Anti-fraud implementation method and device based on risk decision classification model and medium
Through the method based on the risk decision classification model, the integration of risk control expert rules and voice text conversion technology has solved the problem of the risk control middle platform in obtaining multi-dimensional risk levels, and the accurate decision-making and disposal of user risk levels has been achieved, which has improved the wind Control ability.
Patent Information
- Application Number
- CN202411023534.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-29
- Publication Date
- 2024-12-06
AI Technical Summary
When preventing fraud risks, it is difficult for the existing risk control middle platform to effectively obtain the multi-dimensional risk level of customers, resulting in the inability to properly deal with user behavior.
Using a method based on risk decision classification model, the user's trading behavior characteristics are screened by configuring risk control expert rules and risk warning engines, combined with voice text conversion technology to identify user wish tags, integrate multi-dimensional risk tags to make decisions, and output the user's final risk level .
It has improved the anti-fraud identification capabilities, enhanced the risk identification capabilities of the risk control middle platform, effectively dealt with user behavior, and improved the accuracy and efficiency of risk management.
Smart Images

Figure CN119090604A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of computer application technology, in particular to the field of anti-fraud prevention and control, and more particularly to an anti-fraud implementation method, device, equipment and medium based on a risk decision classification model. Background Art
[0002] Risk control platform refers to the construction of a unified risk management platform in the financial field by integrating and consolidating risk management-related technologies, data and business processes to monitor, evaluate and control risks. It provides an efficient, centralized and comprehensive risk management solution for financial institutions and enterprises. Through the risk control platform, financial institutions and enterprises can improve the efficiency and accuracy of risk management, reduce risk losses, and protect the interests of institutions and users. It is widely used in financial fields such as banking, securities, insurance, and payment, and with the continuous advancement of technology and the enrichment of data, the application of risk control platform is also evolving and expanding.
[0003] At present, in order to prevent fraud risks, the risk control center collects multi-dimensional data and applies big data and intelligent AI technology to accurately identify customer intentions. However, the multi-dimensional risk labels cannot effectively obtain the customer's final risk level, resulting in the inability to effectively give the most appropriate and effective behavioral treatment to the user's final behavior. Therefore, a classification model is urgently needed to summarize and learn the risk labels of multi-dimensional inputs, and to obtain the final risk level of this user, which is a technical problem that has been waiting to be solved. Summary of the invention
[0004] To this end, the present application provides a technical solution to provide new solutions and technical methods to make up for the above-mentioned defects of the prior art, and use the risk decision classification model to input and analyze the multi-dimensional risk labels of transaction users in fraud scenarios, and finally obtain the final risk level of the user based on the multi-dimensional risk labels. The present invention provides an anti-fraud implementation method, device and storage medium based on the risk decision classification model to solve at least one of the above-mentioned technical problems.
[0005] To achieve the above objectives, in a first aspect, the present invention provides an anti-fraud implementation method based on a risk decision classification model, comprising:
[0006] Step S1: configuring risk control expert rules, wherein the risk control expert rules formulate risk control measures, and the risk control measures include risk prevention and risk response measures;
[0007] Step S2: Start screening user transactions through the risk warning engine, generate the user's behavior intention according to the user's transaction behavior characteristics, screen the user's transaction behavior characteristics, obtain corresponding risk characteristics, and convert the risk characteristics into corresponding user risk tags;
[0008] Step S3: Collect the user risk tags, which include indicators such as hit list, hit rules, risk level, and user transaction characteristics;
[0009] Step S4: Collecting user behavior intentions: Using speech-to-text conversion technology and the intention classification model to identify the user's conversation rounds or text chat messages, find relevant voices for voice matching based on the content of the conversation rounds, find keywords for keyword matching based on the content of the text chat messages, and mine the customer's intention labels;
[0010] Step S5: Integrate the multi-dimensional risk labels and behavior intention labels collected and acquired in steps S3 and S4, use the risk decision classification model to make risk decisions, output the user risk level, and perform corresponding risk disposal according to the obtained risk level through the risk control center.
