CN115994586A

CN115994586A - Method, device, electronic equipment and medium for recommending initialization parameters of algorithm

Info

Publication number: CN115994586A
Application number: CN202211597150.6A
Authority: CN
Inventors: 陈晗
Original assignee: China Construction Bank Corp; CCB Finetech Co Ltd
Current assignee: China Construction Bank Corp; CCB Finetech Co Ltd
Priority date: 2022-12-12
Filing date: 2022-12-12
Publication date: 2023-04-21

Abstract

The disclosure provides an initialization parameter recommendation method, device, equipment, storage medium and program product of an algorithm, relates to the technical field of computers, and can be applied to the technical field of finance. The method comprises the following steps: in response to receiving a recommendation request from a user, obtaining input information corresponding to the recommendation request, the input information including a target algorithm and a first input model feature; according to first historical data corresponding to a target algorithm, the first historical data comprise numerical vectors of a plurality of historical second input model features, numerical vectors of target historical second input model features matched with the numerical vectors of the first input model features are determined, and target first initialization parameter values corresponding to the target historical second input model features are obtained; according to second historical data corresponding to the target algorithm, the second historical data comprise a plurality of historical second initialization parameter values and model training representation values corresponding to each historical second initialization parameter value respectively, determining a target model training representation value and obtaining a target second initialization parameter value corresponding to the target model training representation value; and determining an initialization parameter recommended value of the target algorithm according to the target first initialization parameter value and the target second initialization parameter value, and recommending the initialization parameter recommended value to the user.

Description

Method, device, electronic equipment and medium for recommending initialization parameters of algorithm

Technical Field

The present disclosure relates to the field of computer technology, and may be applied to the field of financial technology, and more particularly, to an algorithm initialization parameter recommendation method, apparatus, device, medium, and program product.

Background

Machine learning related work requires training of many machine learning algorithms, implementing code with already written algorithms, large amounts of data, and infrastructure. The machine learning platform can quickly position the most suitable machine learning algorithm, adjust algorithm parameters and deploy large-scale model training into an enterprise or cloud cluster, so that a complete system for managing and monitoring model training processes is provided.

One important step in machine learning is configuring and adjusting algorithm parameters. A machine learning model generates a plurality of model parameters (model parameters) based on data learning. In short, what machine learning needs to "learn" is the implicit model parameters in the data, and then predict and judge new data in the future using these implicit parameters. The algorithm parameters (hyperparameters) are set before model training, and are not "learned" from the data, and reflect the characteristics of model complexity, training time, etc. Different machine learning algorithms have different algorithm parameters.

Disclosure of Invention

In view of the foregoing, the present disclosure provides an initialization parameter recommendation method, apparatus, device, medium, and program product for an algorithm, which may avoid too much reliance on a single factor while recommending an initialization parameter, such as determining, by a first policy, a numerical vector of a target historical second input model feature matching a numerical vector of a first input model feature in response to receiving a recommendation request from a user, to obtain a target first initialization parameter value corresponding to the target historical second input model feature; and determining a target model training representation value according to second historical data corresponding to a target algorithm through a second strategy to obtain a target second initialization parameter value corresponding to the target model training representation value, and determining an initialization parameter recommended value of the target algorithm according to the target first initialization parameter value and the target second initialization parameter value, so that fusion of a plurality of parameters to obtain the target parameter recommended value is realized, and better effect of the model is achieved.

According to a first aspect of the present disclosure, there is provided an initialization parameter recommendation method of an algorithm, including: in response to receiving a recommendation request from a user, obtaining input information corresponding to the recommendation request, wherein the input information comprises a target algorithm and a first input model feature; according to first historical data corresponding to the target algorithm, the first historical data comprise numerical vectors of a plurality of historical second input model features, numerical vectors of target historical second input model features matched with the numerical vectors of the first input model features are determined, and target first initialization parameter values corresponding to the target historical second input model features are obtained; according to second historical data corresponding to the target algorithm, the second historical data comprise a plurality of historical second initialization parameter values and model training performance values corresponding to each historical second initialization parameter value respectively, and a target model training performance value is determined to obtain a target second initialization parameter value corresponding to the target model training performance value; and determining an initialization parameter recommended value of the target algorithm according to the target first initialization parameter value and the target second initialization parameter value, and recommending the initialization parameter recommended value to the user.

