Object identification method, device, equipment and readable storage medium, program product

Abstract

This application provides an object recognition method, apparatus, device, readable storage medium, and program product. Embodiments of this application can be applied to various scenarios such as cloud technology, artificial intelligence, smart transportation, and vehicle-mounted systems, and involve big data technology and blockchain technology. The method includes: acquiring object operation data, object resource data, and historical resource transfer data of an object to be identified; predicting the degree of anomaly of the object to be identified based on the object operation data, object resource data, and historical resource transfer data, and obtaining a prediction result; wherein, the prediction result characterizes the probability of anomalies in the resource transfer of the object to be identified, and the probability of anomalies in the usage status of the object to be identified; and identifying the target object from the objects to be identified based on the prediction result. This application can improve the intelligence level of object recognition.

Description

Technical Field

[0001] This application relates to big data technology, and in particular to an object recognition method, apparatus, device, readable storage medium, and program product. Background Technology

[0002] Target objects exhibiting abnormal behavior can pose a significant threat to resource transfer scenarios and require focused management. Before management, it's crucial to identify these abnormal targets. Most related technologies rely on manually verifying information about different objects to determine if they are indeed targets. However, manually verifying information to identify target objects suffers from low accuracy and efficiency, ultimately resulting in limited intelligence in object recognition. Summary of the Invention

[0003] This application provides an object recognition method, apparatus, device, readable storage medium, and program product, which can improve the intelligence level of object recognition.

[0004] The technical solution of this application embodiment is implemented as follows:

[0005] This application provides an object recognition method, including:

[0006] Acquire object operation data, object resource data, and historical resource transfer data of the object to be identified;

[0007] Based on the object operation data, the object resource data, and the historical resource transfer data, the degree of anomaly of the object to be identified is predicted, and a prediction result is obtained.

[0008] The prediction results represent the probability that the resource transfer of the object to be identified is abnormal, and the probability that the usage status of the object to be identified is abnormal.

[0009] Based on the prediction results, the target object is identified from the objects to be identified.

[0010] This application provides an object recognition device, including:

[0011] The data acquisition module is used to acquire object operation data, object resource data, and historical resource transfer data of the object to be identified.

[0012] An anomaly prediction module is used to predict the degree of anomaly of the object to be identified based on the object operation data, the object resource data, and the historical resource transfer data, and to obtain a prediction result; wherein, the prediction result characterizes the probability that the resource transfer of the object to be identified is abnormal, and the probability that the usage status of the object to be identified is abnormal.

[0013] The object recognition module is used to identify the target object from the objects to be identified based on the prediction results.

[0014] In some embodiments of this application, the anomaly prediction module is further configured to: predict the probability of resource transfer anomalies of the object to be identified based on the object resource data and the historical resource transfer data, and obtain a first probability value; predict the probability of usage status anomalies of the object to be identified based on the object operation data, and obtain a second probability value; and determine the prediction result of the object to be identified based on the first probability value and the second probability value.

[0015] In some embodiments of this application, the anomaly prediction module is further configured to determine the current transfer anomaly stage of the object to be identified from the historical resource transfer data; and to predict the first probability value of the object to be identified switching from the current transfer anomaly stage to the next transfer anomaly stage using the object resource data; wherein the object resource data includes at least resource harvesting information and resource expenditure information, and the historical resource transfer data includes at least resource transfer behavior history and resource acquisition behavior history.

[0016] In some embodiments of this application, the anomaly prediction module is further configured to: determine that the first prediction result of the object to be identified is switching from the current transfer anomaly stage to the next transfer anomaly stage when the first probability value is greater than the first probability threshold; filter out the second prediction result of the object to be identified from the plurality of state anomaly stages based on the second probability value and the second probability threshold of each of the plurality of state anomaly stages of the usage state; and integrate the first prediction result and the second prediction result into the prediction result of the object to be identified.

[0017] In some embodiments of this application, the object recognition module is further configured to, based on the first prediction result and the second prediction result in the prediction results, filter out the matching anomaly level corresponding to the object to be identified from multiple anomaly levels; and determine the object whose matching anomaly level among the objects to be identified reaches the preset anomaly level as the target object.

[0018] In some embodiments of this application, the object recognition device further includes: an information repair module; the information repair module is used to identify a target object from the objects to be identified based on the prediction result, and then determine an information repair method for the target object based on the target anomaly level corresponding to the target object; and use the information repair method to repair the information of the target object to obtain target repair information.

[0019] In some embodiments of this application, the information repair method includes: information repair based on identity identifier; the information repair module is further configured to: mine associated objects of the target object from a preset database based on the identity identifier of the target object; statistically analyze the operation frequency of the associated objects to obtain the first activity level of the associated objects; calculate the similarity between the associated objects and the target object based on the attribute information of the target object and the attribute information of the associated objects; and determine the object information of the N objects with the highest first activity level determined based on the first activity level of the associated objects, and the object information of the N objects with the highest similarity level determined based on the similarity with the target object, as the target repair information, thereby realizing information repair based on the identity identifier; N is a positive integer.

[0020] In some embodiments of this application, the information repair method includes: information repair based on login device identifier; the information repair module is further configured to obtain the login device identifier corresponding to the target object; extract objects with the same login device as the target object from a preset database based on the login device identifier; count the operation frequency of the objects with the same login device to obtain the second activity level of the objects with the same login device; and determine the object information of the N objects with the highest second activity level as the target repair information, thereby realizing the information repair based on login device identifier.

[0021] In some embodiments of this application, the information repair method includes: information repair based on interactive objects; the information repair module is further configured to find interactive objects that have interactive behavior with the target object from a preset database; perform frequency statistics on the interactive behavior of the interactive objects and the target object to obtain the interaction frequency between the interactive objects and the target object; and determine the object information of the N objects with the highest interaction frequency with the target object, as determined according to the interaction frequency, as the target repair information, thereby realizing the information repair based on interactive objects.

[0022] In some embodiments of this application, the information repair method includes: information repair based on social relationships; the information repair module is further configured to: mine candidate objects that have social relationships with the target object from the obtained identity relationship data of the target object; calculate the intimacy between the candidate object and the target object based on the number of social interactions between the target object and the candidate object; statistically analyze the operation frequency of the candidate object to obtain the third activity level of the candidate object; and determine the object information of the N objects with the highest intimacy with the target object determined based on the intimacy between the candidate object and the target object, and the object information of the N objects with the highest third activity level determined based on the third activity level, as the target repair information, thereby realizing the information repair based on social relationships.

[0023] In some embodiments of this application, the object identification device further includes: a prompting module; the prompting module is used to repair the target object using the information repair method, and after obtaining target repair information, to issue a prompt message to the target object indicating that a resource transfer anomaly has occurred.

[0024] This application provides an object recognition device, including:

[0025] Memory, used to store executable instructions;

[0026] The processor, when executing executable instructions stored in the memory, implements the object recognition method provided in the embodiments of this application.

[0027] This application provides a computer-readable storage medium storing executable instructions for inducing a processor to execute and implement the object identification method provided in this application.

[0028] This application provides a computer program product, including a computer program or instructions, which, when executed by a processor, implement the object recognition method provided in this application.

[0029] The embodiments of this application have the following beneficial effects: the object recognition device can first acquire data of the object to be recognized in various dimensions, and then use the data in various dimensions to predict the probability of the object to be recognized experiencing resource transfer anomalies and usage status anomalies. Finally, based on the prediction results, the target object can be more accurately identified from the objects to be recognized. Thus, during object recognition, the data of the object to be recognized in various dimensions is fully utilized, improving the accuracy of object recognition. Furthermore, during the recognition process, there is no need to manually confirm the object to be recognized, thereby reducing the lag during recognition, improving the efficiency of object recognition, and ultimately improving the intelligence level of object recognition. Attached Figure Description

[0030] Figure 1A This is a schematic diagram of the architecture of the object recognition system provided in this application embodiment;

[0031] Figure 1B This is a schematic diagram of the architecture of the object recognition system provided in the embodiments of this application. Figure 2 ;

[0032] Figure 2 This is a schematic diagram of the structure of the object recognition device provided in the embodiments of this application;

[0033] Figure 3 This is a flowchart illustrating the object recognition method provided in this application embodiment;

[0034] Figure 4 This is a flowchart illustrating the object recognition method provided in the embodiments of this application. Figure 2 ;

[0035] Figure 5 This is a schematic diagram illustrating the process of predicting the first probability value provided in an embodiment of this application;

[0036] Figure 6 This is a flowchart illustrating the object recognition method provided in the embodiments of this application. Figure 3 ;

[0037] Figure 7 This is a schematic diagram of the system framework for identifying and recovering from loss of contact provided in the embodiments of this application;

[0038] Figure 8 This is a schematic diagram of multi-layer disconnection repair provided in an embodiment of this application;

[0039] Figure 9 This is a schematic diagram of the process for identifying and recovering from loss of contact provided in the embodiments of this application. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0041] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0042] In the following description, the terms "first, second, third" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first, second, third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0043] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0044] In the implementation of this application, the collection and processing of relevant data should strictly comply with the requirements of relevant laws and regulations, obtain the informed consent or separate consent of the personal information subject, and carry out subsequent data use and processing within the scope of laws and regulations and the authorization of the personal information subject.

[0045] Before providing a further detailed description of the embodiments of this application, the nouns and terms involved in the embodiments of this application will be explained, and the nouns and terms involved in the embodiments of this application shall be interpreted as follows.

[0046] 1) Cloud computing refers to the delivery and usage model of IT infrastructure, meaning obtaining necessary resources in an on-demand and easily scalable manner through a network. In a broader sense, cloud computing services refer to obtaining necessary services in an on-demand and easily scalable manner through a network. These services can be IT and software related, internet-related, or other services. Cloud computing is a product of the development and integration of traditional computer and network technologies such as grid computing, distributed computing, parallel computing, utility computing, network storage technologies, virtualization, and load balancing.