[0011] Furthermore, the creation process engine obtains the business system tenant information, including:
[0012] A tenant table is created through the process engine, and the process engine is called by performing tenant registration; a process deployment table is created through the process engine, and the process deployment table includes key information in the deployment; a process definition table is created through the process engine, and the process definition table includes key information in the process.
[0013] Furthermore, the user transaction behavior in step S2 includes: investment and financial management behavior, user online ordering, user online selling and related transaction behaviors.
[0014] Furthermore, the speech-to-text conversion technology includes TTS and ASR, which analyzes the text input by the user into language units, including words, syllables or phonemes; and generates corresponding speech signals through language rules and acoustic models.
[0015] Furthermore, the generation process includes the following steps:
[0016] Step S401: input text, the text is analyzed or parsed into language units, the language units are processed by a speech synthesizer, and the pronunciation and grammatical structure of each language unit are determined by using lexical analysis, syntactic analysis and semantic analysis techniques;
[0017] Step S402: Generate a corresponding speech signal through a speech synthesizer according to the input text and the result obtained by the analysis of step S401 through synthesis rules, splicing technology or statistical model; wherein the synthesis rules use a set of rules of phonetics and linguistics knowledge, the splicing technology uses pre-recorded speech fragments to splice together to generate speech, and the statistical model learns the speech generation pattern from the speech data through a machine learning algorithm.
[0018] In a second aspect, the present invention also provides an anti-fraud implementation device based on a risk decision classification model, the device comprising: a rule configuration unit, a process designer building unit, an information import unit, an information export unit, and an information publishing unit;
[0019] A rule configuration unit 10 is used to configure risk control expert rules, wherein the risk control expert rules formulate risk control measures, and the risk control measures include risk prevention and risk response measures;
[0020] Feature conversion unit 20: used to start screening user transactions through the risk warning engine, generate the user's behavior intention according to the user's transaction behavior characteristics, screen the user's transaction behavior characteristics, obtain corresponding risk characteristics, and convert the risk characteristics into corresponding user risk tags;
[0021] Risk collection unit 30: used to collect the user risk tags, the user risk tags include: hit list, hit rules, risk level, user transaction characteristics and other indicators;
[0022] Behavior collection unit 40: used to collect user behavior intentions: through voice-to-text conversion technology, the intention classification model identifies the user's round conversation or text chat information, finds relevant voices for voice matching according to the content of the round conversation, finds keywords for keyword matching according to the content of the text chat information, and mines the customer's intention label;
[0023] Data integration unit 50: used to integrate the multi-dimensional risk labels and behavior intention labels collected by the risk collection unit 30 and the behavior collection unit 40, make risk decisions using the risk decision classification model, output the user risk level, and perform corresponding risk disposal according to the obtained risk level through the risk control center.
[0024] Furthermore, the user transaction behavior in the feature conversion unit 20 includes: investment and financial management behavior, user online ordering, user online selling and related transaction behaviors.
[0025] Furthermore, the speech-to-text conversion technology includes TTS and ASR, which analyzes the text input by the user into language units, including words, syllables or phonemes; and generates corresponding speech signals through language rules and acoustic models.
[0026] Furthermore, the generation process includes the following steps:
[0027] Step S401: input text, the text is analyzed or parsed into language units, the language units are processed by a speech synthesizer, and the pronunciation and grammatical structure of each language unit are determined by using lexical analysis, syntactic analysis and semantic analysis techniques;
[0028] Step S402: Generate a corresponding speech signal through a speech synthesizer according to the input text and the result obtained by the analysis of step S401 through synthesis rules, splicing technology or statistical model; wherein the synthesis rules use a set of rules of phonetics and linguistics knowledge, the splicing technology uses pre-recorded speech fragments to splice together to generate speech, and the statistical model learns the speech generation pattern from the speech data through a machine learning algorithm.