According to an embodiment of the present disclosure, the second history data further includes a model evaluation AUC index corresponding to each of the history second initial parameter values, a degree of improvement index corresponding to each of the history second initial parameter values, and a degree of differentiation KS index corresponding to each of the history second initial parameter values, the method including: and determining the model training representation values corresponding to each historical second initial parameter value according to the model evaluation AUC index, the lifting degree index and the differentiation degree KS index.

According to an embodiment of the disclosure, according to first history data corresponding to the target algorithm, the first history data includes numerical vectors of a plurality of historical second input model features, determining a numerical vector of a target historical second input model feature matching the numerical vector of the first input model feature, and obtaining a target first initialization parameter value corresponding to the target historical second input model feature, including: and determining the numerical vector of the target historical second input model feature matched with the numerical vector of the first input model feature by adopting a cosine similarity calculation method according to the first historical data corresponding to the target algorithm, and obtaining a target first initialization parameter value corresponding to the target historical second input model feature.

According to an embodiment of the disclosure, the determining an initialization parameter recommendation value of the target algorithm according to the target first initialization parameter value and the target second initialization parameter value, and recommending the initialization parameter recommendation value to the user includes: calculating an average value according to the target first initialization parameter value and the target second initialization parameter value; and taking the average value as an initialization parameter recommended value of the target algorithm, and recommending the initialization parameter recommended value to the user.

According to an embodiment of the present disclosure, the method further comprises: in response to obtaining a target third initialization parameter value from a target object, an initialization parameter recommendation value for the target algorithm is determined from the target first initialization parameter value, the target second initialization parameter value, and the target third initialization parameter value.

A second aspect of the present disclosure provides an initialization parameter recommendation apparatus of an algorithm, including: the acquisition module is used for responding to a recommendation request received from a user and acquiring input information corresponding to the recommendation request, wherein the input information comprises a target algorithm and a first input model feature; the first determining module is used for determining the numerical vector of the target historical second input model feature matched with the numerical vector of the first input model feature according to the first historical data corresponding to the target algorithm, wherein the first historical data comprises the numerical vectors of a plurality of historical second input model features, and obtaining a target first initialization parameter value corresponding to the target historical second input model feature; the second determining module is used for determining a target model training representation value according to second historical data corresponding to the target algorithm, wherein the second historical data comprises a plurality of historical second initialization parameter values and model training representation values corresponding to each historical second initialization parameter value respectively, and obtaining a target second initialization parameter value corresponding to the target model training representation value; and a third determining module, configured to determine an initialization parameter recommendation value of the target algorithm according to the target first initialization parameter value and the target second initialization parameter value, and recommend the initialization parameter recommendation value to the user.

A third aspect of the present disclosure provides an electronic device, comprising: one or more processors; and a memory for storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to perform the initialization parameter recommendation method of the algorithm described above.

A fourth aspect of the present disclosure also provides a computer-readable storage medium having stored thereon executable instructions that, when executed by a processor, cause the processor to perform the method of initializing parameter recommendation of the algorithm described above.

A fifth aspect of the present disclosure also provides a computer program product comprising a computer program which, when executed by a processor, implements the method of initialising a parameter recommendation for the algorithm described above.

According to the initialization parameter recommendation method of the algorithm, when the initialization parameters are recommended, the situation that the single factor is too dependent is avoided, for example, in response to receiving a recommendation request from a user, the numerical vector of the target historical second input model feature matched with the numerical vector of the first input model feature is determined through a first strategy, and the target first initialization parameter value corresponding to the target historical second input model feature is obtained; and determining a target model training representation value according to second historical data corresponding to a target algorithm through a second strategy to obtain a target second initialization parameter value corresponding to the target model training representation value, and determining an initialization parameter recommended value of the target algorithm according to the target first initialization parameter value and the target second initialization parameter value, so that fusion of a plurality of parameters to obtain the target parameter recommended value is realized, and better effect of the model is achieved.

Drawings

The foregoing and other objects, features and advantages of the disclosure will be more apparent from the following description of embodiments of the disclosure with reference to the accompanying drawings, in which:

FIG. 1 schematically illustrates an application scenario diagram of an initialization parameter recommendation method, apparatus, device, medium and program product of an algorithm according to an embodiment of the present disclosure;

FIG. 2 schematically illustrates a flow chart of an initialization parameter recommendation method of an algorithm according to an embodiment of the present disclosure;

FIG. 3 schematically illustrates an implementation of a user obtaining initialization parameter recommendation values according to an embodiment of the present disclosure;

FIG. 4 schematically illustrates a block diagram of an initialization parameter recommendation device of an algorithm according to an embodiment of the present disclosure; and

fig. 5 schematically illustrates a block diagram of an electronic device adapted to implement an initialization parameter recommendation method of an algorithm according to an embodiment of the present disclosure.