[0047] 2) Big Data refers to data sets that cannot be captured, managed, and processed by conventional software tools within a certain timeframe. It represents massive, rapidly growing, and diverse information assets that require new processing models to achieve stronger decision-making, insight discovery, and process optimization capabilities. With the advent of the cloud era, Big Data has attracted increasing attention. Big Data requires specialized technologies to effectively process large amounts of data within a tolerable timeframe. Technologies suitable for Big Data include massively parallel processing databases, data mining, distributed file systems, distributed databases, cloud computing platforms, and internet-scalable storage systems.

[0048] 3) Artificial Intelligence (AI) is the theory, methods, technology, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a comprehensive technology within computer science that attempts to understand the essence of intelligence and produce a new kind of intelligent machine that can react in a way similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to possess the functions of perception, reasoning, and decision-making.

[0049] Artificial intelligence (AI) is a comprehensive discipline encompassing a wide range of fields, including both hardware and software technologies. Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies primarily include computer vision, speech processing, natural language processing, as well as machine learning / deep learning, autonomous driving, and intelligent transportation.

[0050] 4) Machine Learning (ML) is an interdisciplinary field that studies multiple disciplines, including probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how computers can simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to endow computers with intelligence; its applications span all areas of artificial intelligence. Machine learning and deep learning typically include techniques such as artificial neural networks, belief learning, reinforcement learning, transfer learning, and inductive learning.

[0051] 5) The target object refers to the object that has an abnormal phenomenon. This abnormal behavior can be due to an abnormality in the use state, such as the user losing contact, or an abnormality in the transfer of resources, such as the failure to repay a loan on time.

[0052] 6) Information repair refers to the method of retrieving valid information about the target object from information about objects related to the target object. For example, retrieving valid contact information of the target object from information about objects related to the target object.

[0053] Target objects exhibiting abnormal behavior can pose a significant threat to resource transfer scenarios and require focused management. Before management, it is essential to first identify the target objects exhibiting abnormal behavior.

[0054] In most related technologies, the identification of a target is manually confirmed based on information stored by the object. For example, multiple manual contacts (multiple phone calls, text messages, etc.) are made using the stored contact information, and the feedback is analyzed to determine whether the object is the target.

[0055] However, determining whether an object is a target based on stored object information is often susceptible to interference from unexpected situations. For example, a user might miss a phone call due to a meeting or other reasons, leading to the misidentification of that object as a target. Furthermore, relying solely on object information to determine target status only uses data from the feedback dimension, without analyzing other dimensions, resulting in low accuracy in object recognition.

[0056] Furthermore, when determining whether an object is the target object based on the stored object information, multiple manual confirmations are required to arrive at a judgment result, indicating a certain lag in recognition and thus resulting in low efficiency in object recognition.

[0057] In summary, the related technologies that identify target objects by using the object information corresponding to the object suffer from low object recognition accuracy and low recognition efficiency, ultimately resulting in a low level of intelligence in object recognition.

[0058] Furthermore, after identifying the target object, information repair is also necessary to promptly alert the target object to any anomalies. Related technologies primarily involve extracting relationships between entities from the target data, constructing an entity relationship graph based on these relationships, and then identifying relationships between multiple accounts, family and friends, and devices. Based on the entity relationship graph and the corresponding correlations, the closeness of the target object to other entities is determined. Information repair is then performed on the target object based on the degree of closeness, allowing for the re-establishment of valid object information by contacting friends, relatives, and other related individuals.

[0059] However, this method indiscriminately repairs information about the target object, simultaneously extracting information from multiple users with the same identity, family members, colleagues, etc. While this method can achieve broader information repair for the target object, it includes information from many users whose connection to the target object is not very strong. Therefore, contacting all these users would undoubtedly increase the cost of information repair and could potentially cause inconvenience to these less relevant users. Thus, it is evident that related technologies are less effective at information repair targeting specific individuals.

[0060] This application provides an object recognition method, apparatus, device, readable storage medium, and program product, which can improve the intelligence level of object recognition. The following describes exemplary applications of the object recognition device provided in this application. The object recognition device provided in this application can be implemented as various types of terminals such as laptops, tablets, desktop computers, set-top boxes, mobile devices, and vehicle terminals. It can also be implemented as a server, or as a device cluster composed of terminals and servers. The following will describe exemplary applications when the object recognition device is implemented as a server.

[0061] See Figure 1A , Figure 1A This is a schematic diagram of the architecture of the object recognition system provided in this application embodiment. To support an object recognition application, in the object recognition system 100 of Figure 1, terminal 400-1 is connected to server 200 through network 300, and big data storage platform 400-2 is connected to server 200 through network 300. Network 300 can be a wide area network or a local area network, or a combination of both.

[0062] Terminal 400-1 is used to record object operation data of the object to be identified based on the user's operations on the graphical interface. Big data storage platform 400-2 is used to store the object operation data of the object to be identified sent by terminal 400-1, as well as object resource data and historical resource transfer data generated from the object to be identified.

[0063] Server 200 is used to obtain object operation data of the object to be identified from terminal 400-1, and object resource data and historical resource transfer data from big data storage platform 400-2; based on the object operation data, object resource data and historical resource transfer data, the server predicts the degree of abnormality of the object to be identified and obtains the prediction result, wherein the prediction result represents the probability of abnormal resource transfer of the object to be identified and the probability of abnormal usage status of the object to be identified; based on the prediction result, the server identifies the target object from the objects to be identified.

[0064] In some embodiments, server 200 may be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. Terminal 400-1 may be a smartphone, tablet computer, laptop computer, desktop computer, smart speaker, smartwatch, smart home appliance, vehicle terminal, etc., but is not limited thereto. The terminal and server can be directly or indirectly connected through wired or wireless communication, which is not limited in this embodiment of the invention.

[0065] This application's embodiments can also be implemented using blockchain technology. See [link to relevant documentation]. Figure 1B , Figure 1B This is a schematic diagram of the architecture of the object recognition system provided in the embodiments of this application. Figure 2 . Figure 1B In the object recognition system 100, the target object, such as an account that may be overdue and out of contact, is put on the blockchain through the server to ensure that the target object can be recorded and that the record cannot be tampered with.

[0066] In some embodiments, server 200 can join blockchain network 700 and become a node within it. The type of blockchain network 700 is flexible and diverse; for example, it can be any of a public blockchain, private blockchain, or consortium blockchain. Taking a public blockchain as an example, any electronic device of any business entity can access blockchain network 700 without authorization to act as a consensus node in blockchain network 700. For example, terminal 600 can be mapped to consensus node 700-1 in blockchain network 700 to achieve consensus when adding a target object to the blockchain.

[0067] Taking blockchain network 700 as a consortium blockchain as an example, terminal 600 can connect to blockchain network 700 and become a node after obtaining authorization. Server 200 obtains object operation data, object resource data, and historical resource transfer data of the object to be identified. Based on the object operation data, object resource data, and historical resource transfer data, it predicts the degree of anomaly of the object to be identified, obtains the prediction result, identifies the target object from the objects to be identified, and sends the target object to blockchain network 700 for consensus. When consensus is passed, blockchain network 700 allocates a new block for the target object, writes the target object into the new block, and adds the new block to the end of the blockchain to complete the on-chaining of the target object.

[0068] See Figure 2 , Figure 2 This is a schematic diagram of the structure of the object recognition device provided in the embodiments of this application. Figure 2The object identification device 500 shown includes at least one processor 510, a memory 550, at least one network interface 520, and a user interface 530. The various components in the object identification device 500 are coupled together via a bus system 540. It is understood that the bus system 540 is used to implement communication between these components. In addition to a data bus, the bus system 540 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in… Figure 2 The general labeled all buses as Bus System 540.

[0069] The processor 510 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0070] User interface 530 includes one or more output devices 531 that enable the presentation of media content, including one or more speakers and / or one or more visual displays. User interface 530 also includes one or more input devices 532, including user interface components that facilitate user input, such as a keyboard, mouse, microphone, touch screen display, camera, other input buttons and controls.

[0071] The memory 550 may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid-state storage, hard disk drives, optical disk drives, etc. The memory 550 may optionally include one or more storage devices physically located away from the processor 510.

[0072] The memory 550 may include volatile memory or non-volatile memory, or both. The non-volatile memory may be read-only memory (ROM), and the volatile memory may be random access memory (RAM). The memory 550 described in this application embodiment is intended to include any suitable type of memory.

[0073] In some embodiments, memory 550 is capable of storing data to support various operations, examples of which include programs, modules, and data structures or subsets or supersets thereof, as illustrated below.

[0074] Operating system 551 includes system programs for handling various basic system services and performing hardware-related tasks, such as the framework layer, core library layer, driver layer, etc., for implementing various basic business functions and handling hardware-based tasks;

[0075] The network communication module 552 is used to reach other computing devices via one or more (wired or wireless) network interfaces 520, exemplary network interfaces 520 including: Bluetooth, Wi-Fi, and Universal Serial Bus (USB), etc.

[0076] Presentation module 553 is used to enable the presentation of information (e.g., user interface for operating peripheral devices and displaying content and information) via one or more output devices 531 (e.g., display screen, speaker, etc.) associated with user interface 530.

[0077] The input processing module 554 is used to detect and translate one or more user inputs or interactions from one or more input devices 532.

[0078] In some embodiments, the object recognition device provided in this application can be implemented in software. Figure 2 An object recognition device 555 stored in memory 550 is shown. This device can be software in the form of programs and plug-ins, and includes the following software modules: a data acquisition module 5551, an anomaly prediction module 5552, an object recognition module 5553, an information repair module 5554, and a prompt delivery module 5555. These modules are logically connected and can therefore be arbitrarily combined or further separated according to their implemented functions. The functions of each module will be described below.

[0079] In other embodiments, the object recognition device provided in this application can be implemented in hardware. As an example, the object recognition device provided in this application can be a processor in the form of a hardware decoding processor, which is programmed to execute the object recognition method provided in this application. For example, the processor in the form of a hardware decoding processor can be one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

[0080] For example, an embodiment of this application provides an object recognition device, including:

[0081] Memory, used to store executable instructions;

[0082] The processor, when executing executable instructions stored in the memory, implements the object recognition method provided in the embodiments of this application.

[0083] In some embodiments, the object recognition device can implement the object recognition method provided in this application by running a computer program. For example, the computer program can be a native program or software module in an operating system; it can be a native application (APP), i.e., a program that needs to be installed in the operating system to run; it can also be a mini-program, i.e., a program that only needs to be downloaded to a browser environment to run; or it can be a mini-program that can be embedded in any APP. In short, the above-mentioned computer program can be any form of application, module, or plugin.

[0084] The object recognition method provided in this application can be applied to various scenarios such as cloud technology, artificial intelligence, smart transportation, and vehicle-mounted systems. Below, we will describe the object recognition method provided in this application with reference to exemplary applications and implementations of the object recognition device provided in this application.

[0085] See Figure 3 , Figure 3 This is a flowchart illustrating the object recognition method provided in this application embodiment, which will be combined with... Figure 3 The steps shown are explained.

[0086] S101. Obtain the object operation data, object resource data, and historical resource transfer data of the object to be identified.

[0087] This application embodiment is implemented in a scenario where anomaly analysis is performed on the resource transfer status of different objects to identify objects that may be abnormal. For example, accounts that may be overdue or out of contact are identified from different bank accounts, and accounts that may be overdue or out of contact are identified from financial platforms, so as to handle these accounts in a timely manner and avoid risks. After starting object recognition, the object recognition device first obtains a clear object to be identified, and then obtains the object operation data, object resource data, and historical resource transfer data of the object to be identified, so as to use this data to analyze the risk level of the object to be identified. Here, the object to be identified refers to the financial account to be identified, such as a bank account, a lending platform account, etc., or it can refer to the social account to be identified, such as an account registered on a social platform, or a social account associated with a financial account.

[0088] It should be noted that object operation data represents the usage status of the object to be identified, and may include basic data such as the number of logins, login time, and the identifier of the login device used during login. It may also include data corresponding to resource transfer (borrowing and repayment) behaviors, such as the number of loans, loan amount, loan time, repayment amount, etc. Furthermore, object operation data may include historical operation data, such as the operation data of the object to be identified in a historical time period, such as the number of logins and login time of the object to be identified one day ago; object operation data may also include real-time operation data, such as the identifier of the login device of the object to be identified at the current time, etc., which are not limited in this application.

[0089] Understandably, the object recognition device can obtain object operation data from the login device corresponding to the object to be identified. Furthermore, the object recognition device can obtain real-time operation data from the login device of the object to be identified, and historical operation data from the big data platform.

[0090] The object resource data records the changes in resources within the object to be identified. The object resource data may include resource income information, such as monthly and annual income; resource expenditure information, such as monthly expenditure data; and may also include data on deposit amounts, estimated value of movable and immovable property; as well as data on income sources and the distribution of expenditure categories (e.g., the amount of resources spent on food delivery, the amount of resources spent on bank transfers, the amount of resources spent on salary income, and the amount of resources spent on red envelope income), etc., which are not limited herein.

[0091] Historical resource transfer data characterizes the creditworthiness of the object to be analyzed over a historical period. This data may include the object's resource transfer history, i.e., its performance history (repayment amount, number of repayments, repayment time, etc.); it may include its resource acquisition history, such as loan period and loan amount; it may also include its overdue history (e.g., overdue amount, overdue duration, overdue frequency, and the date of the most recent overdue payment); and it may include other credit information (e.g., multiple borrowings, using illegitimate means to extract resources, etc.), which are not limited herein.

[0092] It is understandable that object recognition devices can obtain object resource data and historical resource transfer data from the login device of the object to be identified, or they can obtain object resource data and historical resource transfer data of the object to be identified from the big data platform.

[0093] S102. Based on object operation data, object resource data, and historical resource transfer data, predict the degree of anomaly of the object to be identified and obtain the prediction result.

[0094] Object operation data indicates whether the object to be identified is being used normally. Normal use typically means whether the user corresponding to the object can be contacted. Therefore, object identification devices can analyze the object operation data to determine if the object's usage status is abnormal, such as the possibility of the user being unreachable. Object resource data indicates whether the object has the ability to transfer resources normally, such as whether an account can repay loans on time. Historical resource transfer data indicates the object's repayment habits, such as whether an account never makes timely payments. Therefore, object identification devices can predict the probability of abnormal resource transfers based on the obtained object resource data and historical resource transfer data. Next, the object identification device integrates these two types of probabilities to obtain the prediction result for the object. In other words, the prediction result represents the probability of abnormal resource transfers and the probability of abnormal usage status of the object.

[0095] It should be noted that for accounts with good repayment habits, such as those that never default or rarely default, it is unnecessary to analyze the possibility of abnormal usage status (these accounts generate operation data when making timely repayments, but this does not constitute an abnormal usage status, such as being out of contact). The object recognition device only needs to analyze the possibility of abnormal usage status for objects with a high probability of abnormal resource transfer. Therefore, in this embodiment, the prediction result obtained by predicting the degree of abnormality of the object to be identified needs to be able to simultaneously characterize the abnormal risks of resource transfer and usage status of the object to be identified, so as to facilitate the subsequent selection of the target objects that truly need attention.

[0096] In some embodiments, the object recognition device can input object operation data, object resource data, and historical resource transfer data into a trained risk level prediction model, and use the probability values ​​of resource transfer anomalies and usage status anomalies corresponding to the object to be identified as output by the risk level prediction model to obtain a prediction result.

[0097] In other embodiments, the object recognition device can also calculate a resource risk score for the remaining resources in the object resource data and an overdue risk score using the overdue frequency in the historical resource transfer data. The resource risk score and the overdue risk score are weighted to obtain a probability value of an abnormal resource transfer. Then, the probability value of an abnormal usage status is calculated based on the time elapsed between the most recent login time and the current time in the object operation data. The two probability values ​​are then weighted to obtain the final prediction result.

[0098] Furthermore, the object recognition device can obtain a resource risk score by the difference between the remaining resources and the repayment amount for this round. For example, the proportion of the difference to the repayment amount can be directly used as the resource risk score. The overdue risk score can be obtained by the relationship between the overdue frequency and the preset frequency. For example, when the overdue frequency is greater than the preset frequency, the overdue risk score can be directly set to 0.8.

[0099] It is understandable that "objects to be identified" is a general term for all objects that require anomaly prediction, not a specific object. In other words, the objects to be identified can include one or more objects. Correspondingly, the prediction result for each object does not specifically refer to the anomaly level of a single object; therefore, the prediction result includes the anomaly levels corresponding to one or more objects individually.

[0100] It should be noted that the prediction result may include the probability values ​​of abnormal resource transfer and abnormal usage status of the object to be identified, or it may refer to the weighted result of the probability values ​​of abnormal resource transfer and abnormal usage status of the object to be identified.

[0101] S103. Based on the prediction results, identify the target object from the objects to be identified.

[0102] After obtaining the prediction results, the object recognition device will determine whether there will be any abnormalities in the resource transfer and usage status of the object to be identified. Then, it will select objects that show abnormalities in resource transfer and usage status as target objects. Thus, the object recognition device can automatically identify target objects using data from the object to be identified across multiple dimensions, making target object identification more accurate and eliminating the need for manual confirmation based on object information such as contact details during the identification process.

[0103] In some embodiments, when the prediction result includes the probability values ​​of abnormal resource transfer and abnormal usage status, the object recognition device can compare the probability values ​​of abnormal resource transfer and abnormal usage status with the corresponding thresholds, thereby identifying objects whose probability values ​​of abnormal resource transfer and abnormal usage status are both greater than the corresponding probability thresholds as target objects.

[0104] In other embodiments, when the prediction result is a weighted result of the probability value of abnormal resource transfer and the probability value of abnormal usage status of the object to be identified, the object identification device can compare the weighted result with a weighted threshold to filter out the target object.

[0105] In this embodiment, the object recognition device can first acquire data on the object to be recognized in various dimensions, then use the data in various dimensions to predict the probability of the object to be recognized experiencing resource transfer anomalies and usage status anomalies, and finally, based on the prediction results, more accurately identify the target object from the objects to be recognized. Thus, during object recognition, the device fully utilizes the data of the object to be recognized in various dimensions, improving the accuracy of object recognition. Furthermore, during the recognition process, there is no need to manually confirm the object to be recognized, thereby reducing the lag in recognition, improving the efficiency of object recognition, and ultimately enhancing the intelligence level of object recognition.

[0106] See Figure 4 , Figure 4 This is a flowchart illustrating the object recognition method provided in the embodiments of this application. Figure 2 In some embodiments of this application, based on object operation data, object resource data, and historical resource transfer data, the degree of anomaly of the object to be identified is predicted to obtain the prediction result. The specific implementation process of S102 may include: S1021-S1023, as follows:

[0107] S1021. Based on the object resource data and historical resource transfer data, predict the probability of abnormal resource transfer for the object to be identified and obtain the first probability value.

[0108] In some embodiments, the object recognition device can input object resource data and historical resource transfer data into a trained delinquency probability prediction model (a sub-model in the risk prediction model responsible for predicting abnormal resource transfers) to obtain a first probability value for the object to be identified. In other embodiments, the object recognition device can also input object resource data into a first sub-model for calculating the delinquency probability value based on resource data to obtain a first sub-probability, input historical resource transfer data into a second sub-model for calculating the delinquency probability value based on historical credit to obtain a second sub-probability, and then weight the first and second sub-probabilities to obtain the final first probability value.

[0109] It should be noted that, in the embodiments of this application, the first probability value may refer to the probability that the resource transfer of the object to be identified is abnormal, or it may refer to the probability that the object to be identified will switch from the current abnormal resource transfer stage, such as the current overdue stage to the next overdue stage, that is, the probability that the abnormal resource transfer will continue to occur, such as the probability of switching from 30 days overdue to 60 days overdue, etc. This application does not limit it here.

[0110] S1022. Based on object operation data, predict the probability of abnormal usage status of the object to be identified, and obtain a second probability value.

[0111] The object recognition device can input object operation data into a loss probability prediction model to predict the second probability value of the corresponding user object. The object recognition device can also calculate the frequency of login, online and other operations of the object to be identified within a preset time period from the object operation data, and then use the frequency of use to determine the second probability value (for example, subtract the frequency of use from 1, and directly determine the second probability value as 0.8 when the frequency of use is less than a certain threshold).

[0112] Understandably, the missing person probability prediction model is a pre-trained model. Before starting object recognition, the object recognition device can first obtain initial samples that are manually categorized by different degrees of anomaly in usage status. Then, it can make phone calls to these samples and confirm the degree of anomaly of the initial samples based on the feedback from the phone calls. For example, if it is determined that no one answers the phone in a certain sample, it is identified as a high-risk missing person sample, thus obtaining training samples of different degrees of anomaly. These samples are then used to train the missing person probability prediction model.

[0113] It should be noted that the execution order of S1021 and S1022 will not affect the final risk level result. Therefore, in some embodiments, the object recognition device can execute S1022 first and then S1021, or it can execute S1021 and S1022 simultaneously.

[0114] S1023. Based on the first probability value and the second probability value, determine the prediction result of the object to be identified.

[0115] The object recognition device can directly weight the first probability value and the second probability value to obtain the prediction result of the object to be identified. Alternatively, it can determine the level of resource transfer anomaly and the level of usage status anomaly of the object to be identified based on the first probability value and the second probability value, and then integrate the level of resource transfer anomaly and the level of usage status anomaly to obtain the prediction result.

[0116] In this embodiment of the application, the object recognition device can first use object resource data and historical resource transfer data to predict the first probability value of the object to be identified, and use object operation data to predict the second probability value of the object to be identified. Then, it can analyze the first probability value and the second probability value together to obtain the prediction result, so as to determine the target objects of multiple abnormal levels based on the prediction result.

[0117] In some embodiments of this application, based on object resource data and historical resource transfer data, the probability of abnormal resource transfer of the object to be identified is predicted to obtain a first probability value. The specific implementation process of S1021 may include: S1021a-S1021b, as follows:

[0118] S1021a. Determine the current abnormal transfer stage of the object to be identified from historical resource transfer data.

[0119] S1021b. Using object resource data, predict the first probability value of the object to be identified switching from the current transition anomaly stage to the next transition anomaly stage.

[0120] It should be noted that object resource data includes at least resource acquisition information and resource expenditure information, and historical resource transfer data includes at least the history of resource transfer behavior and resource acquisition behavior. The resource acquisition information and resource expenditure information may include estimated income and expenditure amounts based on the actions of the object to be identified, or they may include income and expenditure amounts filled in by the user of the object to be identified. Furthermore, since the income and expenditure amounts filled in by the user may not be accurate, the object identification device will primarily use the estimated income and expenditure amounts based on the actions of the object to be identified to calculate the first probability value.

[0121] The resource transfer behavior history records information related to the resource transfer behavior of the object to be identified, such as the number of repayments, repayment amount, overdue amount, and overdue duration. Thus, the object identification device can extract the current abnormal transfer stage of the object to be identified from the resource transfer behavior history in the historical resource transfer data. Then, based on the current abnormal transfer stage, the next abnormal transfer stage can be determined. Finally, the object resource data is input into the overdue probability prediction model to calculate the first probability value of the object to be identified continuing to be overdue, that is, continuing to transfer from the current abnormal transfer stage to the next abnormal transfer stage.

[0122] Understandably, the transfer anomaly stage represents the abnormality of resource transfer of the object to be identified in the time dimension. For example, in 30-day units, the current transfer anomaly stage may be 30 days overdue, and the next transfer anomaly stage may be 60 days overdue, etc. In a quarterly unit, the current transfer anomaly stage may be one quarter overdue, and the next transfer anomaly stage may be two quarters overdue, etc.

[0123] In some embodiments, the overdue probability prediction model may include multiple sub-models corresponding to different transfer anomaly stages. These sub-models are used to predict the probability of switching from their corresponding transfer anomaly stage to the next transfer anomaly stage. In this case, after determining the current transfer anomaly stage, the object recognition device can obtain the sub-model corresponding to the current transfer anomaly stage from the overdue probability prediction model, and then input the object resource data into the sub-stage model corresponding to the current transfer anomaly stage to obtain the probability value of the object to be identified moving from the current transfer anomaly stage to the next transfer anomaly stage. This probability value is the first probability value.

[0124] For example, Figure 5 This is a schematic diagram of the process of predicting the first probability value provided in the embodiments of this application. When the current transfer anomaly stage is M0, that is, 0 days overdue 5-1, the corresponding next transfer anomaly stage is M1, that is, 30 days overdue 5-2. The object recognition device obtains the M0-M1 sub-stage model 5-31 from the overdue probability prediction model 5-3 (including the M0-M1 sub-stage model 5-31, the M1-M2 sub-stage model 5-32, and the M2-M3 sub-stage model 5-33) to calculate the object resource data and obtain the first probability value 5-4.

[0125] Of course, the object recognition device can also use object resource data and historical resource transfer data to calculate the first probability value of the object to be identified switching from the current transfer anomaly stage to the next transfer anomaly stage. That is, the object resource data and historical resource transfer data are simultaneously input into the sub-stage model corresponding to the current transfer anomaly stage to obtain the first probability value.

[0126] In other embodiments, the overdue probability prediction model is a whole, that is, it does not divide the model into corresponding sub-stages according to different transfer anomaly stages. Thus, the object recognition device directly inputs the account resource data into the overdue probability prediction model to output the first probability value of the object to be identified switching from the current transfer anomaly stage to the next transfer anomaly stage.

[0127] In this embodiment of the application, the object recognition device first determines the current abnormal transfer stage of the object to be identified from the historical resource transfer data, and then calculates the first probability value of the object to be identified switching from the current abnormal transfer stage to the next abnormal transfer stage, so as to obtain the prediction result based on the first probability value.

[0128] In some embodiments of this application, the prediction result of the object to be identified is determined based on the first probability value and the second probability value. That is, the specific implementation process of S1023 may include: S1023a-S1023c, as follows:

[0129] S1023a. When the first probability value is greater than the first probability threshold, the first prediction result of the object to be identified is determined to be the switch from the current transition anomaly stage to the next transition anomaly stage.

[0130] The object recognition device obtains a first probability threshold and compares the first probability value with the first probability threshold. When the first probability threshold is greater than the first probability threshold, the first prediction result is determined to be that the object to be identified continues to have resource transfer anomalies until it enters the next transfer anomaly stage. For example, the account to be identified is continuously overdue until it enters the next transfer anomaly stage.

[0131] It is understood that the first probability threshold is preset manually, such as 0.9, 0.6, or other values ​​corresponding to actual needs; the first probability threshold can also be automatically determined by the object recognition device based on the current abnormal transfer stage, for example, when the current abnormal transfer stage is 0 days, the first probability threshold is automatically determined to be 0.8, etc., which is not limited in this application.

[0132] S1023b: Based on the second probability value and the second probability threshold of each of the multiple abnormal state stages of the usage state, select the second prediction result of the object to be identified from the multiple abnormal state stages.

[0133] The object recognition device compares the second probability value with the second probability thresholds corresponding to multiple abnormal states in the usage state, thereby identifying the abnormal state stage matched by the second probability value, and then determining the matched abnormal state stage as the second prediction result for the object to be identified. For example, when the second probability value is greater than the second probability threshold of 0.8 corresponding to the highest abnormal state stage (e.g., high-risk loss of contact), the highest abnormal state stage will be determined as the second prediction result.

[0134] It is understood that the second probability threshold can be preset manually, such as set to 0.5 or 0.7; the second probability threshold can also be automatically determined by the object recognition device for multiple abnormal state levels, such as being set randomly or according to a Gaussian distribution, etc., which are not limited here.

[0135] S1023c: Integrate the first prediction result and the second prediction result into a prediction result for the object to be identified.

[0136] The object recognition device integrates the first and second prediction results to obtain the prediction result of the object to be identified. For example, switching from the current transfer anomaly stage to the next transfer anomaly stage and integrating it with the highest state anomaly stage will result in a prediction result of continuous resource transfer anomalies and a high-risk usage status anomaly.

[0137] In some embodiments of this application, the specific implementation process of identifying the target object from the objects to be identified based on the prediction results, i.e., S103, may include: S1031-S1032, as follows:

[0138] S1031. Based on the first and second prediction results in the prediction results, select the matching anomaly level corresponding to the object to be identified from multiple anomaly levels.

[0139] The object recognition device analyzes the prediction results to obtain a first prediction result and a second prediction result. Then, it matches the first prediction result with the resource transfer anomaly criteria for each of the multiple anomaly levels, and matches the second prediction result with the usage status anomaly criteria for each of the multiple anomaly levels. The anomaly level where both the first and second prediction results match is determined as the matching anomaly level. In this way, the object recognition device can determine the matching anomaly level for each of the one or more objects contained within the object to be identified.

[0140] It should be noted that the anomaly level can be determined based on both the resource transfer anomaly stage and the usage status anomaly stage. Specifically, when both the resource transfer anomaly stage and the usage status anomaly stage are at their highest levels, the anomaly level is the highest anomaly level. When one of the resource transfer anomaly stage and the other is at its lowest level, or when both are at a medium level between the highest and lowest levels, the anomaly level is a medium anomaly level between the highest and lowest levels. When both the resource transfer anomaly stage and the usage status anomaly level are at a low level, the anomaly level is the lowest anomaly level.

[0141] S1032. Identify the objects in the objects to be identified whose anomaly level reaches the preset anomaly level as target objects.

[0142] After obtaining the matching anomaly level, the object recognition device can acquire the preset anomaly level, then compare the matching anomaly level with the preset anomaly level, and extract the accounts whose matching anomaly level reaches the preset anomaly level from the objects to be identified as the target objects.

[0143] Understandably, the preset anomaly level can be set according to actual conditions. For example, when the preset anomaly level is set to the lowest anomaly level, the object recognition device will match objects with anomaly levels above the lowest anomaly level as target objects. When the preset anomaly level is set to the medium anomaly level, the object recognition device will use objects corresponding to the medium and highest anomaly levels as target objects. In this way, the object recognition device realizes the process of identifying target objects based on prediction results.

[0144] based on Figure 3 See Figure 6 , Figure 6 This is a flowchart illustrating the object recognition method provided in the embodiments of this application. Figure 3 In some embodiments of this application, after identifying the target object from the objects to be identified based on the prediction result, i.e., after S103, the method may further include: S104-S105, as follows:

[0145] S104. Based on the target anomaly level corresponding to the target object, determine the information repair method for the target object.

[0146] The target object is a part of the object to be identified. Therefore, the object recognition device filters out the matching anomaly levels corresponding to the target object from the matching anomaly levels of the object to be identified, and uses these matching anomaly levels as the target anomaly levels. To improve the effectiveness of information repair for the target object, in this embodiment, the object recognition device uses the target anomaly level corresponding to the target object to determine the most suitable information repair method for the target object, so as to achieve information repair according to different repair methods for objects with different anomaly levels.

[0147] In some embodiments, when the target anomaly level is the lowest anomaly level, it indicates that the anomaly degree of the target object is not high. Therefore, the object recognition device can perform information repair based on the identity identifier of the target object. That is, for the target object, it can search for objects with the same identity identifier (e.g., search for accounts under the same ID card) and use the contact information and other object information of these objects to perform information repair on the target object.

[0148] In other embodiments, when the target anomaly level is medium, the object recognition device will perform information repair on the target object by finding interactive objects that have had transactions with the target object, or objects that have logged in on the same login device as the target object. That is, it will find object information such as contact information of objects that have a certain connection with the target object, thereby realizing information repair.

[0149] In other embodiments, when the target anomaly level is the highest anomaly level, the object recognition device will perform information repair based on the target object's social relationships, that is, find the contact information and other object information of the objects with which the target object has certain social relationships, thereby expanding the scope of information repair.

[0150] S105. Using information repair methods, information repair is performed on the target object to obtain target repair information.

[0151] After obtaining a suitable information repair method, the object recognition device uses the information repair method adapted to the target object to repair the target object's information, such as repairing contact information or repairing device usage information, thereby regaining the target object's object information, i.e., obtaining the target repair information.

[0152] In this embodiment, after identifying the target object, the object recognition device also determines a suitable information repair method for the target object, and then uses the information repair method to repair the information of the target object to obtain target repair information, thereby realizing the corresponding information repair according to the abnormal level and improving the information repair effect.

[0153] In some embodiments of this application, the information repair method includes: information repair based on identity identifier. In this case, information repair is performed on the target object to obtain target repair information. The specific implementation process of S105 may include: S1051-S1054, as follows:

[0154] S1051. Based on the identity identifier of the target object, extract the associated objects of the target object from the preset database.

[0155] In this embodiment of the application, the object recognition device extracts the identity identifier corresponding to the target object, and then, based on the extracted identity identifier, mines objects with the same identity identifier from a preset database. For example, it mines financial accounts with the same ID number, social accounts with the same ID number, etc., from the preset database to obtain associated objects.

[0156] It is understood that the identification can be an ID card number, an employee ID, a student ID, etc., and this application does not impose any restrictions on it.

[0157] It should be noted that the preset database corresponds to the target object. When the target object is a financial account, the preset database is the financial account database; when the target object is a social media account, the preset database is the social media account database.

[0158] S1052. Statistically analyze the operation frequency of associated objects to obtain the first activity level of associated objects.

[0159] After obtaining the associated account, the object recognition device will acquire the operation data corresponding to the associated object, and then use the operation data to count the operation frequency of the associated object, such as the frequency of login and transaction (e.g., transfer and payment), thereby obtaining the first activity level of the associated object.

[0160] S1053. Based on the attribute information of the target object and the attribute information of the associated object, calculate the similarity between the associated object and the target object.

[0161] The object recognition device acquires the attribute information of the target object and the attribute information of the associated objects, and then calculates the similarity between the target object and the associated objects based on the degree of overlap or similarity between their attribute information.

[0162] It should be noted that in some embodiments, the object recognition device may execute S1053 first and then S1052, or it may execute S1052 and S1053 simultaneously. This application does not limit this.

[0163] S1054. The object information of the N objects with the highest first activity level determined based on the first activity level of the associated objects, and the object information of the N objects with the highest similarity level determined based on the similarity with the target object, are determined as the target repair information to realize information repair based on identity identification.

[0164] The object recognition device filters out the N most active objects (such as financial accounts, social media accounts, etc.) from the associated objects, and then filters out the N objects with the highest similarity to the target object. Next, it obtains object information such as contact information and device information from these filtered objects, and uses this obtained object information to determine the target object's repair information. It should be noted that the value of N can be set according to the actual situation, and N is a positive integer.

[0165] It is understood that in some embodiments, the object recognition device can also extract the N most active objects and the N objects with the highest similarity to the target object, then deduplicate these 2N objects, and finally extract the contact information and other object information of the deduplicated objects as target repair information.

[0166] In this embodiment, the object recognition device can mine associated objects based on the identity identifier of the target object, and then, based on the activity level of the associated objects and their similarity to the target object, select objects for information repair, and determine the object information of these objects as the target repair information. This enables the use of multiple objects under the same identity identifier to complete information repair, thereby improving the effectiveness of information repair.

[0167] In some embodiments of this application, the information repair method includes: information repair based on the login device identifier. In this case, the specific implementation process of repairing the target object to obtain the target repair information, i.e., S105, may include: S1055-S1058, as follows:

[0168] S1055. Obtain the login device identifier corresponding to the target object.

[0169] S1056. Based on the login device identifier, extract the same device object that is the same as the login device of the target object from the preset database.

[0170] In this embodiment of the application, the object recognition device first obtains the identification information of the login device of the target object to obtain the login device identifier, and then mines out the object with the same login device identifier from the preset database, such as mining out the account that has logged in on the same login device as a certain financial account, and takes the mined object as the same device object.

[0171] It is understood that the login device identifier can refer to the IP address of the login device or the unique device identifier of the login device; this application does not limit this.

[0172] S1057. Statistically analyze the operation frequency of objects on the same device to obtain the second activity level of objects on the same device.

[0173] The object recognition device counts the frequency of login, transaction and other operations of objects on the same device to obtain the second activity level of objects on the same device, so as to carry out subsequent information repair.

[0174] S1058. The object information of the N objects with the highest second activity level determined based on the second activity level is identified as the target repair information, and information repair is performed based on the login device identifier.

[0175] Objects on the same device are very likely to share the same user as the target object. In other words, objects on the same device are users of other objects besides the target object. Therefore, the N most active objects among the objects on the same device are likely to be objects frequently used by the target object's user, in addition to the target object. The object recognition device extracts the object information from these objects to obtain the target repair information.

[0176] In this embodiment, the object recognition device can identify the login device of the target object, extract objects with the same login device as the target object from a preset database, and then extract the object information of the N objects with the highest activity from the objects with the same login device to obtain the target repair information. This realizes the use of other objects used by the user to repair the loss of connection of the target object, thus improving the information repair effect.

[0177] In some embodiments of this application, the information repair method includes: information repair based on the interactive object. In this case, information repair is performed on the target object to obtain the target repair information. The specific implementation process of S105 may include: S1059-S10511, as follows:

[0178] S1059. From the preset database, find the interactive objects that have interactive behavior with the target object.

[0179] S10510. Statistical analysis of the interaction behaviors between the interactive object and the target object is performed to obtain the interaction frequency between the interactive object and the target object.

[0180] In this embodiment of the application, the object recognition device first extracts objects that have interactive behaviors with the target object from a preset database, such as accounts that have engaged in transfers or remittances, as interactive objects. Then, it statistically analyzes the frequency of interaction between the target object and the interactive objects to obtain the interaction frequency.

[0181] S10511. The object information of the N objects with the highest interaction frequency with the target object, determined according to the interaction frequency, is identified as the target repair information, thereby realizing information repair based on the interaction objects.

[0182] It should be noted that objects that interact frequently with the target object are very likely to be those with whom the target object has close social or financial dealings. Therefore, the object recognition device extracts the N objects with the highest interaction frequency from the interacting objects and obtains the object information of these N objects to obtain target restoration information and complete the information restoration.

