CN117235633A

CN117235633A - Mechanism classification method, mechanism classification device, computer equipment and storage medium

Info

Publication number: CN117235633A
Application number: CN202311177549.3A
Authority: CN
Inventors: 李想
Original assignee: Ping An Property and Casualty Insurance Company of China Ltd
Current assignee: Ping An Property and Casualty Insurance Company of China Ltd
Priority date: 2023-09-13
Filing date: 2023-09-13
Publication date: 2023-12-15

Abstract

The embodiment of the application belongs to the fields of artificial intelligence and financial science and technology, and relates to a mechanism classification method, a device, computer equipment and a storage medium, wherein the method comprises the following steps: acquiring a full-scale image set including sample images of a plurality of mechanisms and classification evaluation values thereof; dividing the full portrait set to obtain a plurality of training sets and a plurality of verification sets; determining the weight of each training set according to a weight algorithm, wherein the weight determines the influence degree of the training set in training; training each initial classification evaluation model according to each training set to obtain a classification evaluation model; verifying the classification evaluation model according to each verification set to obtain verification deviation; selecting the classification evaluation model with the minimum verification deviation as a target classification evaluation modelThe method comprises the steps of carrying out a first treatment on the surface of the And inputting the organization portrait of the target organization into the target classification evaluation model to obtain an organization classification evaluation value so as to generate an organization classification result. The application also relates to a block chain technology, and the full quantity portrait set can be stored in the block chain _。 The application improves the accuracy of mechanism classification.

Description

Mechanism classification method, mechanism classification device, computer equipment and storage medium

Technical Field

The present application relates to the field of artificial intelligence and financial science and technology, and in particular, to a method and apparatus for classifying institutions, a computer device, and a storage medium.

Background

In production and life, it is often necessary to evaluate and classify a target object, for example, to evaluate and classify a characteristic of a user in a certain aspect based on a user characteristic, or to evaluate and classify a characteristic of an organization in a certain aspect based on an organization characteristic. For example, in the field of financial insurance, insurance companies are evaluated for future insurance fees according to their characteristics, and the operation conditions of the insurance companies are classified. However, the current evaluation and classification of institutions are usually predicted by special persons according to experience, the prediction effect varies from person to person, and the accuracy is low.

Disclosure of Invention

The embodiment of the application aims to provide a mechanism classification method, a mechanism classification device, computer equipment and a storage medium, so as to solve the problem of low mechanism classification and evaluation accuracy.

In order to solve the above technical problems, the embodiment of the present application provides a mechanism classification method, which adopts the following technical scheme:

acquiring a full image set, wherein the full image set comprises sample images of a plurality of mechanisms and corresponding classification evaluation values;

Dividing the full portrait set to obtain a plurality of training sets and a plurality of verification sets;

determining the weight of each training set according to a preset weight algorithm;

training each initial classification evaluation model according to each training set to obtain a classification evaluation model, and determining the influence degree of the training set in training by the weight of the training set;

verifying the classification evaluation model according to each verification set to obtain verification deviation;

selecting a classification evaluation model with the minimum verification deviation as a target classification evaluation model;

inputting the organization portrait of the target organization into the target classification evaluation model to obtain an organization classification evaluation value, and generating an organization classification result according to the organization classification evaluation value.

In order to solve the technical problems, the embodiment of the application also provides a mechanism classification device, which adopts the following technical scheme:

the image set acquisition module is used for acquiring a full image set, wherein the full image set comprises sample images of a plurality of mechanisms and corresponding classification evaluation values;

the portrait set dividing module is used for dividing the full portrait set to obtain a plurality of training sets and a plurality of verification sets;

the weight determining module is used for determining the weight of each training set according to a preset weight algorithm;

The model training module is used for training each initial classification evaluation model according to each training set to obtain a classification evaluation model, and the weight of the training set determines the influence degree of the training set in training;

the model verification module is used for verifying the classification evaluation model according to each verification set to obtain verification deviation;

the target determining module is used for selecting the classification evaluation model with the minimum verification deviation as a target classification evaluation model;

and the mechanism classification module is used for inputting the mechanism portrait of the target mechanism into the target classification evaluation model to obtain a mechanism classification evaluation value, and generating a mechanism classification result according to the mechanism classification evaluation value.