Detailed Description

Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. It should be understood that the description is only exemplary and is not intended to limit the scope of the present disclosure. In the following detailed description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the embodiments of the present disclosure. It may be evident, however, that one or more embodiments may be practiced without these specific details. In addition, in the following description, descriptions of well-known structures and techniques are omitted so as not to unnecessarily obscure the concepts of the present disclosure.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The terms "comprises," "comprising," and/or the like, as used herein, specify the presence of stated features, steps, operations, and/or components, but do not preclude the presence or addition of one or more other features, steps, operations, or components.

All terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art unless otherwise defined. It should be noted that the terms used herein should be construed to have meanings consistent with the context of the present specification and should not be construed in an idealized or overly formal manner.

Where expressions like at least one of "A, B and C, etc. are used, the expressions should generally be interpreted in accordance with the meaning as commonly understood by those skilled in the art (e.g.," a system having at least one of A, B and C "shall include, but not be limited to, a system having a alone, B alone, C alone, a and B together, a and C together, B and C together, and/or A, B, C together, etc.).

The embodiment of the disclosure provides an initialization parameter recommending method and device of an algorithm, which are used for responding to received initialization parameter recommending tasks of the algorithm and determining target equipment corresponding to the initialization parameter recommending tasks of the algorithm; acquiring an index description file corresponding to the target equipment according to the identifier of the target equipment; determining preset information corresponding to the index description file, wherein the preset information comprises target index information; and acquiring the resource data of the target equipment corresponding to the target index information.

Fig. 1 schematically illustrates an application scenario diagram of an initialization parameter recommendation method, apparatus, device, medium and program product of an algorithm according to an embodiment of the present disclosure.

As shown in fig. 1, an application scenario 100 according to this embodiment 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 shopping class applications, web browser applications, search class applications, instant messaging tools, mailbox clients, social platform software, etc. (by way of example only) may be installed on the

terminal devices

101, 102, 103.

The

terminal devices

101, 102, 103 may be a variety of electronic devices having a display screen and supporting web browsing, including but not limited to smartphones, tablets, laptop and desktop computers, and the like.

The server 105 may be a server providing various services, such as a background management server (by way of example only) providing support for websites browsed by users using the

terminal devices

101, 102, 103. The background management server may analyze and process the received data such as the user request, and feed back the processing result (e.g., the web page, information, or data obtained or generated according to the user request) to the terminal device.

It should be noted that, the method for recommending initialization parameters of the algorithm provided by the embodiments of the present disclosure may be generally performed by the server 105. Accordingly, the initialization parameter recommendation device of the algorithm provided by the embodiments of the present disclosure may be generally provided in the server 105. The initialization parameter recommendation method of the algorithm provided by the embodiments of the present disclosure may also be performed by a server or a server cluster that is different from the server 105 and is capable of communicating with the

terminal devices

101, 102, 103 and/or the server 105. Accordingly, the initialization parameter recommendation means of the algorithm provided by the embodiments of the present disclosure may also be provided in a server or a server cluster different from the server 105 and capable of communicating with the

terminal devices

101, 102, 103 and/or the server 105.

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.

The initialization parameter recommendation method of the algorithm of the disclosed embodiment will be described in detail below with reference to the scenario described in fig. 1 by fig. 2.

Fig. 2 schematically illustrates a flowchart of an initialization parameter recommendation method of an algorithm according to an embodiment of the present disclosure.

As shown in fig. 2, this embodiment includes operations S210 to S240, and the initialization parameter recommendation method of the algorithm may be performed by a server.

In the technical scheme of the disclosure, the processes of acquiring, collecting, storing, using, processing, transmitting, providing, disclosing, applying and the like of the data all conform to the regulations of related laws and regulations, necessary security measures are adopted, and the public order harmony is not violated.

In response to receiving a recommendation request from a user, input information corresponding to the recommendation request is acquired, the input information including a target algorithm and a first input model feature in operation S210.

In operation S220, according to the first history data corresponding to the target algorithm, the first history data includes numerical vectors of a plurality of history second input model features, and numerical vectors of target history second input model features matching the numerical vectors of the first input model features are determined, so as to obtain target first initialization parameter values corresponding to the target history second input model features.

In operation S230, according to the second history data corresponding to the target algorithm, the second history data includes a plurality of historical second initialization parameter values and model training performance values corresponding to each of the historical second initialization parameter values, and the target model training performance values are determined to obtain target second initialization parameter values corresponding to the target model training performance values.

In operation S240, an initialization parameter recommended value of the target algorithm is determined according to the target first initialization parameter value and the target second initialization parameter value, and the initialization parameter recommended value is recommended to the user.

It will be appreciated that the algorithm parameters (also called superparameters) differ from the model parameters (model parameters) in that the former are typically specified (default values may be used without specification) and the latter are self-learning by the model. By way of example, a neural network model is used in which both weights (weights) and biases (bias) are model parameters that can be updated automatically during the training of the network or learned by itself. Algorithm parameters are those that cannot be learned by the network by itself in training; for example, the hidden layer has 4 neurons, and the number of the neurons of the hidden layer is an algorithm parameter, which needs to be considered to be specified and cannot be automatically learned.

Typically, specifying this algorithm parameter depends on experience. If uncertain, brute force methods such as grid searching can be used, of course, the grid searching itself also requires specifying algorithm parameters. Furthermore, one model may be reused, and the algorithm parameters of another model may be predicted using this model. However, both empirical and model predictions may result in algorithm initialization being too single, e.g., some algorithms may not have a way to empirically initialize algorithm parameters; the information of some data sets is not easy to obtain, and the data sets cannot be classified, so that the algorithm parameters cannot be initialized. And, there is also a limitation in the choice of algorithm.

For example, it is known that the index of the classification algorithm C trained by 5 users with the A1 parameter and the B data is 0.9, and the index of the classification algorithm C trained by 5 users with the A2 parameter and the B data is 0.5. It is apparent that this approach does not integrate factors that affect how good the algorithm initialization parameters are, and does not fuse multiple factors together.

According to the method for recommending the initialization parameters of the algorithm, the historical record of the operation of the user on the machine learning platform can be collected, for example, which data set is used for modeling by the user, which algorithm is used, which initialization parameters are set, the training result of the model (which data set is used for modeling by the user, which algorithm is used, which initialization parameters are set, and the training result of the model (such as indexes of accuracy rate, recall rate and the like) can be collected, when the recommendation request is sent by the user, for example, after the training data set and algorithm are selected by the user, the input information corresponding to the recommendation request can be obtained, the input information comprises a target algorithm and first input model features, and the first input model features represent the training data set.

Further, the first history data may be specifically retrieved by a first policy, such as determining parameters based on similarity of content; vector conversion processing is carried out on the first input model; then, matching the vectors, determining the characteristics of the second input model of the target history, such as determining the most similar vector, and correspondingly taking the initialization parameters of the most similar vector as the values of the first initialization parameters of the target.

Further, it will be appreciated that, assuming that an initialization parameter with the best effect can be obtained after a certain algorithm uses different initialization parameters on the same data set, combining the parameters with the best effect obtained by the algorithm on different data sets can obtain a fusion of the optimal parameters of the historical algorithm. Therefore, the second strategy such as optimal parameter fusion based on a history algorithm can be used for determining parameters, and specifically, second history data can be called, and the condition of the initialization parameters characterized by the model training expression values corresponding to each history second initial parameter value is utilized to determine the target second initialization parameter value, such as the initialization parameters corresponding to the maximum model training expression values, as the target second initialization parameter value.

And finally, fusing a plurality of parameters to obtain a target parameter recommended value, for example, solving an average value of the target first initialization parameter value and the target second initialization parameter value, and calculating by a dynamic weighting method according to the target first initialization parameter value and the target second initialization parameter value, for example, recommending a proper algorithm initialization parameter value, namely an initialization parameter recommended value, for a user.

The second historical data further includes a model evaluation AUC index corresponding to each of the historical second initial parameter values, a degree of improvement index corresponding to each of the historical second initial parameter values, and a degree of differentiation KS index corresponding to each of the historical second initial parameter values, the method comprising: and determining model training representation values corresponding to each historical second initial parameter value respectively according to the model evaluation AUC index, the lifting degree index and the differentiation degree KS index.

