Intelligent result data screening method and system based on multi-channel neural network

Abstract

The invention relates to the technical field of artificial intelligence and deep learning, and discloses an intelligent result data screening method and system based on a multi-channel neural network. The method comprises the following steps: training a multi-channel RPNN neural network model by using a gradient algorithm: taking historical achievement data used in a talent screening process as input data, and performing model training on the multi-channel RPNN neural network model by using the gradient algorithm to obtain an optimal weight parameter and judge whether the achievement data is an excellent reference threshold; an input layer of the multi-channel RPNN neural network model is expanded from one channel to at least two channels; the collected real-time result data is input into the trained multi-channel RPNN neural network model to output a calculation result, the real-time result data is multi-dimensional data or multi-modal data, and each dimension or modal corresponds to one channel; when the calculation result is greater than or equal to a reference threshold value, determining that the real-time result data is excellent data; otherwise, determining the data as general data.

Description

Technical Field

[0001] The present application relates to the fields of artificial intelligence and deep learning technology, and in particular to a method and system for intelligent screening of achievement data based on a multi-channel neural network. Background Art

[0002] In the process of digital transformation, data governance is one of the most critical links. Data governance focuses on the entire life cycle of data management, including data planning, data acquisition, data quality, data sharing, data labeling, etc., and is the key support for intelligent decision-making applications in various fields.

[0003] In data governance, for the achievement data, these achievement data are usually various types of data such as patents, papers, awards and projects. At the same time, the achievement data can be presented in different data formats such as text, video and audio. Usually, manual screening is used to split the data types and data formats, and the split data are further screened and summarized according to a single type or a single data format using a traditional neural network model, and then the achievement data is manually evaluated. The screening and evaluation process of the above achievement data requires a lot of manpower. In addition, since the traditional neural network model only consists of summation neurons, it cannot process data of different types and modalities in parallel, and can only screen a single type or a single data format. Therefore, for these multi-dimensional and multi-modal achievement data, multiple repeated screenings are required to complete the screening of all achievement data, which is time-consuming. Combined with the time-consuming manual screening, the overall screening efficiency is ultimately low. In addition, due to the inability to parallel processing, the correlation between data of different types and modalities cannot be identified, resulting in low screening accuracy. Therefore, how to use artificial intelligence methods to screen reasonably and effectively to improve screening efficiency and accuracy is an urgent problem that needs to be solved in the current talent screening process. It also determines whether the evaluation of the screening results data can provide an important basis for objective evaluation criteria for talent assessment, and becomes an indispensable part of the work of assessing outstanding talents. Summary of the invention

[0004] In order to overcome the above-mentioned technical defects, the present application provides an intelligent screening method and system for achievement data based on a multi-channel neural network, which has the ability to process multidimensional data or multimodal data and can ensure a high accuracy of data processing and data analysis.

[0005] In a first aspect, the present application provides a method for intelligently screening achievement data based on a multi-channel neural network, the method comprising:

[0006] S1. Use gradient algorithm to train multi-channel RPNN neural network model:

[0007] The historical achievement data used in the talent screening process is used as input data, and the multi-channel RPNN neural network model is trained using a gradient algorithm to obtain the optimal weight parameters and the reference threshold for judging whether the achievement data is excellent. The input layer of the multi-channel RPNN neural network model is expanded from one channel to at least two channels, and each channel corresponds to a dimension or a modality. The historical achievement data is multi-dimensional data or multi-modal data, and the multi-dimensional data includes patents, papers, awards and project participation. The multi-modal data includes text, video and audio data;

[0008] S2. Intelligent screening of real-time results data:

[0009] Input the collected real-time achievement data into the trained multi-channel RPNN neural network model to output the calculation results, wherein the real-time achievement data is multi-dimensional data or multi-modal data, and each dimension or modality corresponds to a channel;

[0010] S3. Determine excellent data based on calculation results:

[0011] When the calculation result is greater than or equal to the reference threshold, the real-time achievement data is determined to be excellent data; otherwise, it is determined to be general data.

[0012] In the second aspect, another aspect of the present application provides an intelligent screening system for achievement data based on a multi-channel neural network, which is used to execute the steps in the method described in any one of the first aspect and its implementation methods.