[0029] In a third aspect, the present invention provides a computing device, comprising:
[0030] A memory for storing program instructions;
[0031] The processor is used to call the program instructions stored in the memory and execute the above-mentioned anti-fraud implementation method based on the risk decision classification model according to the obtained program instructions.
[0032] In a fourth aspect, the present invention provides a computer-readable storage medium, including computer-readable instructions. When a computer reads and executes the computer-readable instructions, the above-mentioned anti-fraud implementation method based on a risk decision classification model is implemented. The readable medium may be multiple, and the multiple readable media can operate independently of each other.
[0033] Compared with the prior art, this solution configures risk control expert rules, and the risk warning engine starts screening user transactions, performs screening according to various behavioral characteristics, and obtains corresponding risk characteristics and behavioral characteristics. Generate user behavioral intentions, convert risk characteristics into risk labels, and collect user risk labels and behavioral intentions: through voice-to-text conversion technology, use the intention classification model to identify the customer's round conversation or text chat information, and mine the customer's intention labels. Integrate the collected multi-dimensional risk labels and intention labels, use the risk decision classification model to make risk decisions, output the user's risk level, and use the risk control middle platform to perform relevant risk disposal for the final risk level. Compared with single feature recognition, this multi-dimensional feature risk decision-making anti-fraud identification capability has obvious advantages. The application of the present invention in anti-fraud scenarios and the risk identification capability of the risk control middle platform system play an important role in supplementing the capabilities of the entire risk control middle platform. BRIEF DESCRIPTION OF THE DRAWINGS
[0034] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0035] Figure 1 A flow chart of an embodiment of the anti-fraud implementation method based on the risk decision classification model provided by the present invention;
[0036] Figure 2 A structural block diagram of an embodiment of an anti-fraud implementation device based on a risk decision classification model provided by the present invention. DETAILED DESCRIPTION
[0037] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be interpreted as limiting the present application.
[0038] It will be understood by those skilled in the art that, unless otherwise stated, the singular forms "a", "an" and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present application refers to the presence of features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. The term "and / or" used herein includes all or any unit and all combinations of one or more associated listed items.
[0039] In order to make the objectives, technical solutions and advantages of the present application more clear, the specific implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.
[0040] For the convenience of description, some nouns or terms involved in the embodiments of the present application are explained below:
[0041] Open source BERT model: BERT (Birectional Encoder Representations from Transformers) is a pre-trained model proposed by Google AI Research Institute in 2018. This model has excellent machine reading comprehension ability. It abandons traditional RNN and CNN, optimizes the word distance conversion to 1 through the Attention mechanism, effectively solves the thorny long-term dependency problem in NLP, and is widely used.
[0042] Droo ls rule open source engine: It is a rule engine based on Java pre-research and development provided by the JBOSS organization. It can free complex and changeable business rules from hard coding, and store them in files or specific storage media (such as databases) in the form of rule scripts, so that changes to business rules can take effect immediately in the production environment without modifying the project code or restarting the server.
[0043] The core of the present invention is to provide an anti-fraud implementation method based on a risk decision classification model, such as Figure 1 The figure is a flow chart of the anti-fraud implementation method based on the risk decision classification model provided in this embodiment. In order to better understand the method, Figure 1 The technical solution of this embodiment is described in detail:
[0044] Step S1: configure risk control expert rules, wherein the risk control expert rules formulate risk control measures, and the risk control measures include risk prevention and risk response measures.
[0045] In a preferred implementation, risk assessments are performed regularly to identify potential and existing risks. The sources, nature, probability, and impact of risks are determined. Clear risk control measures are developed, including risk prevention and risk response measures. Effective implementation and monitoring of risk control measures are ensured. A monitoring mechanism is established to regularly monitor risk indicators and risk events. Risk situations are reported in a timely manner to ensure that decision makers and relevant departments are aware of risk situations.