It will be appreciated that the model evaluation AUC index, the boost index, and the differentiation KS index are all computable. If AUC is the area under the ROC curve, the AUC is used as an evaluation index of the classification model; the KS index is used for evaluating the distinguishing capability of the model on the good and bad clients, and calculating the maximum difference between the accumulated bad clients and the accumulated good client percentage.

For example, taking the case of a two-class modeling scenario of overdue prediction as an example, assuming algorithm u, n model training (including the result of search parameter tuning) is performed on data set d, the data that has been trained only once in the data set can be removed, then i (i)>1) The algorithm used for the model training was initialized with parameter p_ (i, i e {1, 2..n }), and the ith time then calculated, the AUC index of the algorithm on this dataset was

Degree of elevation index C _i The ks index is->

The combination of the indexes can quantitatively determine the advantages and disadvantages of the model using the initialization parameters, and then the algorithm optimizes the initialization parameters under the data set as follows:

wherein sig is a sigmoid function that converts different indices to the same dimension.

According to the initialization parameter recommendation method of the algorithm, the model training representation values corresponding to each historical second initial parameter value can be determined according to the model evaluation AUC index, the lifting degree index and the differentiation degree KS index, so that the fusion of the parameters with optimal historical performance of the target algorithm is facilitated, and the initialization parameters with best historical synthesis are obtained.

According to first historical data corresponding to a target algorithm, the first historical data comprises numerical vectors of a plurality of historical second input model features, the numerical vectors of target historical second input model features matched with the numerical vectors of the first input model features are determined, and target first initialization parameter values corresponding to the target historical second input model features are obtained, and the method comprises the following steps: according to first historical data corresponding to a target algorithm, the first historical data comprise numerical vectors of a plurality of historical second input model features, a cosine similarity calculation method is adopted to determine the numerical vectors of the target historical second input model features matched with the numerical vectors of the first input model features, and target first initialization parameter values corresponding to the target historical second input model features are obtained.

For example, taking the case of a two-class modeling scenario of overdue prediction as an example, after a user selects a certain modeling feature (e.g., a first input model feature) using a certain algorithm (e.g., a random forest), the modeling feature may be sampled (< 1000 lines), e.g., by ebedding, to transform the sampled modeling feature into a vector V of a specific dimension (e.g., 1×10). Then, all the in-mold features of the algorithm are used to perform the same process, and the in-mold feature vectors V '1, V ' 2..v ' n, which have historically used the algorithm, are calculated. Through nearest neighbor search, a vector which is most similar to the current input model feature (namely the first input model feature) of the user is calculated by using a cosine similarity, and then an initialization parameter corresponding to the vector is P_feature, and the initialization parameter is a series

Key pairs of columns, such as { "lr":0.9, "tree_deep":10}.

According to the initialization parameter recommendation method of the algorithm, which is provided by the embodiment, the first input model feature currently used by a user is sampled, the first input model feature is vectorized, then the numerical vector of the target historical second input model feature matched with the numerical vector of the first input model feature is determined by utilizing a cosine similarity calculation method, and therefore the target first initialization parameter value corresponding to the target historical second input model feature is obtained.

Determining an initialization parameter recommended value of a target algorithm according to the target first initialization parameter value and the target second initialization parameter value, and recommending the initialization parameter recommended value to a user, wherein the method comprises the following steps: calculating an average value according to the target first initialization parameter value and the target second initialization parameter value; and taking the average value as an initialization parameter recommended value of the target algorithm, and recommending the initialization parameter recommended value to a user.

It can be appreciated that, in order to achieve a better effect, the average value may be calculated according to the target first initialization parameter value and the target second initialization parameter value, so that the initialization parameter recommended value of the target algorithm is more reasonable by taking the average value as the initialization parameter recommended value of the target algorithm.

According to the method for recommending the initialization parameters of the algorithm, the average value obtained through calculation is used as the initialization parameter recommended value of the target algorithm, so that the initialization parameter recommended value of the target algorithm is more reasonable, and a better effect can be achieved for the model.

The initialization parameter recommendation method of the algorithm further comprises the following steps: in response to obtaining the target third initialization parameter value from the target object, an initialization parameter recommendation value for the target algorithm is determined based on the target first initialization parameter value, the target second initialization parameter value, and the target third initialization parameter value.