[0013] This application adopts the above technical solution to achieve the following technical effects: First, since the input layer uses at least two channels, each channel corresponds to a dimension or a mode, so that multi-modal and multi-dimensional results data can be input into the model at the same time; secondly, compared with the traditional neural network model which only contains summation neurons The multi-channel RPNN neural network model introduces a high-order neural network model and contains summation neurons and quadrature neurons The network structure has a strong nonlinear processing capability. The data of the same dimension or the same mode are usually linear, but the data of different dimensions, different modes, or one dimension and one mode usually have nonlinear relationships. and quadrature neurons Data of different dimensions or different modalities can be processed in parallel to improve processing efficiency. While processing in parallel, linear and nonlinear relationships between data can also be processed to improve the accuracy of data processing. Finally, after the screening of the results data is completed, the reference threshold of the multi-channel RPNN neural network model can be used to evaluate whether the results data is excellent data. Therefore, the technical solution of this application can ensure a high accuracy of data processing and data analysis, and introduce a multi-channel mode, which has the ability to process multidimensional data or multi-modal data. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The accompanying drawings exemplarily illustrate the embodiments and constitute a part of the specification, and together with the text description of the specification, are used to explain the exemplary implementation of the embodiments. The embodiments shown are for illustrative purposes only and do not limit the scope of the claims. In all drawings, the same reference numerals refer to similar but not necessarily identical elements.

[0015] Figure 1 A screening process diagram of a method for intelligent screening of achievement data based on a multi-channel neural network in an embodiment of the present application;

[0016] Figure 2 A flowchart of the application of intelligent screening methods for achievement data in the talent screening process;

[0017] Figure 3 A schematic diagram of the Pi-Sigma neural network structure;

[0018] Figure 4 A structural diagram of the Ridge Polynomial neural network structure;

[0019] Figure 5 A schematic diagram of multi-channel input in a Ridge Polynomial neural network;

[0020] Figure 6 It is a screening process diagram of the intelligent screening system for achievement data based on a multi-channel neural network in an embodiment of the present application. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.

[0022] It should be noted that the descriptions involving "first", "second", etc. in the embodiments of the present application are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0023] In the description of the present application, it should be understood that the numerical labels before the steps do not indicate the order in which the steps are executed, but are only used to facilitate the description of the present application and to distinguish each step, and therefore should not be understood as a limitation on the present application.

[0024] The neurons contained in the traditional neural network are summation neurons However, in practical applications, many problems that need to be solved are nonlinear, which greatly limits the application of the network. Scholars have proposed another network structure that also contains summation neurons. and quadrature neurons This kind of high-order neural network has strong nonlinear capabilities and is more suitable for practical problems.

[0025] Embodiment 1

[0026] In order to screen the performance data in the talent screening process, this application introduces a high-order neural network model into the neural network model, which includes both summation neurons. and quadrature neurons , and further introduced a multi-channel mode, which has the ability to process multi-dimensional data or multi-modal data.

[0027] like Figure 1 As shown, the intelligent screening method of achievement data based on multi-channel neural network includes:

[0028] S1. Use gradient algorithm to train multi-channel RPNN neural network model:

[0029] The historical achievement data used in the talent screening process is used as input data, and the gradient algorithm is used to train the multi-channel RPNN neural network model to obtain the optimal weight parameters and the reference threshold for judging whether the achievement data is excellent. The input layer of the multi-channel RPNN neural network model is expanded from one channel to at least two channels.

[0030] Optionally, in some embodiments, the multi-channel RPNN neural network model includes summation neurons and product neurons, and the product neurons are obtained by forming a Pi-sigma neural network (PSNN).

[0031] S2. Intelligent screening of real-time results data:

[0032] The collected real-time performance data is input into the trained multi-channel RPNN neural network model to output the calculation results. The real-time performance data is multi-dimensional data or multi-modal data, and each dimension or each modality corresponds to a channel.

[0033] Specifically, it is the achievement data of an individual, which is multi-dimensional, such as patents, papers, awards, reference projects, etc., or multi-modal, such as text, video, audio, etc. Each dimensional data or each modal data corresponds to a channel.

[0034] Optionally, in some embodiments, the number of neuron nodes in the input layer, hidden layer, and output layer in the multi-channel RPNN neural network model is N, K, and 1, respectively, the weight between the hidden layer and the output layer is 1, d represents the channel, and its value range is [1, D], M1, M2…M d …M D are the number of input nodes of the input layer of the dth channel respectively. The number of input nodes between each channel can be different. The input data x of the input layer in the channel RPNN neural network model is:

[0035] ;

[0036] The corresponding weight parameter w in the hidden layer is:

[0037] ; ;

[0038] in, represents the weights corresponding to the input layer and hidden layer of the dth channel, Represents the weight between the kth node of the input layer and the jth node of the hidden layer of the dth channel.

[0039] S3. Determine priority data based on calculation results:

[0040] When the calculation result is greater than or equal to the reference threshold, the real-time achievement data is determined to be excellent data; otherwise, it is determined to be general data.

[0041] For example, Figure 2 As shown in the screening flow chart, the real-time results data is input into the multi-channel RPNN neural network with fixed weights. , and the output value is calculated to be y.

[0042] Then make the following screening judgments:

[0043] when , indicating that the person's performance data is excellent, and the corresponding person is included in the potential talent pool; when , indicating that the person's achievement data is general data and no subsequent operations will be performed.

[0044] In the embodiment of the present application, compared with the traditional neural network model which only includes summation neurons The multi-channel RPNN neural network model introduces a high-order neural network model and contains summation neurons and quadrature neurons The network structure has super nonlinear processing capabilities, which can ensure high accuracy of data processing and data analysis, and introduces a multi-channel mode, which has the ability to process multidimensional data or multimodal data.

[0045] Embodiment 2

[0046] Ridge Polynomial Neural Network (RPNN) is a high-order neural network formed by combining Pi-Sigma Neural Network (PSNN). It inherits the fast learning ability of Pi-Sigma neural network and can also approximate any continuous function. It has been studied and applied to a certain extent.

[0047] See also Figure 3 and Figure 4 The network structure of the RPNN neural network in the embodiment of the present application is described as follows:

[0048] Figure 3 The figure shows the structure of Pi-Sigma neural network, where the number of neuron nodes in each layer of Pi-Sigma neural network is N, K and 1 respectively, the weight between hidden layer and output layer is 1, and the input sample is , represents the threshold value, represents the weight between the kth node of the input layer and the jth node of the hidden layer, , represents the weight connected to the threshold node. The output of the unit is:

[0049] ;

[0050] The final output of the Pi-Sigma neural network is:

[0051] ;

[0052] in is the activation function.

[0053] Figure 4 As shown in the figure, in the hidden layer of the Ridge Polynomial neural network, each node is a PSNN. , the subscript indicates that it contains i hidden nodes. Assume that the number of input layer nodes of RPNN is d, the number of hidden layer nodes is k, and the number of output layer nodes is 1. is the weight between the kth hidden layer node and the jth input node in the i-th PSNN, Indicates each The output of RPNN is:

[0054] ; Where g is the activation function.

[0055] Embodiment 3

[0056] Next, the technical solution for multi-channel expansion of the input layer of the Ridge Polynomial neural network is described. For the Ridge Polynomial neural network, the initial input layer is only one-dimensional and can only input one type of data at a time. In order to be able to process and analyze multi-dimensional data or multi-modal data at one time, the network input layer is improved and expanded, that is, the one-channel input is converted into a multi-channel input.

[0057] like Figure 5 As shown, D represents the number of channels, M1, M2…M d …M D are the number of input nodes of the input layer of the dth channel, and the number of input nodes between each channel can be different. Through multi-channel expansion, the input value of the entire network is changed to:

[0058] ;

[0059] The corresponding weight changes are:

[0060] ;

[0061] ;

[0062] in, represents the weights corresponding to the input layer and hidden layer of the dth channel, Represents the weight between the kth node of the input layer and the jth node of the hidden layer of the dth channel.