[0046] Step S2: Start screening user transactions through the risk warning engine, generate the user's behavioral intention based on the user's transaction behavior characteristics, screen the user's transaction behavior characteristics, obtain corresponding risk characteristics, and convert the risk characteristics into corresponding user risk tags.
[0047] In a preferred embodiment, the user transaction behavior in step S2 includes: investment and financial management behavior, user online ordering, user online selling and related transaction behaviors.
[0048] Step S3: Collect the user risk tags, which include indicators such as hit list, hit rules, risk level, and user transaction characteristics.
[0049] Step S4: Collect user behavior intentions: Use speech-to-text conversion technology and the intention classification model to identify the user's conversation rounds or text chat messages, find relevant voices for voice matching based on the content of the conversation rounds, find keywords for keyword matching based on the content of the text chat messages, and mine the customer's intention tags.
[0050] In a preferred embodiment, the speech-to-text conversion technology includes TTS and ASR, which analyzes the text input by the user into language units, including words, syllables or phonemes; and generates corresponding speech signals through language rules and acoustic models.
[0051] Step S5: Integrate the multi-dimensional risk labels and behavior intention labels collected and acquired in steps S3 and S4, use the risk decision classification model to make risk decisions, output the user risk level, and perform corresponding risk disposal according to the obtained risk level through the risk control center.
[0052] In a preferred implementation, the risk control center performs relevant risk treatment for the final risk level. For example, if it is judged to be high risk, transaction blocking and account control can be performed. If it is judged to be medium risk, manual risk screening can be performed and the manual verification process can be followed. If it is judged to be low risk, the transaction can be directly released to complete normal user transactions.
[0053] In one embodiment, the generation process further comprises the following steps:
[0054] Step S401: input text, the text is analyzed or parsed into language units, the language units are processed by a speech synthesizer, and the pronunciation and grammatical structure of each language unit are determined by using lexical analysis, syntactic analysis and semantic analysis techniques;
[0055] Step S402: Generate a corresponding speech signal through a speech synthesizer according to the input text and the result obtained by the analysis of step S401 through synthesis rules, splicing technology or statistical model; wherein the synthesis rules use a set of rules of phonetics and linguistics knowledge, the splicing technology uses pre-recorded speech fragments to splice together to generate speech, and the statistical model learns the speech generation pattern from the speech data through a machine learning algorithm.
[0056] In this embodiment, by configuring the risk control expert rules, the risk warning engine starts screening user transactions, screens according to various behavioral characteristics, and obtains the corresponding risk characteristics and behavioral characteristics. Generate the user's behavioral intention, convert the risk characteristics into risk labels, and collect user risk labels and behavioral intentions: through voice-to-text conversion technology, use the intention classification model to identify the customer's round conversation or text chat information, and mine the customer's intention labels. Integrate the collected multi-dimensional risk labels and intention labels, use the risk decision classification model to make risk decisions, output the user's risk level, and use the risk control middle station to perform relevant risk disposal for the final risk level. Compared with single feature recognition, this multi-dimensional feature risk decision-making anti-fraud identification capability has obvious advantages. The application of the present invention in anti-fraud scenarios and the risk identification capability of the risk control middle station system play an important role in supplementing the capabilities of the entire risk control middle station.
[0057] In this embodiment, a "self-supervision" method is implemented through the risk decision classification model and the BERT pre-training mechanism. The BERT model is trained using a small amount of labeled data in the early stage to achieve 85% accurate risk identification. After going online, the multi-dimensional data features of real customers are used for labeling, and the model is repeatedly trained. After multiple iterations, the accuracy of the risk decision classification model reaches 95%. It can be seen that this solution has a high accuracy.
[0058] In one embodiment, the present application provides an anti-fraud implementation device based on a risk decision classification model, see Figure 2 The figure shows a structural block diagram of an anti-fraud implementation device based on a risk decision classification model provided in this embodiment, wherein the device comprises: a rule configuration unit 10, a process designer building unit 20, an information import unit 30, an information export unit 40, and an information publishing unit 50.
[0059] The engine creation unit 10 is used to configure risk control expert rules, wherein the risk control expert rules formulate risk control measures, and the risk control measures include risk prevention and risk response measures.
[0060] In a preferred embodiment, the engine creation unit 10 can regularly conduct risk assessments to identify potential risks and existing risks. Determine the source, nature, probability and impact of risks. Formulate clear risk control measures, including risk prevention and risk response measures. Ensure the effective implementation and monitoring of risk control measures. Establish a monitoring mechanism to regularly monitor risk indicators and risk events. Report risk situations in a timely manner to ensure that decision makers and relevant departments are aware of risk situations.
[0061] Feature conversion unit 20: used to start screening user transactions through the risk warning engine, generate the user's behavior intention according to the user's transaction behavior characteristics, screen the user's transaction behavior characteristics, obtain corresponding risk characteristics, and convert the risk characteristics into corresponding user risk tags.
[0062] In a preferred embodiment, the user transaction behavior in the feature conversion unit 20 includes: investment and financial management behavior, user online ordering, user online selling and related transaction behaviors.
[0063] The risk collection unit 30 is used to collect the user risk tags, which include indicators such as the hit list, hit rules, risk level, and user transaction characteristics.
[0064] Behavior collection unit 40: used to collect user behavior intentions: through speech-to-text conversion technology, identify the user's turn conversation or text chat information through the intention classification model, find relevant voices for voice matching according to the content of the round conversation, find keywords for keyword matching according to the content of the text chat information, and mine the customer's intention tags.
[0065] In a preferred embodiment, the speech-to-text conversion technology includes TTS and ASR, which analyzes the text input by the user into language units, including words, syllables or phonemes; and generates corresponding speech signals through language rules and acoustic models.
[0066] Data integration unit 50: used to integrate the multi-dimensional risk labels and behavior intention labels collected by the risk collection unit 30 and the behavior collection unit 40, make risk decisions using the risk decision classification model, output the user risk level, and perform corresponding risk disposal according to the obtained risk level through the risk control center.
[0067] In a preferred implementation, the risk control center performs relevant risk treatment for the final risk level. For example, if it is judged to be high risk, transaction blocking and account control can be performed. If it is judged to be medium risk, manual risk screening can be performed and the manual verification process can be followed. If it is judged to be low risk, the transaction can be directly released to complete normal user transactions.
[0068] In this embodiment, the generation process also has the following functions:
[0069] Parsing unit 401: Input text, the text is analyzed or parsed into language units, the language units are processed by a speech synthesizer, and the pronunciation and grammatical structure of each language unit are determined using lexical analysis, syntactic analysis and semantic analysis techniques.
[0070] Speech generation unit 402: generates a corresponding speech signal through a speech synthesizer according to the input text and the analysis result obtained by the parsing unit 401 through synthesis rules, splicing technology or statistical model; wherein the synthesis rules use a set of rules of phonetics and linguistics knowledge, the splicing technology uses pre-recorded speech fragments to splice together to generate speech, and the statistical model learns the speech generation pattern from the speech data through a machine learning algorithm.
[0071] In this embodiment, by configuring the risk control expert rules, the risk warning engine starts to screen user transactions, screens according to various behavioral characteristics, and obtains the corresponding risk characteristics and behavioral characteristics. Generate the user's behavioral intention, convert the risk characteristics into risk labels, and collect user risk labels and behavioral intentions: through voice-to-text conversion technology, use the intention classification model to identify the customer's round conversation or text chat information, and mine the customer's intention label. Integrate the collected multi-dimensional risk labels and intention labels, use the risk decision classification model to make risk decisions, output the user's risk level, and use the risk control middle platform to perform relevant risk disposal for the final risk level. Use the risk decision classification model to input, analyze and decide on the multi-dimensional risk labels of transaction users in fraud scenarios, and finally obtain the user's final risk level (high risk, medium risk, low risk) based on the multi-dimensional risk labels, and use the risk level to deal with the user's behavior accordingly in risk disposal.