In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical schemes:

In order to solve the above technical problems, an embodiment of the present application further provides a computer readable storage medium, which adopts the following technical schemes:

Compared with the prior art, the embodiment of the application has the following main beneficial effects: acquiring a full image set, wherein the full image set comprises sample images of a plurality of mechanisms and corresponding classification evaluation values; dividing the full portrait set to obtain a plurality of training sets and a plurality of verification sets; measuring the characteristics of each training set in a certain aspect according to a preset weight algorithm, so as to add weight to each training set; training each initial classification evaluation model according to each training set to obtain a classification evaluation model, wherein the larger the weight of the training set is, the more important the representative training set is, the greater the influence degree on the model in training is, the more the model needs to learn the characteristics from the data sets in training, so that the model is effectively trained according to the training set, and the accuracy of the obtained classification evaluation model is ensured; verifying the classification evaluation model according to each verification set to obtain verification deviation; the smaller the verification deviation is, the smaller the error of the classification evaluation model on the verification set is, the better the prediction effect of the representative model is, the classification evaluation model with the minimum verification deviation is selected as the target classification evaluation model, and therefore the current most accurate model is selected; the mechanism portrait of the target mechanism is input into the target classification evaluation model, so that a mechanism classification evaluation value can be accurately obtained, a mechanism classification result is generated according to the mechanism classification evaluation value, and the accuracy of mechanism classification is improved; according to the application, the model obtained through training is used for automatically carrying out the mechanism evaluation and classification, so that the manual interference is reduced, and the accuracy of the mechanism evaluation and classification is improved.

Drawings

In order to more clearly illustrate the solution of the present application, a brief description will be given below of the drawings required for the description of the embodiments of the present application, it being apparent that the drawings in the following description are some embodiments of the present application, and that other drawings may be obtained from these drawings without the exercise of inventive effort for a person of ordinary skill in the art.

FIG. 1 is an exemplary system architecture diagram in which the present application may be applied;

FIG. 2 is a flow chart of one embodiment of a facility classification method according to the present application;

FIG. 3 is a schematic view of the structure of one embodiment of a mechanical sorting device according to the present application;

FIG. 4 is a schematic structural diagram of one embodiment of a computer device in accordance with the present application.

Detailed Description

Unless defined otherwise, 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 in the description of the applications herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having" and any variations thereof in the description of the application and the claims and the description of the drawings above are intended to cover a non-exclusive inclusion. The terms first, second and the like in the description and in the claims or in the above-described figures, are used for distinguishing between different objects and not necessarily for describing a sequential or chronological order.

Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the application. The appearances of such phrases in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. Those of skill in the art will explicitly and implicitly appreciate that the embodiments described herein may be combined with other embodiments.

In order to make the person skilled in the art better understand the solution of the present application, the technical solution of the embodiment of the present application will be clearly and completely described below with reference to the accompanying drawings.

As shown in fig. 1, a system architecture 100 may include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 is used as a medium to provide communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, among others.

The user may interact with the server 105 via the network 104 using the terminal devices 101, 102, 103 to receive or send messages or the like. Various communication client applications, such as a web browser application, a shopping class application, a search class application, an instant messaging tool, a mailbox client, social platform software, etc., may be installed on the terminal devices 101, 102, 103.

The terminal devices 101, 102, 103 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smartphones, tablet computers, electronic book readers, MP3 players (Moving Picture Experts Group Audio Layer III, dynamic video expert compression standard audio plane 3), MP4 (Moving Picture Experts Group Audio Layer IV, dynamic video expert compression standard audio plane 4) players, laptop and desktop computers, and the like.

The server 105 may be a server providing various services, such as a background server providing support for pages displayed on the terminal devices 101, 102, 103.

It should be noted that, the method for classifying institutions provided in the embodiment of the present application is generally executed by a server, and accordingly, the device for classifying institutions is generally disposed in the server.

It should be understood that the number of terminal devices, networks and servers in fig. 1 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for implementation.

With continued reference to FIG. 2, a flow chart of one embodiment of an organization sorting method according to the present application is shown. The mechanism classification method comprises the following steps:

Step S201, a full image set is obtained, wherein the full image set comprises sample images of a plurality of mechanisms and corresponding classification evaluation values.

In this embodiment, the electronic device (e.g., the server shown in fig. 1) on which the mechanism classification method operates may communicate with the terminal device through a wired connection or a wireless connection. It should be noted that the wireless connection may include, but is not limited to, 3G/4G/5G connection, wiFi connection, bluetooth connection, wiMAX connection, zigbee connection, UWB (ultra wideband) connection, and other now known or later developed wireless connection.

Specifically, a full-volume image set is obtained, wherein the full-volume image set comprises sample images of a plurality of mechanisms and classification evaluation values corresponding to the sample images; each sample representation contains a plurality of mechanism features of the mechanism. The present application classifies the mechanism according to the image of the mechanism. The classification evaluation value of the sample portrait represents the evaluation value of the organization on the preset attribute, and the organization can be classified according to the evaluation value. The classification evaluation value corresponds to a label.