The target object may be a certain expert, and the target third initialization parameter value may be an empirical value given by the expert.

Fig. 3 schematically illustrates an implementation diagram of a user to obtain an initialization parameter recommendation value according to an embodiment of the present disclosure, see fig. 3. First, the target object sends the target third initialization parameter value, e.g., via the client 310. The server 320, in response to receiving the target third initialization parameter value, forwards the target third initialization parameter value to the initialization parameter recommender 330 of the algorithm. The initialization parameter recommendation device 330 of the algorithm further obtains the target first initialization parameter value and the target second initialization parameter value, and determines an initialization parameter recommendation value of the target algorithm according to the target first initialization parameter value, the target second initialization parameter value and the target third initialization parameter value. Finally, the initialization parameter recommendation value is transmitted to the client 310 corresponding to the user.

For example, the recommended algorithm parameter (the target first initialization parameter value) based on the similarity of the content is p_feature, the historical algorithm optimal parameter fusion result (the target second initialization parameter value) is p_index, and the expert experience recommended algorithm parameter (the target third initialization parameter value) is p_profile. Then the recommended value of the initialization parameter of the target algorithm recommended to the user at this time may be the average of the above three results, assuming that p_feature= { "lr":0.9, "tree_deep":10}, p_index= { "lr":0.7, "tree_deep":8}, p_feature= { "lr":0.6, "tree_deep":5}, then the initialization parameter recommended to the user may be p= { "lr":0.733, "tree_deep":8}.

According to the method for recommending the initialization parameters of the algorithm, the first initialization parameter value of the target corresponding to the similarity degree of modeling content, the second initialization parameter value of the target corresponding to the optimal parameter fusion result of the historical algorithm and the third initialization parameter value of the target corresponding to the expert experience parameter are combined to be used as the initialization parameters of the algorithm of the user recommendation, and the recommendation value of the initialization parameters obtained through multi-strategy fusion calculation is high in robustness and universality.

The invention further provides an initialization parameter recommendation device of the algorithm based on the initialization parameter recommendation method of the algorithm. The device will be described in detail below in connection with fig. 4.

Fig. 4 schematically shows a block diagram of the initialization parameter recommendation device of the algorithm according to an embodiment of the present disclosure.

As shown in fig. 4, the initialization parameter recommendation apparatus 400 of the algorithm of this embodiment includes an acquisition module 410, a first determination module 420, a second determination module 430, and a third determination module 440.

An obtaining module 410, configured to obtain, in response to receiving a recommendation request from a user, input information corresponding to the recommendation request, where the input information includes a target algorithm and a first input model feature; a first determining module 420, configured to determine, according to first historical data corresponding to the target algorithm, the first historical data including numerical vectors of a plurality of historical second input model features, the numerical vector of a target historical second input model feature that matches the numerical vector of the first input model feature, and obtain a target first initialization parameter value corresponding to the target historical second input model feature; a second determining module 430, configured to determine a target model training performance value according to second historical data corresponding to the target algorithm, where the second historical data includes a plurality of historical second initialization parameter values and model training performance values corresponding to each historical second initialization parameter value respectively, so as to obtain a target second initialization parameter value corresponding to the target model training performance value; and a third determining module 440, configured to determine an initialization parameter recommendation value of the target algorithm according to the target first initialization parameter value and the target second initialization parameter value, and recommend the initialization parameter recommendation value to the user.

In some embodiments, the second historical data further includes a model evaluation AUC index corresponding to each historical second initial parameter value, a degree of improvement index corresponding to each historical second initial parameter value, and a degree of differentiation KS index corresponding to each historical second initial parameter value, the apparatus comprising: and the first calculation module is used for determining the model training representation values corresponding to each historical second initial parameter value according to the model evaluation AUC index, the lifting degree index and the differentiation degree KS index.

In some embodiments, the first determining module includes: and the first determining submodule is used for determining the numerical vector of the target historical second input model feature matched with the numerical vector of the first input model feature by adopting a cosine similarity calculation method according to the first historical data corresponding to the target algorithm, so as to obtain a target first initialization parameter value corresponding to the target historical second input model feature.

In some embodiments, the second determining module includes: the second calculation module is used for calculating an average value according to the target first initialization parameter value and the target second initialization parameter value; and the recommending module is used for taking the average value as an initialization parameter recommending value of the target algorithm and recommending the initialization parameter recommending value to the user.