[0072] The technical solution provided by this embodiment solves the problem that effective decision-making cannot be made based on the multi-dimensional risk characteristics of users. The multi-dimensional risk labels include user willingness, transaction purpose, transaction counterparty, tone of voice, whether it belongs to a certain risk list, whether it hits the relevant risk control rules, etc. By aggregating the data of these labels and using the classification decision algorithm, the final risk level of the user can be decided.
[0073] The anti-fraud implementation device based on the risk decision classification model of this embodiment is used to implement the aforementioned anti-fraud implementation method based on the risk decision classification model. Therefore, the specific implementation method of the anti-fraud implementation device based on the risk decision classification model can be seen in the specific implementation method of the anti-fraud implementation method based on the risk decision classification model in the previous text. For example, the rule configuration unit 10 is used to implement step S1 in the above-mentioned anti-fraud implementation method based on the risk decision classification model, the feature conversion unit 20 is used to implement step S2 in the above-mentioned anti-fraud implementation method based on the risk decision classification model, the risk collection unit 30 is used to implement step S3 in the above-mentioned anti-fraud implementation method based on the risk decision classification model, the behavior collection unit 40 is used to implement step S4 in the above-mentioned anti-fraud implementation method based on the risk decision classification model, and the data integration unit 50 is used to implement step S5 in the above-mentioned anti-fraud implementation method based on the risk decision classification model. Therefore, its specific implementation method can refer to the description of the corresponding various parts of the embodiments. The device provided by the present invention is attached to the method provided by the present invention, and its specific effects can refer to the above-mentioned method, which will not be repeated here.
[0074] In one embodiment, a computing device is provided, comprising:
[0075] A memory for storing program instructions;
[0076] A processor is used to implement the steps of the anti-fraud implementation method based on the risk decision classification model as described in any of the above when executing the computer program. The anti-fraud implementation method based on the risk decision classification model of the present invention has obvious advantages and beneficial effects compared with the prior art. By configuring the risk control expert rules, the risk warning engine starts to screen user transactions, screens according to various behavioral characteristics, and obtains the corresponding risk characteristics and behavioral characteristics. Generate the user's behavioral intention, convert the risk characteristics into risk labels, collect user risk labels and behavioral intentions: through voice-to-text conversion technology, use the intention classification model to identify the customer's round conversation or text chat information, and mine the customer's intention label. Integrate the collected multi-dimensional risk labels and intention labels, use the risk decision classification model to make risk decisions, output the user's risk level, and use the risk control middle station to perform relevant risk disposal for the final risk level. Use the risk decision classification model to input, analyze and decide on the multi-dimensional risk labels of the transaction users in the fraud scenario, and finally obtain the user's final risk level (high risk, medium risk, low risk) according to the multi-dimensional risk labels, and use the risk level to deal with the user's behavior accordingly in risk disposal.
[0077] The technical solution provided by this embodiment solves the problem that effective decision-making cannot be made based on the multi-dimensional risk characteristics of users. The multi-dimensional risk labels include user willingness, transaction purpose, transaction counterparty, tone of voice, whether it belongs to a certain risk list, whether it hits the relevant risk control rules, etc. By aggregating the data of these labels and using the classification decision algorithm, the final risk level of the user can be decided.
[0078] In one embodiment, a computer-readable storage medium is provided, including computer-readable instructions. When a computer reads and executes the computer-readable instructions, the steps of the anti-fraud implementation method based on the risk decision classification model described above are implemented. The anti-fraud implementation method based on the risk decision classification model of the present invention has obvious advantages and beneficial effects compared with the prior art. By configuring the risk control expert rules, the risk warning engine starts to screen user transactions, screens according to various behavioral characteristics, and obtains corresponding risk characteristics and behavioral characteristics. Generate the user's behavioral willingness, convert the risk characteristics into risk labels, collect user risk labels and behavioral willingness: through voice-to-text conversion technology, use the willingness classification model to identify the customer's round conversation or text chat information, and mine the customer's willingness label. Integrate the collected multi-dimensional risk labels and willingness labels, use the risk decision classification model to make risk decisions, output the user's risk level, and use the risk control middle station to perform relevant risk disposal for the final risk level. Use the risk decision classification model to input, analyze and decide on the multi-dimensional risk labels of transaction users in the fraud scenario, and finally obtain the user's final risk level (high risk, medium risk, low risk) according to the multi-dimensional risk labels, and use the risk level to deal with the user's behavior accordingly in risk disposal.
[0079] The technical solution provided by this embodiment solves the problem that effective decision-making cannot be made based on the multi-dimensional risk characteristics of users. The multi-dimensional risk labels include user willingness, transaction purpose, transaction counterparty, tone of voice, whether it belongs to a certain risk list, whether it hits the relevant risk control rules, etc. By aggregating the data of these labels and using the classification decision algorithm, the final risk level of the user can be decided.
[0080] Through the description of the above implementation methods, technical personnel in the relevant field can understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the specific device can be divided into different functional modules to complete all or part of the functions described above.
[0081] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.
[0082] In several embodiments covered by the present application, it should be understood that the disclosed devices, equipment and methods can be implemented in other ways. For example, the division of modules is only a logical function division, and there may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another device, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or modules, or it can be an electrical, mechanical or other form of connection.
[0083] The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the embodiments of the present application.
[0084] In addition, each functional module in each embodiment of the present application can be implemented in the form of hardware or in the form of software functional modules. If these functional modules are implemented in the form of software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present application is essentially or part of the prior art that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a computer program product, which includes one or more computer instructions. When loading and executing computer program instructions on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer program product is stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, a computer instruction can be transmitted from a website site, a computer, a server or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or a data center that includes one or more available media integrations. The available media may be magnetic media (eg, floppy disks, hard disks, magnetic tapes), optical media (eg, DVDs), or semiconductor media (eg, solid state drives (SSDs)).
[0085] The above contents are merely examples and explanations of the structure of the present invention. The technicians in this technical field may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the structure of the invention or exceed the scope defined by the claims, they should all fall within the protection scope of the present invention.
Claims
1. An anti-fraud implementation method based on a risk decision classification model, characterized in that: include: Step S1: configuring risk control expert rules, wherein the risk control expert rules formulate risk control measures, and the risk control measures include risk prevention and risk response measures; Step S2: Start screening user transactions through the risk warning engine, generate the user's behavior intention according to the user's transaction behavior characteristics, screen the user's transaction behavior characteristics, obtain corresponding risk characteristics, and convert the risk characteristics into corresponding user risk tags; Step S3: Collect the user risk tags, which include indicators such as hit list, hit rules, risk level, and user transaction characteristics; Step S4: Collecting user behavior intentions: Using speech-to-text conversion technology and the intention classification model to identify the user's conversation rounds or text chat messages, find relevant voices for voice matching based on the content of the conversation rounds, find keywords for keyword matching based on the content of the text chat messages, and mine the customer's intention labels; Step S5: Integrate the multi-dimensional risk labels and behavior intention labels collected and acquired in steps S3 and S4, use the risk decision classification model to make risk decisions, output the user risk level, and perform corresponding risk disposal according to the obtained risk level through the risk control center.
2. The anti-fraud implementation method based on risk decision classification model according to claim 1 is characterized in that: The user transaction behaviors in step S2 include: investment and financial management behaviors, user online ordering, user online selling and related transaction behaviors.
3. The anti-fraud implementation method based on risk decision classification model according to claim 1 is characterized in that: The speech-to-text conversion technology includes TTS and ASR, which analyzes the text input by the user into language units, including words, syllables or phonemes; and generates corresponding speech signals through language rules and acoustic models.
4. The anti-fraud implementation method based on risk decision classification model according to claim 3 is characterized in that: The generation process includes the following steps: Step S401: input text, the text is analyzed or parsed into language units, the language units are processed by a speech synthesizer, and the pronunciation and grammatical structure of each language unit are determined by using lexical analysis, syntactic analysis and semantic analysis techniques; Step S402: Generate a corresponding speech signal through a speech synthesizer according to the input text and the result obtained by the analysis of step S401 through synthesis rules, splicing technology or statistical model; wherein the synthesis rules use a set of rules of phonetics and linguistics knowledge, the splicing technology uses pre-recorded speech fragments to splice together to generate speech, and the statistical model learns the speech generation pattern from the speech data through a machine learning algorithm. 5.An anti-fraud implementation device based on risk decision classification model, characterized in that: include: Rule configuration unit, process designer building unit, information import unit, information export unit, information publishing unit; A rule configuration unit 10 is used to configure risk control expert rules, wherein the risk control expert rules formulate risk control measures, and the risk control measures include risk prevention and risk response measures; Feature conversion unit 20: used to start screening user transactions through the risk warning engine, generate the user's behavior intention according to the user's transaction behavior characteristics, screen the user's transaction behavior characteristics, obtain corresponding risk characteristics, and convert the risk characteristics into corresponding user risk tags; Risk collection unit 30: used to collect the user risk tags, the user risk tags include: hit list, hit rules, risk level, user transaction characteristics and other indicators; Behavior collection unit 40: used to collect user behavior intentions: through voice-to-text conversion technology, the intention classification model identifies the user's round conversation or text chat information, finds relevant voices for voice matching according to the content of the round conversation, finds keywords for keyword matching according to the content of the text chat information, and mines the customer's intention label; Data integration unit 50: used to integrate the multi-dimensional risk labels and behavior intention labels collected by the risk collection unit 30 and the behavior collection unit 40, make risk decisions using the risk decision classification model, output the user risk level, and perform corresponding risk disposal according to the obtained risk level through the risk control center.
6. The anti-fraud implementation device based on risk decision classification model according to claim 5 is characterized in that: The user transaction behaviors in the feature conversion unit 20 include: investment and financial management behaviors, user online ordering, user online selling and related transaction behaviors.
7. The anti-fraud implementation device based on risk decision classification model according to claim 5 is characterized in that: The speech-to-text conversion technology includes TTS and ASR, which analyzes the text input by the user into language units, including words, syllables or phonemes; and generates corresponding speech signals through language rules and acoustic models.
8. The anti-fraud implementation device based on risk decision classification model according to claim 7 is characterized in that: The generation process includes the following steps: Step S401: input text, the text is analyzed or parsed into language units, the language units are processed by a speech synthesizer, and the pronunciation and grammatical structure of each language unit are determined by using lexical analysis, syntactic analysis and semantic analysis techniques; Step S402: Generate a corresponding speech signal through a speech synthesizer according to the input text and the result obtained by the analysis of step S401 through synthesis rules, splicing technology or statistical model; wherein the synthesis rules use a set of rules of phonetics and linguistics knowledge, the splicing technology uses pre-recorded speech fragments to splice together to generate speech, and the statistical model learns the speech generation pattern from the speech data through a machine learning algorithm.
9. A computing device, characterized in that include: A memory for storing program instructions; A processor is used to call the program instructions stored in the memory, and execute the method according to any one of claims 1 to 4 according to the obtained program instructions.
10. A computer-readable storage medium, characterized in that: It includes computer-readable instructions. When a computer reads and executes the computer-readable instructions, it implements the method as claimed in any one of claims 1 to 4. There can be multiple readable media, and the multiple readable media can operate independently of each other.