For example, in the field of financial insurance, an organization may be an insurance company, and the organization features in the sample image may include, in addition to basic features of the organization (such as a business type, the number of employees, etc.), features related to the business, such as a premium income share, a premium income increase share, whether to offer insurance business for a new energy automobile, etc. The application can predict the premium which needs to be set in the future by the insurance company, takes the premium as a classification evaluation value, classifies the insurance company according to the premium obtained by prediction, and the classification result represents the operation condition of the insurance company in the future.

And step S202, dividing the full portrait set to obtain a plurality of training sets and a plurality of verification sets.

Specifically, the full representation set is divided to obtain a plurality of training sets and a plurality of verification sets, wherein the training sets and the verification sets are all subsets of the full representation set, and different training sets and verification sets can contain sample representations of the same organization.

Step S203, determining the weight of each training set according to a preset weight algorithm.

Specifically, the application adds weights to each training set, and the weights can be obtained through a preset weight algorithm. The application sets a plurality of weight algorithms, and the different weight algorithms measure the characteristics of different aspects of the data set and give the weight of the training set.

Step S204, training each initial classification evaluation model according to each training set to obtain a classification evaluation model, wherein the weight of the training set determines the influence degree of the training set in training.

Specifically, the application sets up a plurality of training sets and a plurality of different initial classification evaluation models, such as linear regression, decision trees, random forests, support vector machines, neural networks, etc.

For each initial classification evaluation model, training the initial classification evaluation model according to each training set is needed, and the classification evaluation model is obtained after training is finished. The greater the weight of the training set, the more important the representation of the training set, and the greater the degree of impact on the model during training, the more the model needs to learn the features from these data sets during training.

Step S205, verifying the classification evaluation model according to each verification set to obtain verification deviation.

Specifically, after obtaining a plurality of classification evaluation models, each classification evaluation model needs to be verified: and respectively inputting each verification set into a classification evaluation model to obtain the difference between the output value of the classification evaluation model on the sample image in the verification set and the classification evaluation value thereof, and obtaining verification deviation according to the difference of the classification evaluation model on all the verification sets.

Step S206, selecting the classification evaluation model with the minimum verification deviation as the target classification evaluation model.

Specifically, the verification deviation may be a numerical value, and the smaller the numerical value, the smaller the error of the classification evaluation model on the verification set, and the better the prediction effect of the representative model. And selecting the classification evaluation model with the minimum verification deviation from the multiple classification evaluation models as a target classification evaluation model, and putting the target classification evaluation model into application.

Step S207, inputting the organization image of the target organization into the target classification evaluation model to obtain an organization classification evaluation value, and generating an organization classification result according to the organization classification evaluation value.

Specifically, a mechanism portrait of a target mechanism to be classified is acquired and input into a target classification evaluation model to obtain a mechanism classification evaluation value.

Multiple evaluation value intervals can be preset, and different evaluation value intervals represent different institution types, so that the evaluation value intervals where the institution classification evaluation values are located can be used for generating an institution classification result.

In this embodiment, a full-scale image set is obtained, the full-scale image set including sample images of a plurality of mechanisms and classification evaluation values corresponding to the sample images; dividing the full portrait set to obtain a plurality of training sets and a plurality of verification sets; measuring the characteristics of each training set in a certain aspect according to a preset weight algorithm, so as to add weight to each training set; training each initial classification evaluation model according to each training set to obtain a classification evaluation model, wherein the larger the weight of the training set is, the more important the representative training set is, the greater the influence degree on the model in training is, the more the model needs to learn the characteristics from the data sets in training, so that the model is effectively trained according to the training set, and the accuracy of the obtained classification evaluation model is ensured; verifying the classification evaluation model according to each verification set to obtain verification deviation; the smaller the verification deviation is, the smaller the error of the classification evaluation model on the verification set is, the better the prediction effect of the representative model is, the classification evaluation model with the minimum verification deviation is selected as the target classification evaluation model, and therefore the current most accurate model is selected; the mechanism portrait of the target mechanism is input into the target classification evaluation model, so that a mechanism classification evaluation value can be accurately obtained, a mechanism classification result is generated according to the mechanism classification evaluation value, and the accuracy of mechanism classification is improved; according to the application, the model obtained through training is used for automatically carrying out the mechanism evaluation and classification, so that the manual interference is reduced, and the accuracy of the mechanism evaluation and classification is improved.

Further, before the step S201, the method may further include: acquiring an initial portrait set; acquiring mechanism features contained in all sample images in the initial image set; inputting the obtained mechanism features into a trained random forest to obtain importance scores of the mechanism features; and performing feature screening on all sample images according to the obtained importance scores to obtain a full image set.

