575601A (en) 2023-11-21 2023-11-21 Transaction behavior identification method and device, electronic equipment and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311554691.5A CN117575601A (en) 2023-11-21 2023-11-21 Transaction behavior identification method and device, electronic equipment and storage medium

Publications (1)

Publication Number Publication Date
CN117575601A true CN117575601A (en) 2024-02-20

Family

ID=89883887

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311554691.5A Pending CN117575601A (en) 2023-11-21 2023-11-21 Transaction behavior identification method and device, electronic equipment and storage medium

Country Status (1)

Country Link
CN (1) CN117575601A (en)

Similar Documents

Publication Publication Date Title
CN108021934B (en) Method and device for recognizing multiple elements
CN110309299B (en) Communication anti-fraud method, device, computer readable medium and electronic equipment
CN111753090A (en) Document auditing method, device, equipment and medium based on RPA and AI
CN111522916A (en) Voice service quality detection method, model training method and device
CN111177307A (en) Test scheme and system based on semantic understanding similarity threshold configuration
CN112235470B (en) Incoming call client follow-up method, device and equipment based on voice recognition
CN114399379A (en) Artificial intelligence-based collection behavior recognition method, device, equipment and medium
CN112434501A (en) Work order intelligent generation method and device, electronic equipment and medium
CN113240510A (en) Abnormal user prediction method, device, equipment and storage medium
CN114549241A (en) Contract examination method, device, system and computer readable storage medium
CN117251547A (en) User question response method and device, equipment and medium thereof
CN112288584B (en) Insurance report processing method and device, computer readable medium and electronic equipment
CN113051384A (en) User portrait extraction method based on conversation and related device
CN117520503A (en) Financial customer service dialogue generation method, device, equipment and medium based on LLM model
CN117575601A (en) Transaction behavior identification method and device, electronic equipment and storage medium
CN111080444B (en) Information auditing method and device
CN114549177A (en) Insurance letter examination method, device, system and computer readable storage medium
CN113449506A (en) Data detection method, device and equipment and readable storage medium
CN117575602A (en) Transaction behavior identification method and device, electronic equipment and storage medium
CN113887214A (en) Artificial intelligence based wish presumption method and related equipment thereof
CN114254088A (en) Method for constructing automatic response model and automatic response method
CN112907360A (en) Risk assessment method, apparatus, storage medium, and program product
CN113255361B (en) Automatic voice content detection method, device, equipment and storage medium
CN117113947B (en) Form filling system, method, electronic equipment and storage medium
US20220319496A1 (en) Systems and methods for training natural language processing models in a contact center

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination