Disclosure of Invention
The embodiment of the specification aims to provide a service processing method, a model training method, a device and equipment, so as to solve the technical problem of how to accurately verify the identity of a user when processing a service.
In order to solve the above technical problems, an embodiment of the present disclosure provides a service processing method, including:
receiving a service processing request, wherein the service processing request comprises service application information which corresponds to at least two modal characteristics;
respectively extracting modal feature data corresponding to the modal features from the service application information;
fusing the feature data of each mode to obtain fused feature data;
Judging whether the service processing request is an abnormal request or not according to the fusion characteristic data;
and processing the service corresponding to the service processing request under the condition that the service processing request is not an abnormal request.
The embodiment of the specification also provides a service processing device, which comprises:
the system comprises a request receiving module, a service processing module and a service processing module, wherein the request receiving module is used for receiving a service processing request, and the service processing request comprises service application information which corresponds to at least two modal characteristics;
The modal feature data extraction module is used for respectively extracting modal feature data corresponding to the modal features from the service application information;
the modal feature data fusion module is used for fusing the modal feature data to obtain fused feature data;
the judging module is used for judging whether the service processing request is an abnormal request or not according to the fusion characteristic data;
And the service processing module is used for processing the service corresponding to the service processing request under the condition that the service processing request is not an abnormal request.
The embodiment of the specification also provides service processing equipment which comprises a memory and a processor, wherein the memory is used for storing computer program instructions, the processor is used for executing the computer program instructions to realize the following steps of receiving a service processing request, wherein the service processing request comprises service application information, the service application information corresponds to at least two mode characteristics, respectively extracting mode characteristic data corresponding to the mode characteristics from the service application information, fusing the mode characteristic data to obtain fused characteristic data, judging whether the service processing request is an abnormal request according to the fused characteristic data, and processing a service corresponding to the service processing request under the condition that the service processing request is not the abnormal request.
In order to solve the above technical problem, an embodiment of the present disclosure further provides a model training method, including:
acquiring service application sample data, wherein the service application sample data corresponds to at least two modal characteristics;
Respectively extracting modal feature sample data corresponding to the modal features from the service application sample data;
fusing the modal feature sample data to obtain a fused sample vector;
training a pre-constructed abnormal service identification model according to the fusion sample vector.
The embodiment of the specification also provides a model training device, which comprises:
The system comprises a sample data acquisition module, a service application sample data processing module, a service application data processing module and a service application data processing module, wherein the sample data acquisition module is used for acquiring service application sample data, wherein the service application sample data is marked with abnormal conditions;
the modal feature sample data extraction module is used for respectively extracting modal feature sample data corresponding to the modal features from the service application sample data;
the sample vector fusion module is used for fusing the modal characteristic sample data to obtain a fused sample vector;
And the model training module is used for training an abnormal service identification model according to the fusion sample vector and the corresponding abnormal condition.
The embodiment of the specification also provides model training equipment, which comprises a memory and a processor;
the memory is used for storing computer program instructions;
The processor is used for executing the computer program instructions to achieve the following steps of obtaining service application sample data, marking abnormal conditions on the service application sample data, enabling the service application sample data to correspond to at least two modal characteristics, respectively extracting modal characteristic sample data corresponding to the modal characteristics from the service application sample data, fusing the modal characteristic sample data to obtain fused sample vectors, and training an abnormal service identification model according to the fused sample vectors and the corresponding abnormal conditions.
As can be seen from the technical solutions provided in the embodiments of the present disclosure, service application information corresponding to at least two modal features is obtained, modal feature data corresponding to the modal features are extracted respectively, and then the modal feature data are fused, and whether a service processing request is an abnormal request is determined by using the fused feature data. By the method, when judging service application information, the plurality of modal characteristics can be comprehensively considered, the situation that the data type is single can be avoided by identifying the data fused with the plurality of modal characteristics, and the fused data has rich characteristics, so that the difficulty of forging the data is increased, and the accuracy of verifying the user application is improved.
Detailed Description
The technical solutions of the embodiments of the present specification will be clearly and completely described below with reference to the drawings in the embodiments of the present specification, and it is apparent that the described embodiments are only some embodiments of the present specification, not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are intended to be within the scope of the present disclosure.
In order to solve the above technical problems, first, an embodiment of a model training method according to the present embodiment will be described with reference to fig. 1. The execution subject of the method may be a model training device including, but not limited to, a server, an industrial computer, a PC, etc. The method comprises the following specific implementation steps.
S110, acquiring service application sample data, wherein the service application sample data correspond to at least two modal characteristics.
The service application sample data is sample data obtained by training an abnormal service identification model. The service application sample data may be sample data provided for a service to be identified, for example, when the service is a credit card application service, service application information submitted by a user in a history record may be collected, and the service application information is used as the service application sample data.
The service application sample data may correspond to an abnormal condition label, where the abnormal condition label is used to label whether the service application corresponding to the service application sample data is an abnormal application. Correspondingly, the abnormal condition labeling and service application sample data can be based on the follow-up model training by adopting a supervised learning method. The service application sample data can also be provided with no corresponding label, and the preset classifier is directly trained by using an unsupervised learning method in the subsequent training. The specific machine learning manner may be selected according to the actual situation requirement, which is not described herein.
The modal characteristics are used for representing the types corresponding to the data contained in the service application sample data. In some implementations, the modal features can include at least two of image modal features, fingerprint modal features, audio modal features, and text modal features. The image mode characteristics may correspond to image data, such as photographs, certificate images, etc. of the applicant, the fingerprint mode characteristics may correspond to fingerprint data provided by the applicant, such as fingerprints acquired by a terminal device, the audio mode characteristics may correspond to audio sent by the applicant, such as voice instructions preset by the applicant, or voice information sent by the applicant itself, and the text mode characteristics may correspond to text submitted by the applicant, such as text information of application materials, personal identity information, etc. The above examples are only for introducing the modal feature, and other information types may be selected as the modal feature in practical applications, which is not described herein.
And S120, respectively extracting the modal characteristic sample data corresponding to the modal characteristics from the service application sample data.
The modal characteristic sample data is the data corresponding to the modal characteristics. For example, when the modal feature is an image modal feature, an image in the modal feature sample data may be extracted as modal feature sample data, and when the modal feature is an audio modal feature, an audio file in the modal feature sample data may be extracted as modal feature sample data. The specific method for extracting the modal feature sample data can be adjusted according to the actual application requirement, and will not be described in detail.
In some embodiments, when the mode feature sample data is extracted, data extraction can be performed on feature dimensions with stronger business meaning, for example, when text mode feature sample data is extracted, text information corresponding to certain specific contents can be selected as the text mode feature sample data, and when data corresponding to image mode features is selected, pictures with stronger business meaning, such as certificate images and the like, can be identified. By combining the business meaning of the modal feature sample data to screen, the effectiveness of the selected modal feature sample data is improved, and the recognition accuracy of the trained model is improved.
S130, fusing the modal characteristic sample data to obtain a fused sample vector.
After the modal feature sample data corresponding to at least two modal features is acquired, the modal feature sample data may be fused. The specific fusion method can be to fuse the mode characteristic sample data into a characteristic vector, and then map the fused characteristic vector into a preset vector subspace, so as to obtain a fused sample vector. The above method is only an exemplary introduction to the process of fusing the modal feature sample data, and other methods for obtaining the fused sample vector may be selected according to the needs in practical application, which is not described herein.
In some embodiments, the modal feature sample data may be preprocessed before being fused, so as to reduce noise in the modal feature sample data and improve accuracy of training results. The preprocessing may include denoising processing and normalization processing.
And denoising, namely removing noise sample data in the modal characteristic sample data according to the modal characteristic sample data. Specifically, the noise sample data can be determined in the modal feature sample data according to a preset noise data template so as to be removed, or the noise sample data can be directly screened out from the modal feature sample data based on the characteristics of continuity and the like of the modal feature sample data. The specific manner of determining the noise sample data may be adjusted according to the actual application requirement, and is not limited to the above example, and will not be described herein.
The standardization process can adjust the modal characteristic sample data to the same data distribution range, so that the sample data are distributed consistently, and the processing of the modal characteristic sample data in the subsequent process is facilitated. The normalization process may be, for example, a normalization process, and divides the data into distribution ranges of [0,1], so that the application of the data is more convenient.
And S140, training a pre-constructed abnormal service identification model according to the fusion sample vector.
The abnormal traffic recognition model may be a model for recognizing abnormal traffic. The abnormal service identification model may be a mathematical model, which is used to identify whether the corresponding service is an abnormal service according to different modal characteristics of the service application data. The abnormal service identification model may be a bayesian classification model, a support vector machine classification model (Support Vector Machine, SVM), a convolutional neural network classification model (Convolutional Neural Networks, CNN), or the like, which is not limited.
After the fusion sample vector is obtained, the abnormal service identification model can be trained according to the fusion sample vector, so that the trained model can identify the abnormal condition of the service based on the training sample.
In some embodiments, multiple abnormal service identification models may be pre-built, and training of the abnormal service identification models may be performed based on the same service application sample data. For the abnormal service identification models, verification data are respectively input into the abnormal service identification models to obtain verification results corresponding to the models. And judging the accuracy of each abnormal service identification model according to the verification result, and determining the final abnormal service identification model applied to service processing according to the accuracy of different models.
According to the model training method, the service application information corresponding to at least two modal characteristics is acquired, the modal characteristic data corresponding to the modal characteristics are respectively extracted, then the modal characteristic data are fused, and the fused characteristic data are utilized to train the abnormal service identification model, so that the abnormal service identification model can identify the abnormal condition of the service. By the method, when judging service application information, the plurality of modal characteristics can be comprehensively considered, the situation that the data type is single can be avoided by identifying the data fused with the plurality of modal characteristics, and the fused data has rich characteristics, so that the difficulty of forging the data is increased, and the accuracy of verifying the user application is improved.
Based on the embodiment of the model training method, the embodiment of the specification also provides a service processing method. The execution main body of the service processing method is service processing equipment, and the service processing equipment comprises, but is not limited to, a server, an industrial personal computer, a PC (personal computer) and the like. As shown in fig. 2, the method includes the following specific implementation steps.
S210, receiving a service processing request, wherein the service processing request comprises service application information, and the service application information corresponds to at least two modal characteristics.
The service processing request is a request provided by the user for processing a specific service, for example, when the user needs to apply for a credit card, the service processing request may be a request for applying for a credit card.
The service processing request contains service application information, which may be data provided corresponding to the service to be processed, for example, may include personal identity information, certificate information, biometric information, etc. of the user.
The modal characteristics are used for representing the types corresponding to the data contained in the service application sample data. In some implementations, the modal features can include at least two of image modal features, fingerprint modal features, audio modal features, and text modal features. The image mode characteristics may correspond to image data, such as photographs, certificate images, etc. of the applicant, the fingerprint mode characteristics may correspond to fingerprint data provided by the applicant, such as fingerprints acquired by a terminal device, the audio mode characteristics may correspond to audio sent by the applicant, such as voice instructions preset by the applicant, or voice information sent by the applicant itself, and the text mode characteristics may correspond to text submitted by the applicant, such as text information of application materials, personal identity information, etc. The above examples are only for introducing the modal feature, and other information types may be selected as the modal feature in practical applications, which is not described herein.
S220, respectively extracting the modal characteristic data corresponding to the modal characteristics from the service application information.
The modal characteristic data is the data corresponding to the modal characteristic. For example, when the modal feature is an image modal feature, an image in the modal feature data may be extracted as modal feature data, and when the modal feature is an audio modal feature, an audio file in the modal feature data may be extracted as modal feature data. The specific method for extracting the modal feature data can be adjusted according to the actual application requirement, and will not be described in detail.
In some embodiments, the data extraction may be performed for feature dimensions with strong business meaning when extracting the modal feature data, for example, when extracting text modal feature data, text information corresponding to certain specific content may be selected as the text modal feature data, and when selecting data corresponding to image modal feature, a picture with strong business meaning, for example, a certificate image, etc., may be identified from the text information. By combining the service meaning of the modal feature data to screen, the effectiveness of the selected modal feature data is improved, and the recognition accuracy of the trained model is improved.
S230, fusing the feature data of each mode to obtain fused feature data.
After the modal feature sample data corresponding to at least two modal features is acquired, the modal feature sample data may be fused. The specific fusion method can be to fuse the mode characteristic sample data into a characteristic vector, and then map the fused characteristic vector into a preset vector subspace, so as to obtain a fused sample vector. The above method is only an exemplary introduction to the process of fusing the modal feature sample data, and other methods for obtaining the fused sample vector may be selected according to the needs in practical application, which is not described herein.
In some embodiments, the modal feature sample data may be preprocessed before being fused, so as to reduce noise in the modal feature sample data and improve accuracy of training results. The preprocessing may include denoising processing and normalization processing.
And denoising, namely removing noise sample data in the modal characteristic sample data according to the modal characteristic sample data. Specifically, the noise sample data can be determined in the modal feature sample data according to a preset noise data template so as to be removed, or the noise sample data can be directly screened out from the modal feature sample data based on the characteristics of continuity and the like of the modal feature sample data. The specific manner of determining the noise sample data may be adjusted according to the actual application requirement, and is not limited to the above example, and will not be described herein.
The standardization process can adjust the modal characteristic sample data to the same data distribution range, so that the sample data are distributed consistently, and the processing of the modal characteristic sample data in the subsequent process is facilitated. The normalization process may be, for example, a normalization process, and divides the data into distribution ranges of [0,1], so that the application of the data is more convenient.
S240, judging whether the service processing request is an abnormal request according to the fusion characteristic data.
In some embodiments, determining whether the service processing request is an abnormal request according to the converged feature data may be to identify whether the service processing request corresponding to the converged feature data is an abnormal request by using an abnormal service identification model. The abnormal service identification model may be a model trained based on the corresponding embodiment of fig. 1. The description of the abnormal service identification model may refer to the description in the corresponding embodiment of fig. 1, and will not be repeated herein.
S250, processing the service corresponding to the service processing request under the condition that the service processing request is not an abnormal request.
If the service processing request is identified to be not an abnormal request based on the fusion characteristic data, the service processing request is indicated to be a normal request, no higher risk exists in processing the service processing request, and the service can be processed based on a normal processing flow. The specific processing manner can be combined with the characteristics of the service in the practical application to process the service, which is not described herein.
If the service processing request is identified to be an abnormal request based on the fusion characteristic data, the condition that a certain risk exists in the service corresponding to the service processing request is indicated, the data or funds are easy to leak is easily caused, the service processing request can be refused to be processed, and corresponding refusal information is fed back. The specific processing manner may also be set according to the requirements in practical applications, and will not be described herein.
According to the service processing method, the service application information corresponding to at least two modal characteristics is obtained, the modal characteristic data corresponding to the modal characteristics are respectively extracted, then the modal characteristic data are fused, and whether the service processing request is an abnormal request is judged by utilizing the fused characteristic data. By the method, when judging service application information, the plurality of modal characteristics can be comprehensively considered, the situation that the data type is single can be avoided by identifying the data fused with the plurality of modal characteristics, and the fused data has rich characteristics, so that the difficulty of forging the data is increased, and the accuracy of verifying the user application is improved.
Based on the embodiment of the model training method, the embodiment of the present specification further provides a model training device, where the model training device may be integrated with the model training apparatus, as shown in fig. 3, and the model training device may include the following modules.
The system comprises a sample data acquisition module 310, a service application sample data processing module and a service application data processing module, wherein the sample data acquisition module 310 is used for acquiring service application sample data, wherein the service application sample data is marked with abnormal conditions;
a modal feature sample data extraction module 320, configured to extract modal feature sample data corresponding to the modal features from the service application sample data, respectively;
a sample vector fusion module 330, configured to fuse the modal feature sample data to obtain a fused sample vector;
the model training module 340 is configured to train an abnormal service identification model according to the fused sample vector and the corresponding abnormal situation.
Based on the above embodiments of the service processing method, the embodiments of the present disclosure further provide a service processing apparatus, where the service processing apparatus may be integrated with the service processing device, as shown in fig. 4, and the service processing apparatus may include the following modules.
A request receiving module 410, configured to receive a service processing request, where the service processing request includes service application information, and the service application information corresponds to at least two modal features;
a modal feature data extraction module 420, configured to extract modal feature data corresponding to the modal features from the service application information respectively;
the modal feature data fusion module 430 is configured to fuse the modal feature data to obtain fused feature data;
a judging module 440, configured to judge whether the service processing request is an abnormal request according to the fusion feature data;
And the service processing module 450 is configured to process a service corresponding to the service processing request if the service processing request is not an abnormal request.
Based on the above model training method, as shown in fig. 5, the embodiment of the present disclosure further provides a model training device. The model training apparatus may include a memory and a processor.
In this embodiment, the memory may be implemented in any suitable manner. For example, the memory may be a read-only memory, a mechanical hard disk, a solid state hard disk, or a usb disk. The memory may be used to store computer program instructions.
In this embodiment, the processor may be implemented in any suitable manner. For example, a processor may take the form of, for example, a microprocessor or processor, and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, application SPECIFIC INTEGRATED Circuits (ASICs), programmable logic controllers, and embedded microcontrollers, among others.
The processor can execute the computer program instructions to realize the following steps of obtaining service application sample data, marking abnormal conditions on the service application sample data, enabling the service application sample data to correspond to at least two modal characteristics, respectively extracting modal characteristic sample data corresponding to the modal characteristics from the service application sample data, fusing the modal characteristic sample data to obtain fused sample vectors, and training an abnormal service identification model according to the fused sample vectors and the corresponding abnormal conditions.