[0063] Embodiment 4

[0064] Next, the training process of the multi-channel Ridge Polynomial neural network model in the embodiment of the present application is introduced as follows:

[0065] The gradient algorithm is used to train the multi-channel Ridge Polynomial neural network model. Assume that the historical results data in the talent screening process is recorded as the sample set , then the error function of RPNN output is:

[0066] ;

[0067] remember:

[0068]

[0069] From the above two formulas, we can get:

[0070] ;

[0071] The ultimate goal of using the gradient algorithm to train the network is to find a ,make , right Find the gradient:

[0072] ;

[0073] During the training process, at the mth iteration, the weights are updated as follows:

[0074] ;

[0075] Weight update variables for:

[0076] ;

[0077] ;

[0078] is the learning rate.

[0079] Through training, we get the optimal weights And judgment threshold ;

[0080] Further screening and judgment:

[0081] when , indicating that the person's achievement data is excellent, and the corresponding person is included in the potential talent pool;

[0082] when , indicating that the person's achievement data is general data and no subsequent operations will be performed.

[0083] Embodiment 5

[0084] Figure 6 The schematic diagram of the hardware architecture of a computer device 10000 suitable for implementing the intelligent screening method for achievement data according to the third embodiment of the present application is schematically shown. In some embodiments, the computer device 10000 may be a terminal device such as a smart phone, a wearable device, a tablet computer, a personal computer, a vehicle terminal, a game console, a virtual device, a workbench, a digital assistant, a set-top box, a robot, etc. In other embodiments, the computer device 10000 may be a rack server, a blade server, a tower server, or a cabinet server (including an independent server, or a server cluster composed of multiple servers), etc. Figure 6 As shown, the computer device 10000 includes but is not limited to: a memory 10010, a processor 10020, and a network interface 10030 that can communicate with each other through a system bus. Among them:

[0085] The memory 10010 includes at least one type of computer-readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (such as SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 10010 can be an internal storage module of the computer device 10000, such as a hard disk or memory of the computer device 10000. In other embodiments, the memory 10010 can also be an external storage device of the computer device 10000, such as a plug-in hard disk equipped on the computer device 10000, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. Of course, the memory 10010 can also include both the internal storage module of the computer device 10000 and its external storage device. In this embodiment, the memory 10010 is generally used to store the operating system and various application software installed in the computer device 10000, such as the program code of the achievement data intelligent screening method, etc. In addition, the memory 10010 can also be used to temporarily store various data that have been output or will be output.

[0086] In some embodiments, the processor 10020 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other chips. The processor 10020 is generally used to control the overall operation of the computer device 10000, such as performing control and processing related to data interaction or communication with the computer device 10000. In this embodiment, the processor 10020 is used to run the program code stored in the memory 10010 or process data.

[0087] The network interface 10030 may include a wireless network interface or a wired network interface, and the network interface 10030 is generally used to establish a communication link between the computer device 10000 and other computer devices. For example, the network interface 10030 is used to connect the computer device 10000 to an external terminal through a network, and to establish a data transmission channel and a communication link between the computer device 10000 and the external terminal. The network may be a wireless or wired network such as an intranet, the Internet, the Global System of Mobile communication (GSM), Wideband Code Division Multiple Access (WCDMA), 4G network, 5G network, Bluetooth, Wi-Fi, etc.

[0088] It should be pointed out that Figure 6 Only a computer device having components 10010 - 10030 is shown, but it should be understood that implementation of all of the components shown is not a requirement, and more or fewer components may alternatively be implemented.

[0089] In this embodiment, the intelligent screening method of achievement data stored in the memory 10010 can also be divided into one or more program modules and executed by one or more processors (such as processor 10020) to complete the embodiment of the present application.

[0090] Embodiment 6

[0091] An embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored, wherein when the computer program is executed by a processor, the steps of the method for intelligent screening of achievement data in the embodiment are implemented.

[0092] In this embodiment, the computer-readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the computer-readable storage medium can be an internal storage unit of a computer device, such as a hard disk or memory of the computer device. In other embodiments, the computer-readable storage medium can also be an external storage device of a computer device, such as a plug-in hard disk equipped on the computer device, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. Of course, the computer-readable storage medium can also include both the internal storage unit of the computer device and its external storage device. In this embodiment, the computer-readable storage medium is generally used to store an operating system and various application software installed on the computer device, such as the program code of the intelligent screening method of achievement data in the embodiment. In addition, the computer-readable storage medium can also be used to temporarily store various types of data that have been output or are to be output.