In some embodiments, the apparatus further comprises: and the fourth determining module is used for determining an initialization parameter recommended value of the target algorithm according to the target first initialization parameter value, the target second initialization parameter value and the target third initialization parameter value in response to obtaining the target third initialization parameter value from the target object.

According to an embodiment of the present disclosure, any of the acquisition module 410, the first determination module 420, the second determination module 430, and the third determination module 440 may be combined in one module to be implemented, or any of the modules may be split into a plurality of modules. Alternatively, at least some of the functionality of one or more of the modules may be combined with at least some of the functionality of other modules and implemented in one module. According to embodiments of the present disclosure, at least one of the acquisition module 410, the first determination module 420, the second determination module 430, and the third determination module 440 may be implemented at least in part as hardware circuitry, such as a Field Programmable Gate Array (FPGA), a Programmable Logic Array (PLA), a system on a chip, a system on a substrate, a system on a package, an Application Specific Integrated Circuit (ASIC), or may be implemented in hardware or firmware in any other reasonable manner of integrating or packaging the circuitry, or in any one of or a suitable combination of three of software, hardware, and firmware. Alternatively, at least one of the acquisition module 410, the first determination module 420, the second determination module 430, and the third determination module 440 may be at least partially implemented as computer program modules, which when executed, may perform the respective functions.

As shown in fig. 5, an electronic device 500 according to an embodiment of the present disclosure includes a processor 501 that can perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM) 502 or a program loaded from a storage section 508 into a Random Access Memory (RAM) 503. The processor 501 may include, for example, a general purpose microprocessor (e.g., a CPU), an instruction set processor and/or an associated chipset and/or a special purpose microprocessor (e.g., an Application Specific Integrated Circuit (ASIC)), or the like. The processor 501 may also include on-board memory for caching purposes. The processor 501 may comprise a single processing unit or a plurality of processing units for performing different actions of the method flows according to embodiments of the disclosure.

In the RAM 503, various programs and data required for the operation of the electronic apparatus 500 are stored. The processor 501, ROM502, and RAM 503 are connected to each other by a bus 504. The processor 501 performs various operations of the method flow according to the embodiments of the present disclosure by executing programs in the ROM502 and/or the RAM 503. Note that the program may be stored in one or more memories other than the ROM502 and the RAM 503. The processor 501 may also perform various operations of the method flow according to embodiments of the present disclosure by executing programs stored in the one or more memories.

According to an embodiment of the present disclosure, the electronic device 500 may also include an input/output (I/O) interface 505, the input/output (I/O) interface 505 also being connected to the bus 504. The electronic device 500 may also include one or more of the following components connected to the I/O interface 505: an input section 506 including a keyboard, a mouse, and the like; an output portion 507 including a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and the like, and a speaker, and the like; a storage portion 508 including a hard disk and the like; and a communication section 509 including a network interface card such as a LAN card, a modem, or the like. The communication section 509 performs communication processing via a network such as the internet. The drive 510 is also connected to the I/O interface 505 as needed. A removable medium 511 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 510 as needed so that a computer program read therefrom is mounted into the storage section 508 as needed.

The present disclosure also provides a computer-readable storage medium that may be embodied in the apparatus/device/system described in the above embodiments; or may exist alone without being assembled into the apparatus/device/system. The computer-readable storage medium carries one or more programs which, when executed, implement methods in accordance with embodiments of the present disclosure.

According to embodiments of the present disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium, which may include, for example, but is not limited to: a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this disclosure, a computer-readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. For example, according to embodiments of the present disclosure, the computer-readable storage medium may include ROM 502 and/or RAM 503 and/or one or more memories other than ROM 502 and RAM 503 described above.

Embodiments of the present disclosure also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowcharts. The program code, when executed in a computer system, is operative to cause the computer system to implement an initialization parameter recommendation method for an algorithm provided by embodiments of the present disclosure.

The above-described functions defined in the system/apparatus of the embodiments of the present disclosure are performed when the computer program is executed by the processor 501. The systems, apparatus, modules, units, etc. described above may be implemented by computer program modules according to embodiments of the disclosure.

In one embodiment, the computer program may be based on a tangible storage medium such as an optical storage device, a magnetic storage device, or the like. In another embodiment, the computer program may also be transmitted, distributed, and downloaded and installed in the form of a signal on a network medium, and/or installed from a removable medium 511 via the communication portion 509. The computer program may include program code that may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the foregoing.