Specifically, an initial portrait set, which is a full portrait set in an initial state, is acquired. Acquiring mechanism features contained in all sample images in the initial image set; and inputting the obtained mechanism features into a random forest to obtain importance scores of the mechanism features.

Random forests need to be pre-trained in which an initial portrait set is assumed, which contains the following mechanical features: the city of the company, the number of branch companies, the number of staff, the premium income comparison, the premium income speed-increasing comparison, whether to set up insurance business aiming at new energy automobiles, and the like.

First, a data set is prepared and divided into a feature matrix (X) containing individual mechanism features of all the sample images and a target variable (y) representing a classification evaluation value of the mechanism to which the sample image corresponds.

Next, training and feature selection are performed by a random forest algorithm. The random forest algorithm automatically evaluates the importance of each institutional feature and assigns a corresponding score thereto. For example, after training a random forest model, the importance score of each organization feature is obtained by the feature_importances_attribute of the model, which returns an array in which each element corresponds to an organization feature and the value represents the importance score of that organization feature.

The magnitude of the importance score indicates the contribution of the mechanism feature to the mechanism evaluation and classification, so that the mechanism feature contained in all the sample images in the initial image set can be subjected to feature screening according to the obtained importance score, for example, the mechanism feature with the importance score being greater than or equal to a preset threshold value is reserved, and the feature with the importance less than the preset threshold value is deleted from the initial image set, so that the full image set is obtained.

In the embodiment, mechanism features contained in all sample images in an initial image set are obtained, importance scores of the mechanism features are output through random forests, and the magnitude of the importance scores represents the contribution of the mechanism features to the evaluation and classification of a prediction mechanism; according to the importance score, the mechanism features contained in the initial portrait set can be subjected to feature screening, the mechanism features with high contribution to mechanism evaluation and classification are reserved, the data set can be simplified, the quality of the data set is improved, and the accuracy of subsequent mechanism evaluation and classification is ensured.

Further, the step S202 may include: randomly dividing the full portrait set to obtain a plurality of initial training sets and a plurality of initial verification sets; and randomly carrying out feature elimination on the sample portraits in the plurality of initial training sets and the plurality of initial verification sets to obtain a plurality of training sets and a plurality of verification sets.

Specifically, the full portrait set is randomly divided to obtain a plurality of initial training sets and a plurality of initial verification sets. The initial training set and the initial verification set are recorded as initial sets, so that the number of sample images in the initial sets is not limited, and the number of the sample images contained in different initial sets can be different; and the feature elimination can be randomly carried out on the sample portrait in each initial set to obtain a plurality of training sets and a plurality of verification sets. When the feature elimination is performed, at least one type of mechanism feature is completely deleted from the initial set, and the initial set does not have the feature value of the mechanism feature any more; or, deleting at least one type of mechanism feature from a portion of the sample representation in the initial set; alternatively, some of the mechanical features in all/part of the sample images in the initial set are deleted randomly.

It will be appreciated that different random feature culling may be performed for different initial sets.

In the embodiment, the full portrait set is randomly divided to obtain a plurality of initial training sets and a plurality of initial verification sets; and randomly carrying out feature elimination on sample images in a plurality of initial training sets and a plurality of initial verification sets to obtain a plurality of training sets and a plurality of verification sets, so that the training sets and the sample sets with rich feature combinations can be obtained, various real conditions can be simulated, and the accuracy of the model can be improved.

Further, the step S203 may include: clustering each training set by a density-based clustering algorithm to obtain a clustering result of each training set; and determining the weight of each training set according to the number of clusters in each training aggregation class result.

Specifically, clustering is carried out on each training set through a clustering algorithm based on density, so that clustering results of each training set are obtained. The Density-based clustering algorithm may predefine the number of clusters, with one of the most common algorithms being DBSCAN (Density-Based Spatial Clustering of Applications with Noise). DBSCAN is a non-parameterized algorithm that automatically determines the number of clusters based on the density distribution of the data and enables any shape of clusters to be found.

The clustering result shows the number of clusters in the training set, and different clusters can be regarded as sample portraits of different categories. In general, a wide variety of institutional images may be encountered in real-world applications, i.e., across various types of institutional images. Therefore, it can be considered that the more the number of clusters in the training set is, the more types of mechanism portraits can be covered by the representative training set, and the more the training set is close to the real situation; these training sets are also more important.

Therefore, the weight of the training set can be determined according to the number of clusters in the clustering result, and the more the number of clusters is, the larger the weight of the training set is.

In the embodiment, clustering is performed on each training set through a clustering algorithm based on density, so as to obtain a clustering result of each training set; the more clusters in the clustering result, the more types of mechanism portraits are covered by the representative training set, and the more the training set is close to the real situation, the more important; and determining the weight of the training set according to the number of clusters in the clustering result, so that the accuracy of weight addition is ensured.

Further, the step S203 may further include: respectively calculating the eigenvalue missing degree of each training set; and determining the weight of each training set according to the characteristic value deficiency degree of each training set.

Specifically, for each sample image in the training set, the sample image may have a missing feature value, for example, one sample image lacks a feature value with a premium income ratio, and the other sample image lacks two feature values with a premium income ratio and a branch company number. And taking the training set as a whole, calculating the total number of missing characteristic values in the training set, and dividing the total number by the number of due characteristic values in the training set to obtain the characteristic value missing degree of the training set. The number of due characteristic values is calculated according to the mechanism characteristics contained after the characteristic screening is carried out according to the random forest, for example, after the characteristic screening is carried out on the random forest, 20 characteristics are contained in each sample image in the full-quantity image set, and the training set has 100 sample images in total, so that the total number of due characteristic values is 20 multiplied by 100=2000; if there are 100 missing eigenvalues in the training set, the eigenvalue missing degree is 100/2000=0.05.

It can be understood that the higher the degree of missing feature values, the more feature values that are missing in the training set, the worse the quality of the training set, and the more errors and misleading the training set may bring, and the influence brought by these training sets needs to be controlled. The weight of the training set can be determined according to the degree of missing feature values of the training set, and in general, the higher the degree of missing feature values, the lower the weight.

In this embodiment, the feature value missing degree of each training set is calculated respectively; the higher the degree of missing the characteristic value, the more the characteristic value which is missing in the training set, the worse the quality of the training set, and the influence caused by the training set needs to be reduced; and determining the weight of the training set according to the degree of missing of the characteristic values of each training set, so that the accuracy of the weight is ensured.

Further, the step S204 may include: inputting each training set into each initial classification evaluation model to obtain a prediction result of the initial classification evaluation model on each training set, wherein the prediction result comprises a prediction evaluation value of the initial classification evaluation model on each sample portrait in each training set; calculating model loss of the initial classification evaluation model according to the weight of each training set, the classification evaluation value of each sample image in each training set and the prediction result of the initial classification evaluation model on each training set; and adjusting the initial classification evaluation model according to the model loss until the model loss meets the training stop condition, and obtaining the classification evaluation model.

Specifically, for each initial classification evaluation model, each training set is input into the initial classification evaluation model to obtain a prediction result of the model on each training set, wherein the prediction result comprises a prediction evaluation value of the model on each sample portrait in each training set.

For each sample portrait in each training set, the sample portrait has a classification evaluation value as a label and a prediction evaluation value output by the model, and the error of the model on the sample portrait can be calculated according to the classification evaluation value and the prediction evaluation value; and accumulating errors on the images of the samples to obtain an error E1 of the model on the training set under the condition of no weight. As discussed above, the weight determines the influence degree of the training set in training, and the error E2 of the model on the model training set under the condition of weight can be obtained by multiplying the weight of the training set with the error E1 of the model on the training set. Therefore, the model loss of the initial classification evaluation model can be obtained by carrying out weighted summation on the error E1 of the model on each training set and the weight of each training set.

And aiming at reducing the model loss, adjusting the model parameters of the initial classification evaluation model, and carrying out iterative training on the initial classification evaluation model after parameter adjustment until the model loss meets a training stop condition to obtain the classification evaluation model, wherein the training stop condition can be that the model loss is less than or equal to a preset loss threshold value.

In the embodiment, each training set is respectively input into each initial classification evaluation model to obtain a prediction result of the initial classification evaluation model on each training set, wherein the prediction result comprises a prediction evaluation value of the initial classification evaluation model on each sample portrait in each training set; according to the weight of each training set, the classification evaluation value and the prediction evaluation value of each sample image in each training set, the model loss of the initial classification evaluation model can be calculated; and adjusting the initial classification evaluation model according to the model loss until the model loss meets the training stop condition, so that an accurate classification evaluation model can be obtained.

Further, the step S205 may include: inputting each verification set into each classification evaluation model to obtain a verification result of the classification evaluation model on each verification set, wherein the verification result comprises verification evaluation values of the classification evaluation model on each sample image in each verification set; and calculating the verification deviation of the classification evaluation model on each verification set according to the classification evaluation value of each sample image in each verification set and the verification result of the classification evaluation model on each verification set.

Specifically, for each classification evaluation model, each verification set is input into the classification evaluation model to obtain a verification result of the classification evaluation model on each verification set, wherein the verification result comprises verification evaluation values of the classification evaluation model on each sample image in each verification set.

For each sample portrait in each verification set, the sample portrait has a classification evaluation value as a label, and a verification evaluation value output by a classification evaluation model, and a deviation of the classification evaluation model on the sample portrait can be calculated according to the classification evaluation value and the verification evaluation value; and accumulating errors on the images of the samples to obtain the deviation of the classification evaluation model on the verification set.

And accumulating the deviation of the classification evaluation model on each verification set to obtain the verification deviation of the classification evaluation model on the verification set.

In one embodiment, weights can be added to each verification set through a preset weight algorithm, and then the deviation of the classification evaluation model on each verification set and the weights of each verification set are weighted and summed to obtain the verification deviation of the classification evaluation model on the verification set.

In the embodiment, each verification set is input into each classification evaluation model to obtain a verification result of the classification evaluation model on each verification set, wherein the verification result comprises verification evaluation values of the classification evaluation model on each sample image in each verification set; according to the classification evaluation value and the verification evaluation value of each sample image in each verification set, verification deviation of the classification evaluation model on each verification set can be calculated, and then the trained model is evaluated.

It is emphasized that the full representation set may also be stored in a blockchain node in order to further ensure privacy and security of the full representation set.

The blockchain is a novel application mode of computer technologies such as distributed data storage, point-to-point transmission, consensus mechanism, encryption algorithm and the like. The Blockchain (Blockchain), which is essentially a decentralised database, is a string of data blocks that are generated by cryptographic means in association, each data block containing a batch of information of network transactions for verifying the validity of the information (anti-counterfeiting) and generating the next block. The blockchain may include a blockchain underlying platform, a platform product services layer, an application services layer, and the like.

The embodiment of the application can acquire and process the related data based on the artificial intelligence technology. Among these, artificial intelligence (Artificial Intelligence, AI) is the theory, method, technique and application system that uses a digital computer or a digital computer-controlled machine to simulate, extend and extend human intelligence, sense the environment, acquire knowledge and use knowledge to obtain optimal results.

Artificial intelligence infrastructure technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a robot technology, a biological recognition technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and other directions.

Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by computer readable instructions stored in a computer readable storage medium that, when executed, may comprise the steps of the embodiments of the methods described above. The storage medium may be a nonvolatile storage medium such as a magnetic disk, an optical disk, a Read-Only Memory (ROM), or a random access Memory (Random Access Memory, RAM).

It should be understood that, although the steps in the flowcharts of the figures are shown in order as indicated by the arrows, these steps are not necessarily performed in order as indicated by the arrows. The steps are not strictly limited in order and may be performed in other orders, unless explicitly stated herein. Moreover, at least some of the steps in the flowcharts of the figures may include a plurality of sub-steps or stages that are not necessarily performed at the same time, but may be performed at different times, the order of their execution not necessarily being sequential, but may be performed in turn or alternately with other steps or at least a portion of the other steps or stages.

With further reference to fig. 3, as an implementation of the method shown in fig. 2 described above, the present application provides an embodiment of a mechanism sorting device, which corresponds to the method embodiment shown in fig. 2, and which is particularly applicable to various electronic devices.

As shown in fig. 3, the mechanism classifying device 300 according to the present embodiment includes: a representation set acquisition module 301, a representation set partitioning module 302, a weight determination module 303, a model training module 304, a model verification module 305, a target determination module 306, and an institution classification module 307, wherein:

The image set acquisition module 301 is used for acquiring a full image set, wherein the full image set comprises sample images of a plurality of mechanisms and corresponding classification evaluation values.

The portrait set dividing module 302 is configured to divide the full portrait set to obtain a plurality of training sets and a plurality of verification sets.

The weight determining module 303 is configured to determine the weight of each training set according to a preset weight algorithm.

The model training module 304 is configured to train each initial classification evaluation model according to each training set, to obtain a classification evaluation model, and the weight of the training set determines the influence degree of the training set in training.

The model verification module 305 is configured to verify the classification evaluation model according to each verification set, so as to obtain a verification deviation.

The target determining module 306 is configured to select the classification evaluation model with the smallest verification deviation as the target classification evaluation model.

The mechanism classification module 307 is used for inputting the mechanism portrait of the target mechanism into the target classification evaluation model to obtain a mechanism classification evaluation value, and generating a mechanism classification result according to the mechanism classification evaluation value.

In some alternative implementations of the present embodiment, the mechanism classification device 300 may further include: initial acquisition module, characteristic input module and characteristic screening module, wherein:

and the initial acquisition module is used for acquiring the initial portrait set.

And the feature acquisition module is used for acquiring the mechanism features contained in all sample images in the initial image set.

And the feature input module is used for inputting the obtained mechanism features into the trained random forest to obtain importance scores of the mechanism features.

And the feature screening module is used for carrying out feature screening on all sample images according to the obtained importance scores to obtain a full image set.

In some alternative implementations of the present embodiment, the representation set partitioning module 302 may include: randomly dividing a sub-module and a feature eliminating sub-module, wherein:

and the random dividing sub-module is used for randomly dividing the full portrait set to obtain a plurality of initial training sets and a plurality of initial verification sets.

And the feature elimination sub-module is used for randomly carrying out feature elimination on sample images in the plurality of initial training sets and the plurality of initial verification sets to obtain a plurality of training sets and a plurality of verification sets.

In some alternative implementations of the present embodiment, the weight determination module 303 may include: training the aggregation class sub-module and the first determination sub-module, wherein:

the training aggregation sub-module is used for clustering each training set through a density-based clustering algorithm to obtain a clustering result of each training set.

The first determining submodule is used for determining the weight of each training set according to the number of clusters in each training aggregation class result.

In other alternative implementations of the present embodiment, the weight determination module 303 may include: a deficiency degree calculation sub-module and a second determination sub-module, wherein:

and the missing degree calculation submodule is used for calculating the missing degree of the characteristic value of each training set respectively.

And the second determining submodule is used for determining the weight of each training set according to the characteristic value deficiency degree of each training set.

In some alternative implementations of the present embodiment, model training module 304 may include: training set input submodule, loss calculation submodule and model adjustment submodule, wherein:

the training set input sub-module is used for respectively inputting each training set into each initial classification evaluation model to obtain the prediction result of the initial classification evaluation model on each training set, wherein the prediction result comprises the prediction evaluation value of the initial classification evaluation model on each sample portrait in each training set.

The loss calculation sub-module is used for calculating the model loss of the initial classification evaluation model according to the weight of each training set, the classification evaluation value of each sample image in each training set and the prediction result of the initial classification evaluation model on each training set.

And the model adjustment sub-module is used for adjusting the initial classification evaluation model according to the model loss until the model loss meets the training stop condition to obtain the classification evaluation model.

In some alternative implementations of the present embodiment, the model verification module 305 may include: verification set input submodule and deviation calculation submodule, wherein:

the verification set input sub-module is used for respectively inputting each verification set into each classification evaluation model to obtain verification results of the classification evaluation models on each verification set, wherein the verification results comprise verification evaluation values of the classification evaluation models on each sample image in each verification set.

And the deviation calculation sub-module is used for calculating the verification deviation of the classification evaluation model on each verification set according to the classification evaluation value of each sample image in each verification set and the verification result of the classification evaluation model on each verification set.

In order to solve the technical problems, the embodiment of the application also provides computer equipment. Referring specifically to fig. 4, fig. 4 is a basic structural block diagram of a computer device according to the present embodiment.

The computer device 4 comprises a memory 41, a processor 42, a network interface 43 communicatively connected to each other via a system bus. It should be noted that only computer device 4 having components 41-43 is shown in the figures, but it should be understood that not all of the illustrated components are required to be implemented and that more or fewer components may be implemented instead. It will be appreciated by those skilled in the art that the computer device herein is a device capable of automatically performing numerical calculations and/or information processing in accordance with predetermined or stored instructions, the hardware of which includes, but is not limited to, microprocessors, application specific integrated circuits (Application Specific Integrated Circuit, ASICs), programmable gate arrays (fields-Programmable Gate Array, FPGAs), digital processors (Digital Signal Processor, DSPs), embedded devices, etc.

The computer equipment can be a desktop computer, a notebook computer, a palm computer, a cloud server and other computing equipment. The computer equipment can perform man-machine interaction with a user through a keyboard, a mouse, a remote controller, a touch pad or voice control equipment and the like.

The memory 41 includes at least one type of readable storage medium including flash memory, hard disk, multimedia card, card memory (e.g., 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, magnetic disk, optical disk, etc. In some embodiments, the storage 41 may be an internal storage unit of the computer device 4, such as a hard disk or a memory of the computer device 4. In other embodiments, the memory 41 may also be an external storage device of the computer device 4, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash Card (Flash Card) or the like, which are provided on the computer device 4. Of course, the memory 41 may also comprise both an internal memory unit of the computer device 4 and an external memory device. In this embodiment, the memory 41 is typically used to store an operating system and various types of application software installed on the computer device 4, such as computer readable instructions of an organization classification method, and the like. Further, the memory 41 may be used to temporarily store various types of data that have been output or are to be output.