[0093] Embodiment 7

[0094] An embodiment of the present application also provides a computer program product, including a computer program, which implements the method in the above embodiment when executed by a processor.

[0095] Obviously, those skilled in the art should understand that the modules or steps of the above-mentioned embodiments of the present application can be implemented by general-purpose computer devices, they can be concentrated on a single computer device, or distributed on a network composed of multiple computer devices, optionally, they can be implemented by executable program codes of computer devices, so that they can be stored in a storage device and executed by the computer device, and in some cases, the steps shown or described can be executed in a different order from that herein, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. In this way, the embodiments of the present application are not limited to any specific combination of hardware and software.

[0096] It should be noted that the above are only preferred embodiments of the present application, and the patent protection scope of the present application is not limited thereto. Any equivalent structure or equivalent process transformation made using the contents of the specification and drawings of the present application, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for intelligent screening of achievement data based on multi-channel neural network, characterized in that: The following steps are involved: S1. Use gradient algorithm to train multi-channel RPNN neural network model: The historical achievement data used in the talent screening process is used as input data, and the multi-channel RPNN neural network model is trained using a gradient algorithm to obtain the optimal weight parameters and the reference threshold for judging whether the achievement data is excellent. The input layer of the multi-channel RPNN neural network model is expanded from one channel to at least two channels, and each channel corresponds to a dimension or a modality. The historical achievement data is multi-dimensional data or multi-modal data, and the multi-dimensional data includes patents, papers, awards and project participation. The multi-modal data includes text, video and audio data; S2. Intelligent screening of real-time results data: Input the collected real-time achievement data into the trained multi-channel RPNN neural network model to output the calculation results, wherein the real-time achievement data is multi-dimensional data or multi-modal data, and each dimension or modality corresponds to a channel; S3. Determine excellent data based on calculation results: When the calculation result is greater than or equal to the reference threshold, the real-time achievement data is determined to be excellent data; otherwise, it is determined to be general data.

2. The method for intelligent screening of achievement data according to claim 1, characterized in that: The multi-channel RPNN neural network model includes summation neurons and quadrature neurons. The quadrature neurons are composed of Pi-sigma neural network PSNN. In the hidden layer of the RPNN neural network, each node is a PSNN. , the subscript indicates that it contains i hidden nodes. Assuming that the number of input layer nodes of RPNN is d, the number of hidden layer nodes is k, and the number of output layer nodes is 1, is the weight between the kth hidden layer node and the jth input node in the i-th PSNN, Indicates each The output of RPNN is: ; Where g is the activation function.

3. The method for intelligent screening of achievement data according to claim 1 or 2, characterized in that: The total number of channels in the multi-channel RPNN neural network model is D, denoted by d as the channel, and its value range is [1, D], M1, M2…M d …M D are the number of input nodes of the input layer of the dth channel respectively. The number of input nodes between each channel can be different. The input data x of its input layer is: ； The corresponding weight parameter w in the hidden layer is: ； ； in, represents the weights corresponding to the input layer and hidden layer of the dth channel, Represents the weight between the kth node of the input layer and the jth node of the hidden layer of the dth channel.

4. The method for intelligent screening of achievement data according to claim 3, characterized in that: During the model training process, the input sample is recorded as , then the error of the RPNN neural network model output is for: ； remember: ; The above two formulas finally give: ; Calculate the error using the gradient algorithm Weight The gradient of , and iteratively calculate the error Minimize The optimal weight of is taken as the optimal weight parameter; error Weight The gradient of is: ； At the mth iteration during training, its weight is updated as: ； Weight Update Variable for: ； ； is the learning rate.

5. An intelligent screening system for achievement data based on multi-channel neural network, characterized in that: The achievement data intelligent screening system is used to execute the steps in the achievement data intelligent screening method based on a multi-channel neural network as described in any one of claims 1 to 4 above.

Images