Based on the above service processing method, as shown in fig. 6, the embodiment of the present disclosure further provides a service processing device. The traffic processing device may include a memory and a processor.
In this embodiment, the memory may be implemented in any suitable manner. For example, the memory may be a read-only memory, a mechanical hard disk, a solid state hard disk, or a usb disk. The memory may be used to store computer program instructions.
In this embodiment, the processor may be implemented in any suitable manner. For example, a processor may take the form of, for example, a microprocessor or processor, and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, application SPECIFIC INTEGRATED Circuits (ASICs), programmable logic controllers, and embedded microcontrollers, among others.
The processor can execute the computer program instructions to realize the following steps of receiving a service processing request, wherein the service processing request comprises service application information, the service application information corresponds to at least two mode characteristics, respectively extracting mode characteristic data corresponding to the mode characteristics from the service application information, fusing the mode characteristic data to obtain fused characteristic data, judging whether the service processing request is an abnormal request according to the fused characteristic data, and processing a service corresponding to the service processing request under the condition that the service processing request is not the abnormal request.
It should be noted that the service processing method, the model training method, the device and the equipment disclosed in the embodiments of the present disclosure may be used in the technical field of artificial intelligence to implement verification and processing of service applications, and of course, the service processing method, the model training method, the device and the equipment may also be applied in other fields, which are not limited.
In the 90 s of the 20 th century, improvements to one technology could clearly be distinguished as improvements in hardware (e.g., improvements to circuit structures such as diodes, transistors, switches, etc.) or software (improvements to the process flow). However, with the development of technology, many improvements of the current method flows can be regarded as direct improvements of hardware circuit structures. Designers almost always obtain corresponding hardware circuit structures by programming improved method flows into hardware circuits. Therefore, an improvement of a method flow cannot be said to be realized by a hardware entity module. For example, a programmable logic device (Programmable Logic Device, PLD) (e.g., field programmable gate array (Field Programmable GATE ARRAY, FPGA)) is an integrated circuit whose logic functions are determined by user programming of the device. A digital system is "integrated" on a PLD by the designer's own programming without requiring the chip manufacturer to design and fabricate application-specific integrated circuit chips 2. Moreover, nowadays, instead of manually manufacturing integrated circuit chips, such programming is mostly implemented with "logic compiler (logic compiler)" software, which is similar to the software compiler used in program development and writing, and the original code before being compiled is also written in a specific programming language, which is called hardware description language (Hardware Description Language, HDL), but HDL is not just one, but a plurality of kinds, such as ABEL(Advanced Boolean Expression Language)、AHDL(Altera Hardware Description Language)、Confluence、CUPL(Cornell University Programming Language)、HDCal、JHDL(Java Hardware Description Language)、Lava、Lola、MyHDL、PALASM、RHDL(Ruby Hardware Description Language), and VHDL (Very-High-SPEED INTEGRATED Circuit Hardware Description Language) and Verilog2 are currently most commonly used. It will also be apparent to those skilled in the art that a hardware circuit implementing the logic method flow can be readily obtained by merely slightly programming the method flow into an integrated circuit using several of the hardware description languages described above.
The system, apparatus, module or unit set forth in the above embodiments may be implemented in particular by a computer chip or entity, or by a product having a certain function. One typical implementation is a computer. In particular, the computer may be, for example, a personal computer, a laptop computer, a cellular telephone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.
From the above description of embodiments, it will be apparent to those skilled in the art that the present description may be implemented in software plus a necessary general purpose hardware platform. Based on this understanding, the technical solution of the present specification may be embodied in essence or a part contributing to the prior art in the form of a software product, which may be stored in a storage medium, such as a ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the embodiments or some parts of the embodiments of the present specification.
In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments. In particular, for system embodiments, since they are substantially similar to method embodiments, the description is relatively simple, as relevant to see a section of the description of method embodiments.
The specification is operational with numerous general purpose or special purpose computer system environments or configurations. Such as a personal computer, a server computer, a hand-held or portable device, a tablet device, a multiprocessor system, a microprocessor-based system, a set top box, a programmable consumer electronics, a network PC, a minicomputer, a mainframe computer, a distributed computing environment that includes any of the above systems or devices, and the like.
The description may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.
Although the present specification has been described by way of example, it will be appreciated by those skilled in the art that there are many variations and modifications to the specification without departing from the spirit of the specification, and it is intended that the appended claims encompass such variations and modifications as do not depart from the spirit of the specification.