[0183] In this embodiment, the object recognition device can determine the target repair information based on the object information of the interactive account that has frequent interaction with the target object, thereby improving the effect of information repair.

[0184] In some embodiments of this application, the information repair method includes: information repair based on social relationships. In this case, information repair is performed on the target object to obtain target repair information. The specific implementation process of S105 may include: S10512-S10515, as follows:

[0185] S10512. From the obtained identity relationship data of the target object, extract candidate objects that have social relationships with the target object.

[0186] The object recognition device acquires the identity relationship data of the target object, then parses the identity relationship data to obtain the social relationship network of the target object, and then extracts the objects that have social relationships with the target object. For example, the accounts of the user's parents and friends can be extracted as candidate objects.

[0187] It is understood that social relationships can be family relationships, friendships, or colleague relationships; this application does not limit this. The identity relationship data includes the target's identity information and relationship profile. Identity information refers to the personal information of the user corresponding to the target, such as name, age, gender, and occupation, while the relationship profile is the social relationship network of the user corresponding to the target, such as the user's parents' financial accounts or their spouse's social media accounts.

[0188] S10513. Calculate the intimacy between the candidate object and the target object based on the number of interactions between the target object and the candidate object.

[0189] Object recognition devices count the number of social interactions between target objects and candidate objects, and then use the counted number directly as the intimacy between the target object and the candidate object. Alternatively, the intimacy between the target object and the candidate object can be obtained by taking the logarithm of the number of social interactions.

[0190] It is understood that the social interaction between the target and the candidate can refer to behaviors such as chatting, liking, and sharing between the target and the candidate, or behaviors such as making payments on behalf of each other and giving gifts between the target and the candidate. This application does not limit this.

[0191] S10514. Statistically analyze the operation frequency of candidate objects to obtain the third activity level of candidate objects.

[0192] S10515. The object information of the N objects with the highest intimacy with the target object, determined based on the intimacy between the candidate object and the target object, and the object information of the N objects with the highest third activity, determined based on the third activity level, are determined as target repair information to realize information repair based on social relationships.

[0193] The object recognition device statistically analyzes the frequency of login and transaction operations of candidate objects to determine their activity level, thus obtaining a third activity level. Next, the device extracts the N objects with the highest third activity level from the candidate objects to identify those that the user is actively using. It also extracts the N objects with the highest affinity to the target object. Finally, it uses the contact information and other object information of these 2N objects as target repair information to complete the information repair process.

[0194] Of course, the object recognition device can also deduplicate these 2N objects and use the contact information and other object information of the deduplicated objects as the target repair information of the target object, thereby completing the information repair.

[0195] In this embodiment, the object recognition device can perform information repair based on object information such as contact information of objects with close social relationships with the target object, thereby improving the effectiveness of information repair.

[0196] It should be noted that in some embodiments, when the target object is at the lowest anomaly level, the object recognition device can only perform information repair based on identity identifiers. When the target object is at a medium anomaly level, between the lowest and highest anomaly levels, the object recognition device can, in addition to information repair based on identity identifiers, also perform information repair based on login device identifiers and information repair based on interactive objects. When the target object is at the highest risk level, the object recognition device can simultaneously perform information repair based on identity identifiers, login device identifiers, interactive objects, and social relationships. In this way, the object recognition device can gradually add a wider range of information repair methods as the anomaly level of the target object increases, ensuring that at the lowest anomaly level, it will not disturb users other than the user of the target object, while ensuring that effective target repair information is obtained at the highest anomaly level, further improving the effectiveness of information repair.

[0197] In some embodiments of this application, after the information repair method is used to repair the target object and obtain the target repair information, i.e. after S105, the method may further include: S106, as follows:

[0198] S106. Based on the target repair information, issue a prompt message to indicate that the target object has experienced a resource transfer anomaly.

[0199] After obtaining the target repair information, the object recognition device can also send resource transfer anomaly prompts to each object information in the target repair information, such as each contact method, to remind the target object that it has been overdue or lost contact, so that the target object can fulfill its obligations in a timely manner, such as timely repayment.

[0200] Understandably, the content of the notification message can be modified based on the level of non-payment the target is experiencing. For example, when the target is at the lowest level of non-payment, the notification message includes the user's contact information, allowing for a more direct notification that the payment has been overdue. This could include the overdue amount, duration, and the impact on future credit scores. Conversely, when the target is at the highest level of non-payment, the notification message includes the contact information of the user's family and friends, providing a more tactful way to protect the user's privacy.

[0201] In this embodiment of the application, the object recognition device can also send a prompt message to the target repair information to remind the user of the target object to fulfill the agreement in a timely manner, thereby realizing the reminder to the target object.

[0202] The following will describe an exemplary application of the embodiments of this application in a real-world application scenario.

[0203] The embodiments of this application are implemented in a scenario where a server (object identification device) identifies and restores the contact of a credit user (the object to be identified) due to the risk of loss of contact.

[0204] The server primarily combines various data accumulated by loan users in social and financial transaction scenarios. Based on big data, artificial intelligence, and graph correlation analysis technologies, it employs multiple intelligent risk identification and ranking algorithms and measurements to comprehensively assess the risk and severity of user disappearance after borrowing. Then, it uses different contact recovery methods for users with different risk levels, targeting different social circles. Finally, it outputs the corresponding contact recovery customer groups for each risk level and takes appropriate action. For example, for users with higher contact risk, it uses methods involving both individuals and social groups to facilitate earlier recall and ensure effective recovery and risk management for this group. For users with low contact risk, it generally only restores contact with individuals to ensure a smooth repayment experience.

[0205] Figure 7 This is a schematic diagram of the system framework for loss of contact risk identification and loss of contact recovery provided in an embodiment of this application. See also... Figure 7The system framework comprises three parts: basic data collection 7-1, loss of contact risk identification 7-2, and loss of contact recovery 7-3. Basic data collection 7-1 primarily involves accumulating core basic data after a user takes out a loan. This core basic data includes user operational relationships and behaviors 7-11 (object operational data), user identity and relationship profiles 7-12 (identity relationship data), user asset data 7-13 (object resource data), and user historical credit 7-14 (historical resource transfer data). User operation relationships and behaviors 7-11 include operation behaviors 7-111 (e.g., number of times the loan app is used, time sequence, and data associated with multiple accounts using the app), transaction relationship networks 7-112 (e.g., number of transactions, amount, time sequence, and relationship between the transacting parties), and borrowing and repayment behaviors 7-113 (number of loans / repayments, amount, etc.); User identity and relationship profile 7-12 includes user identity information 7-121 (e.g., work / study related information, occupation / position / length of service, etc.), relationship networks 7-122 (e.g., data on family, friends, colleagues, etc.), and social responsibilities 7-123 (…). Social network area / scope, etc.); User asset data 7-13, including monthly income / consumption 7-131 (e.g., monthly salary, rent, red envelopes, shopping, takeout, etc. index sorting), savings 7-132 (e.g., investment transfers, cash balance, etc.), and other assets 7-133 (e.g., movable property, immovable property, etc. registered by the user); User credit history 1-4, including default history 7-141 (e.g., default amount, duration, frequency, recent time, etc.), repayment history 7-142 (e.g., number of repayments, amount, etc.), and other risks 7-143 (e.g., multiple borrowing, cashing out, etc.).

[0206] The "Lost Contact Risk Identification 7-2" is mainly used for risk identification and classification of lost contact users. In this application, the server mainly uses the continuous default identification model and the lost contact risk prediction model to predict the continuous default risk and the lost contact risk using the user's basic data. Then, it comprehensively judges the user's lost contact risk level (prediction result), for example, obtaining a slightly lost contact overdue customer group 7-21, a moderately lost contact overdue customer group 7-22, and a high-risk lost contact overdue customer group 7-23.

[0207] The "Recovery from Missing Contact 7-3" method is mainly used to implement multi-layered recovery methods for different levels of risk of missing contact (the level of abnormality corresponding to the target object). For example, for the slightly missing and overdue customer group 7-21, the method of recovery from missing contact with the same natural person 7-31 is used; for the moderately missing and overdue customer group 7-22, the method of recovery from missing contact with shared accounts 7-32 is used; and for the high-risk missing and overdue customer group, the method of recovery from missing contact with a group 7-33 is used.

[0208] Furthermore, the persistent default identification model has sub-stage models corresponding to multiple stages of default risk, such as the M0-M1 transfer prediction model (0 days overdue to 30 days overdue), the M1-M2 transfer prediction model (30 days overdue to 60 days overdue), and the M2-M3 transfer prediction model (60 days overdue to 90 days overdue). The inputs to these models are all collected basic data. For each user, the corresponding stage is first determined, such as M0, M1, or M2 (the current abnormal transition stage of the user). Then, using the aforementioned sub-stage model, the probability value of transitioning to the next stage is predicted, thus obtaining the probability value of continued delinquency risk (the first probability value of transitioning from the current abnormal transition stage to the next abnormal transition stage). The risk level is then judged based on the probability value of continued delinquency risk. For example, if the probability value of the M0-M1 sub-stage model is greater than 0.9, the risk of the user transitioning from continuous default in M0 to M1 is considered high (the first predicted result is switching from the current abnormal transition stage to the next abnormal transition stage). If the probability value is less than 0.2, the risk of the user transitioning from continuous default to M1 is considered low. Furthermore, the continuous default identification model can employ logistic regression and ensemble learning, such as gradient boosting trees and XGBoost models, with training samples as the base data and individual risk customer groups for each stage.

[0209] The model for predicting the level of user contact loss is based on the user's recent operational behavior characteristics, such as app logins / message clicks, to make predictions (based on the user's operational data, the probability of abnormal usage status is predicted, resulting in a first probability value). The risk level is then determined based on different predicted probability values ​​(based on a second probability value and second probability thresholds for multiple abnormal usage status stages, the second prediction result is selected from multiple abnormal status stages). For example, a predicted probability value greater than 0.8 is considered a high-risk user who has lost contact. The model can employ logistic regression and ensemble learning, such as gradient boosting trees and XGBoost models. Training samples can first undergo preliminary multi-level classification of contact loss levels, and then gradually initiate phone follow-ups, using feedback from these calls to determine the contact loss risk level of each sample.

[0210] Finally, the server combines the risk of continued delinquency and the predicted risk of loss of contact to comprehensively determine the current level of loss of contact (integrating the first prediction result and the second prediction result into a single prediction result). For example, for customers with minor delinquency and loss of contact, the risk of both continued delinquency and loss of contact is relatively low; for customers with moderate delinquency and loss of contact, the risk of either continued delinquency or loss of contact is relatively high; and for customers with high risk of loss of contact and loss of contact, the risk of both continued delinquency and loss of contact is extremely high.

[0211] Lost contact recovery for the same natural person refers to calculating the connection between multiple accounts (associated objects) created with the same identity ID (identity identifier) ​​as the lost user, and their associated contact information, thereby achieving recovery in cases where multiple accounts belong to the same person. Shared account lost contact recovery refers to extracting accounts used by the lost user on the same device (same device object), or transaction accounts involved in multiple fund transfers, thereby recovering accounts from people closely connected to the lost user, or accounts transferred after the lost user went missing. Group-based recovery (delicate repair based on social relationships) refers to situations where it is essentially impossible to contact the lost user through natural persons or highly intimate accounts, requiring the restoration of connections with colleagues and friends. In this case, the majority of the recovered users are the original user, and caution should be exercised when notifying users of this type of recovery.

[0212] Figure 8 This is a schematic diagram of multi-layer disconnection repair provided in an embodiment of this application. See also... Figure 8The multi-layered contact recovery process includes individual recovery (8-1), shared account recovery (8-2), and group / circle recovery (8-3). In individual recovery (8-1), the core data (8-A) is the user's associated UIN (associated object) with the same ID card (8-11), the activity level (8-12), and attributes (8-13) of the associated UIN (attribute information of the associated object). The recovery process (8-B) involves obtaining the contact information of the top-N accounts with the highest activity and similarity based on the associated UIN (8-14). This involves identifying the object information of the N most active objects (based on the first activity level of the associated object) and the N most similar objects (based on the similarity to the target object), which are then used as the target recovery information. The recovery effect (8-C) achieved a baseline recoverable rate (8-15), with an individual recovery rate >95% (8-16). Correspondingly, the reminder application (8-D) provides reminders specifically for individuals (8-17). The core data for shared account recovery (8-2) is either the UIN (8-21) of the same device ID (same device object) or the transfer UIN (8-22) of funds transfer. The recovery process (8-B) involves retrieving the contact information (8-23) of the top-N most active accounts from these UINs (based on the second activity level, the object information of the N objects with the highest second activity level) and the contact information (8-24) of the top-N accounts with high transaction frequency (based on interaction frequency, the contact information of the N objects with the highest interaction frequency with the target object is determined as the target recovery information). The recovery effect (8-C) of this method is 10% higher than that of recovery with a natural person (8-1), with a recovery rate >90% (8-26). The corresponding reminder application is to remind close contacts (8-27) or to remind transferred accounts after loss of contact (8-28). The core data for group circle recovery 8-3 (8-A) consists of family relationship (UIN 8-31) and work relationship (UIN 8-32) (candidates with social relationships). The recovery process (8-B) involves taking the contact information of the top-N accounts with the highest intimacy and the top-N accounts with the highest activity (8-33) to determine the target information for repair. The recovery effect (8-C) of this method is 10% higher than that of shared account recovery (8-2). In the corresponding reminder application (8-D), reminders are also sent to colleagues and friends (8-35). The majority of the customers who have been repaired are the individuals themselves.

[0213] The following section explains the process for identifying the risk of losing contact and for restoring contact.

[0214] See Figure 9 , Figure 9 This is a schematic diagram of the process for identifying and recovering lost contact as provided in an embodiment of this application. The process includes:

[0215] S201, Basic Data Acquisition.

[0216] This involves collecting and aggregating user basic behavior, assets, identity, credit, and various relationship data. This can be achieved using data reporting services from smart terminals, big data storage platforms like Redis / HDFS, and big data analytics platforms like SparkSQL. The basic data is aggregated into two categories: offline and online features. For complex analysis and prediction models, shared accounts, and group circles, offline storage can be used for analysis. However, for resolving issues related to the same individual and identifying lost contact groups, both offline and real-time features (such as real-time user login scenarios) can be analyzed together to make the analysis results more accurate.

[0217] S202, prediction of multiple levels of loss of contact risk.

[0218] This process primarily combines the prediction results of the probability of continued default and the probability of loss of contact to classify different levels of loss of contact. Risk models can also include both offline and online methods. For complex models predicting the probability of continued default, offline feature calculations can be used, while for models predicting the probability of loss of contact, a combination of offline and real-time feature calculations can be used.

[0219] S203. Establish a multi-layered disconnection repair model.

[0220] By combining the risk of losing contact with data on relationships across different social circles, a multi-layered model for restoring contact between the individual, transferred accounts, and family and friends is established.

[0221] S204, Multi-layer Disconnection Repair Application.

[0222] Different contact loss repair methods are matched according to different repair layers, and the contact information obtained from the repair is used to urge the recovery (a prompt message is sent to remind the target object that a resource transfer anomaly has occurred).

[0223] The above methods enable more accurate and effective identification of missing customer groups and their risk levels, improving the intelligence of missing customer group identification. Furthermore, different contact restoration methods are adopted for different missing customer groups, thereby enabling differentiated reminder methods for risk management based on different risk levels of missing customer groups, and improving the user experience of reminders.

[0224] The following description continues to illustrate the exemplary structure of the object recognition device 555 provided in the embodiments of this application as a software module. In some embodiments, such as Figure 2 As shown, the software module stored in the object recognition device 555 of the memory 550 may include:

[0225] Data acquisition module 5551 is used to acquire object operation data, object resource data and historical resource transfer data of the object to be identified;

[0226] Anomaly prediction module 5552 is used to predict the degree of anomaly of the object to be identified based on the object operation data, the object resource data and the historical resource transfer data, and obtain a prediction result; wherein, the prediction result characterizes the probability that the resource transfer of the object to be identified is abnormal, and the probability that the usage status of the object to be identified is abnormal.

[0227] The object recognition module 5553 is used to identify the target object from the objects to be identified based on the prediction result.

[0228] In some embodiments of this application, the anomaly prediction module 5552 is further configured to predict the probability of resource transfer anomalies of the object to be identified based on the object resource data and the historical resource transfer data, and obtain a first probability value; predict the probability of usage status anomalies of the object to be identified based on the object operation data, and obtain a second probability value; and determine the prediction result of the object to be identified based on the first probability value and the second probability value.

[0229] In some embodiments of this application, the anomaly prediction module 5552 is further configured to determine the current transfer anomaly stage of the object to be identified from the historical resource transfer data; and to predict the first probability value of the object to be identified switching from the current transfer anomaly stage to the next transfer anomaly stage using the object resource data; wherein the object resource data includes at least resource harvesting information and resource expenditure information, and the historical resource transfer data includes at least resource transfer behavior history and resource acquisition behavior history.

[0230] In some embodiments of this application, the anomaly prediction module 5552 is further configured to: determine that the first prediction result of the object to be identified is switching from the current transfer anomaly stage to the next transfer anomaly stage when the first probability value is greater than the first probability threshold; filter out the second prediction result of the object to be identified from the multiple state anomaly stages based on the second probability value and the second probability threshold of each of the multiple state anomaly stages of the usage state; and integrate the first prediction result and the second prediction result into the prediction result of the object to be identified.

[0231] In some embodiments of this application, the object recognition module 5553 is further configured to, based on the first prediction result and the second prediction result in the prediction results, filter out the matching abnormal level corresponding to the object to be identified from multiple abnormal levels; and determine the object whose matching abnormal level among the objects to be identified reaches the preset abnormal level as the target object.

[0232] In some embodiments of this application, the object recognition device 555 further includes: an information repair module 5554; the information repair module 5554 is used to identify a target object from the objects to be identified based on the prediction result, determine an information repair method for the target object based on the target anomaly level corresponding to the target object, and use the information repair method to repair the information of the target object to obtain target repair information.

[0233] In some embodiments of this application, the information repair method includes: information repair based on identity identifier; the information repair module 5554 is further configured to: mine associated objects of the target object from a preset database based on the identity identifier of the target object; statistically analyze the operation frequency of the associated objects to obtain the first activity level of the associated objects; calculate the similarity between the associated objects and the target object based on the attribute information of the target object and the attribute information of the associated objects; and determine the object information of the N objects with the highest first activity level determined based on the first activity level of the associated objects, and the object information of the N objects with the highest similarity level determined based on the similarity with the target object, as the target repair information, thereby realizing information repair based on the identity identifier; N is a positive integer.

[0234] In some embodiments of this application, the information repair method includes: information repair based on login device identifier; the information repair module 5554 is further configured to obtain the login device identifier corresponding to the target object; extract objects with the same login device as the target object from a preset database based on the login device identifier; count the operation frequency of the objects with the same login device to obtain the second activity level of the objects with the same login device; and determine the object information of the N objects with the highest second activity level as the target repair information, thereby realizing the information repair based on login device identifier.

[0235] In some embodiments of this application, the information repair method includes: information repair based on interactive objects; the information repair module 5554 is further configured to: search for interactive objects that have interactive behavior with the target object from a preset database; perform frequency statistics on the interactive behavior of the interactive objects and the target object to obtain the interaction frequency between the interactive objects and the target object; and determine the object information of the N objects with the highest interaction frequency with the target object, as determined according to the interaction frequency, as the target repair information, thereby realizing the information repair based on interactive objects.

[0236] In some embodiments of this application, the information repair method includes: information repair based on social relationships; the information repair module 5554 is further configured to: mine candidate objects that have social relationships with the target object from the obtained identity relationship data of the target object; calculate the intimacy between the candidate object and the target object based on the number of social interactions between the target object and the candidate object; statistically analyze the operation frequency of the candidate object to obtain the third activity level of the candidate object; and determine the object information of the N objects with the highest intimacy with the target object determined based on the intimacy between the candidate object and the target object, and the object information of the N objects with the highest third activity level determined based on the third activity level, as the target repair information, thereby realizing the information repair based on social relationships.

[0237] In some embodiments of this application, the object identification device 555 further includes: a prompting module 5555; the prompting module 5555 is used to perform information repair on the target object using the information repair method, and after obtaining target repair information, to perform information repair on the target object using the information repair method, and after obtaining target repair information, to issue a prompt message to the target object indicating that a resource transfer anomaly has occurred.

[0238] This application provides a computer program product or computer program that includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the object identification method described in this application.

[0239] This application provides a computer-readable storage medium storing executable instructions. When these executable instructions are executed by a processor, they cause the processor to execute the object recognition method provided in this application. For example, ... Figure 3 The object recognition method is shown.

[0240] In some embodiments, the computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or it may be a variety of devices including one or any combination of the above-mentioned memories.

[0241] In some embodiments, executable instructions may take the form of a program, software, software module, script, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0242] As an example, executable instructions may, but do not necessarily, correspond to files in a file system. They may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a Hyper Text Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple collaborating files (e.g., a file that stores one or more modules, subroutines, or code sections).

[0243] As an example, executable instructions can be deployed to execute on a single object recognition device, or on multiple object recognition devices located at one location, or on multiple object recognition devices distributed across multiple locations and interconnected via a communication network.

[0244] In summary, through the embodiments of this application, the object recognition device can first acquire data on the object to be identified in various dimensions, then use the data in each dimension to predict the probability of the object to be identified experiencing resource transfer anomalies and usage status anomalies, and finally, based on the prediction results, more accurately identify the target object from the objects to be identified. Thus, during object recognition, it fully utilizes the data in each dimension of the object to be identified, improving the accuracy of object recognition. Furthermore, during the recognition process, there is no need for manual confirmation of the object to be identified, thereby reducing the lag in recognition, improving the efficiency of object recognition, and ultimately enhancing the intelligence of object recognition. After determining the target object, the object recognition device also determines an information repair method suitable for the target object based on the target anomaly level corresponding to the target object, and then uses this information repair method to repair the target object, obtaining target repair information. This realizes information repair according to different anomaly levels, improving the effectiveness of information repair.

[0245] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, and improvements made within the spirit and scope of this application are included within the scope of protection of this application.

Claims

1. An object recognition method, characterized in that, The object recognition method includes: Acquire object operation data, object resource data, and historical resource transfer data of the object to be identified; wherein, the object resource data records the resource changes in the object to be identified, and the historical resource transfer data represents the credit status of the object to be identified over a historical period; The current abnormal transfer phase of the object to be identified is determined from the historical resource transfer data; Using the object resource data, a first probability value is predicted for the object to be identified to switch from the current transfer anomaly stage to the next transfer anomaly stage; The frequency of the operations of the object to be identified during a preset time period is calculated from the object operation data, wherein the operations of the object to be identified include logging in and going online; Using the usage frequency, a probability prediction of abnormal usage status is made for the object to be identified, and a second probability value is obtained; When the first probability value is greater than the first probability threshold, the first prediction result of the object to be identified is determined to be switching from the current transfer anomaly stage to the next transfer anomaly stage; Based on the second probability value and the second probability threshold of each of the multiple abnormal state stages of the usage state, a second prediction result of the object to be identified is selected from the multiple abnormal state stages. The first prediction result and the second prediction result are integrated into the prediction result of the object to be identified; wherein, the prediction result represents the probability that the resource transfer of the object to be identified is abnormal, and the probability that the usage status of the object to be identified is abnormal; Based on the prediction results, the target object exhibiting abnormal behavior is identified from the objects to be identified.

2. The method according to claim 1, characterized in that, The step of identifying target objects exhibiting abnormal behavior from the objects to be identified based on the prediction results includes: Based on the first and second prediction results in the prediction results, the matching anomaly level corresponding to the object to be identified is selected from multiple anomaly levels. The objects whose matching anomaly level reaches the preset anomaly level among the objects to be identified are determined as the target objects.

3. The method according to claim 1, characterized in that, After identifying the target object exhibiting abnormal behavior from the objects to be identified based on the prediction result, the method further includes: Based on the target anomaly level corresponding to the target object, an information repair method is determined for the target object; The information repair method described above is used to repair the information of the target object, thereby obtaining target repair information.

4. The method according to claim 3, characterized in that, The information repair method includes: information repair based on identity identifier; the information repair of the target object to obtain target repair information includes: Based on the identity identifier of the target object, related objects of the target object are extracted from a preset database; The operation frequency of the associated object is statistically analyzed to obtain the first activity level of the associated object; Based on the attribute information of the target object and the attribute information of the associated object, the similarity between the associated object and the target object is calculated; The object information of the N objects with the highest first activity level determined based on the first activity level of the associated object, and the object information of the N objects with the highest similarity level determined based on the similarity level with the target object, are determined as the target repair information, thereby realizing information repair based on the identity identifier; N is a positive integer.

5. The method according to claim 3, characterized in that, The information repair method includes: repairing information based on the login device identifier; the step of repairing the target object to obtain target repair information includes: Obtain the login device identifier corresponding to the target object; Based on the login device identifier, extract the same device object that is the same as the login device of the target object from the preset database; The operation frequency of the same device object is statistically analyzed to obtain the second activity level of the same device object; The object information of the N objects with the highest second activity level, determined based on the second activity level, is identified as the target repair information, thereby realizing the information repair based on the login device identifier.

6. The method according to claim 3, characterized in that, The information repair method includes: information repair based on interactive objects; the information repair of the target object to obtain target repair information includes: From the preset database, find the interactive objects that have interactive behavior with the target object; The frequency of interaction between the interactive object and the target object is statistically analyzed to obtain the interaction frequency between the interactive object and the target object. The object information of the N objects with the highest interaction frequency with the target object, determined according to the interaction frequency, is identified as the target repair information, thereby realizing the information repair based on the interaction objects.

7. The method according to claim 3, characterized in that, The information repair method includes: information repair based on social relationships; the information repair of the target object to obtain target repair information includes: From the obtained identity relationship data of the target object, candidate objects that have social relationships with the target object are extracted; The intimacy level between the candidate object and the target object is calculated based on the number of social interactions between the target object and the candidate object. The operation frequency of the candidate objects is statistically analyzed to obtain the third activity level of the candidate objects; The object information of the N objects with the highest intimacy with the target object, determined based on the intimacy between the candidate object and the target object, and the object information of the N objects with the highest third activity, determined based on the third activity level, are determined as the target repair information, thereby realizing the information repair based on social relationships.

8. The method according to claim 3, characterized in that, After using the aforementioned information repair method to repair the target object and obtain the target repair information, the method further includes: In response to the target repair information, a notification message is issued to indicate that the target object has experienced a resource transfer anomaly.

9. An object recognition device, characterized in that, The object recognition device includes: The data acquisition module is used to acquire object operation data, object resource data, and historical resource transfer data of the object to be identified; wherein, the object resource data records the resource changes in the object to be identified, and the historical resource transfer data represents the credit status of the object to be identified over a historical period. An anomaly prediction module is used to determine the current transfer anomaly stage of the object to be identified from the historical resource transfer data; predict a first probability value of the object to be identified switching from the current transfer anomaly stage to the next transfer anomaly stage using the object resource data; calculate the usage frequency of the object to be identified's operations within a preset time period from the object operation data, wherein the operations of the object to be identified include logging in and going online; predict the probability of the object to be identified having an abnormal usage status using the usage frequency to obtain a second probability value; when the first probability value is greater than a first probability threshold, determine that the first prediction result of the object to be identified is switching from the current transfer anomaly stage to the next transfer anomaly stage; based on the second probability value and the second probability thresholds of each of the multiple abnormal usage status stages, select a second prediction result of the object to be identified from the multiple abnormal usage status stages; integrate the first prediction result and the second prediction result into a prediction result of the object to be identified; wherein the prediction result represents the probability of the resource transfer of the object to be identified being abnormal and the probability of the usage status of the object to be identified being abnormal. The object recognition module is used to identify the target object exhibiting abnormal behavior from the objects to be identified based on the prediction results.

10. The apparatus according to claim 9, characterized in that, The object recognition module is further used for: Based on the first and second prediction results in the prediction results, the matching anomaly level corresponding to the object to be identified is selected from multiple anomaly levels. The objects whose matching anomaly level reaches the preset anomaly level among the objects to be identified are determined as the target objects.

11. An object recognition device, characterized in that, The object recognition device includes: Memory, used to store executable instructions; A processor, when executing executable instructions stored in the memory, implements the object identification method according to any one of claims 1 to 8.

12. A computer-readable storage medium storing executable instructions, characterized in that, When the executable instructions are executed by the processor, they implement the object recognition method according to any one of claims 1 to 8.

13. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by a processor, they implement the object recognition method according to any one of claims 1 to 8.

Images