The processor 42 may be a central processing unit (Central Processing Unit, CPU), controller, microcontroller, microprocessor, or other data processing chip in some embodiments. The processor 42 is typically used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is configured to execute computer readable instructions stored in the memory 41 or process data, such as computer readable instructions for executing the mechanism classification method.

The network interface 43 may comprise a wireless network interface or a wired network interface, which network interface 43 is typically used for establishing a communication connection between the computer device 4 and other electronic devices.

The computer device provided in the present embodiment may perform the above-described mechanism classification method. The mechanism classification method here may be the mechanism classification method of each of the above embodiments.

The present application also provides another embodiment, namely, a computer-readable storage medium storing computer-readable instructions executable by at least one processor to cause the at least one processor to perform the steps of the mechanism classification method as described above.

From the above description of the embodiments, it will be clear to those skilled in the art that the above-described embodiment method may be implemented by means of software plus a necessary general hardware platform, but of course may also be implemented by means of hardware, but in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art in the form of a software product stored in a storage medium (e.g. ROM/RAM, magnetic disk, optical disk) comprising instructions for causing a terminal device (which may be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to perform the method according to the embodiments of the present application.

It is apparent that the above-described embodiments are only some embodiments of the present application, but not all embodiments, and the preferred embodiments of the present application are shown in the drawings, which do not limit the scope of the patent claims. This application may be embodied in many different forms, but rather, embodiments are provided in order to provide a thorough and complete understanding of the present disclosure. Although the application has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that modifications may be made to the embodiments described in the foregoing description, or equivalents may be substituted for elements thereof. All equivalent structures made by the content of the specification and the drawings of the application are directly or indirectly applied to other related technical fields, and are also within the scope of the application.

Claims

1. A method of institutional classification comprising the steps of:

2. The institutional classification method of claim 1, further comprising, prior to the step of acquiring the full representation set:

acquiring an initial portrait set;

obtaining mechanism features contained in all sample images in the initial image set;

Inputting the obtained mechanism features into a trained random forest to obtain importance scores of the mechanism features;

and performing feature screening on all the sample images according to the obtained importance score to obtain a full image set.

3. The method of claim 1, wherein the step of dividing the full representation set to obtain a plurality of training sets and a plurality of validation sets comprises:

randomly dividing the full portrait set to obtain a plurality of initial training sets and a plurality of initial verification sets;

and randomly carrying out feature elimination on the sample portraits in the initial training sets and the initial verification sets to obtain a plurality of training sets and a plurality of verification sets.

4. The method of claim 1, wherein the step of determining weights for each training set according to a preset weight algorithm comprises:

clustering each training set by a density-based clustering algorithm to obtain a clustering result of each training set;

and determining the weight of each training set according to the number of clusters in each training aggregation class result.

5. The method of claim 1, wherein the step of determining weights for each training set according to a preset weight algorithm further comprises:

Respectively calculating the eigenvalue missing degree of each training set;

and determining the weight of each training set according to the characteristic value missing degree of each training set.

6. The method of claim 1, wherein the step of training each initial classification evaluation model based on the training sets to obtain classification evaluation models comprises:

inputting each training set into each initial classification evaluation model to obtain a prediction result of the initial classification evaluation model on each training set, wherein the prediction result comprises a prediction evaluation value of each sample portrait in each training set by the initial classification evaluation model;

calculating model loss of the initial classification evaluation model according to the weight of each training set, the classification evaluation value of each sample image in each training set and the prediction result of the initial classification evaluation model on each training set;

and adjusting the initial classification evaluation model according to the model loss until the model loss meets the training stopping condition, so as to obtain the classification evaluation model.

7. The method of claim 1, wherein the step of validating the classification evaluation model according to each validation set to obtain a validation bias comprises:

Inputting each verification set into each classification evaluation model to obtain a verification result of the classification evaluation model on each verification set, wherein the verification result comprises verification evaluation values of each sample image in each verification set by the classification evaluation model;

and calculating verification deviation of the classification evaluation model on each verification set according to the classification evaluation value of each sample image in each verification set and the verification result of the classification evaluation model on each verification set.

8. A mechanism sorting device, comprising:

9. A computer device comprising a memory having stored therein computer readable instructions which when executed by a processor implement the steps of the method of mechanism classification of any of claims 1 to 7.

10. A computer readable storage medium having stored thereon computer readable instructions which when executed by a processor implement the steps of the mechanism classification method of any of claims 1 to 7.