In such an embodiment, the computer program may be downloaded and installed from a network via the communication portion 509, and/or installed from the removable media 511. The above-described functions defined in the system of the embodiments of the present disclosure are performed when the computer program is executed by the processor 501. The systems, devices, apparatus, modules, units, etc. described above may be implemented by computer program modules according to embodiments of the disclosure.

According to embodiments of the present disclosure, program code for performing computer programs provided by embodiments of the present disclosure may be written in any combination of one or more programming languages, and in particular, such computer programs may be implemented in high-level procedural and/or object-oriented programming languages, and/or assembly/machine languages. Programming languages include, but are not limited to, such as Java, c++, python, "C" or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, partly on a remote computing device, or entirely on the remote computing device or server. In the case of remote computing devices, the remote computing device may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., connected via the Internet using an Internet service provider).

The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams or flowchart illustration, and combinations of blocks in the block diagrams or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

Those skilled in the art will appreciate that the features recited in the various embodiments of the disclosure and/or in the claims may be combined in various combinations and/or combinations, even if such combinations or combinations are not explicitly recited in the disclosure. In particular, the features recited in the various embodiments of the present disclosure and/or the claims may be variously combined and/or combined without departing from the spirit and teachings of the present disclosure. All such combinations and/or combinations fall within the scope of the present disclosure.

The embodiments of the present disclosure are described above. However, these examples are for illustrative purposes only and are not intended to limit the scope of the present disclosure. Although the embodiments are described above separately, this does not mean that the measures in the embodiments cannot be used advantageously in combination. The scope of the disclosure is defined by the appended claims and equivalents thereof. Various alternatives and modifications can be made by those skilled in the art without departing from the scope of the disclosure, and such alternatives and modifications are intended to fall within the scope of the disclosure.

Claims

1. An initialization parameter recommendation method of an algorithm comprises the following steps:

in response to receiving a recommendation request from a user, obtaining input information corresponding to the recommendation request, wherein the input information comprises a target algorithm and a first input model feature;

According to first historical data corresponding to the target algorithm, the first historical data comprise numerical vectors of a plurality of historical second input model features, numerical vectors of target historical second input model features matched with the numerical vectors of the first input model features are determined, and target first initialization parameter values corresponding to the target historical second input model features are obtained;

according to second historical data corresponding to the target algorithm, the second historical data comprise a plurality of historical second initialization parameter values and model training performance values corresponding to each historical second initialization parameter value respectively, and a target model training performance value is determined to obtain a target second initialization parameter value corresponding to the target model training performance value; and

and determining an initialization parameter recommended value of the target algorithm according to the target first initialization parameter value and the target second initialization parameter value, and recommending the initialization parameter recommended value to the user.

2. The method of claim 1, wherein the second historical data further includes a model evaluation AUC index corresponding to each historical second initial parameter value, a degree of improvement index corresponding to each historical second initial parameter value, and a degree of differentiation KS index corresponding to each historical second initial parameter value, the method comprising:

And determining the model training representation values corresponding to each historical second initial parameter value according to the model evaluation AUC index, the lifting degree index and the differentiation degree KS index.

3. The method of claim 1, wherein the determining, from the first historical data corresponding to the target algorithm, the first historical data including numerical vectors of a plurality of historical second input model features, the numerical vector of a target historical second input model feature that matches the numerical vector of the first input model feature, results in a target first initialization parameter value corresponding to the target historical second input model feature, comprising:

and determining the numerical vector of the target historical second input model feature matched with the numerical vector of the first input model feature by adopting a cosine similarity calculation method according to the first historical data corresponding to the target algorithm, and obtaining a target first initialization parameter value corresponding to the target historical second input model feature.

4. The method of claim 1, wherein the determining an initialization parameter recommendation value for the target algorithm from the target first initialization parameter value and the target second initialization parameter value and recommending the initialization parameter recommendation value to the user comprises:

Calculating an average value according to the target first initialization parameter value and the target second initialization parameter value; and

and taking the average value as an initialization parameter recommended value of the target algorithm, and recommending the initialization parameter recommended value to the user.

5. The method of claim 1, further comprising:

in response to obtaining a target third initialization parameter value from a target object, an initialization parameter recommendation value for the target algorithm is determined from the target first initialization parameter value, the target second initialization parameter value, and the target third initialization parameter value.

6. An initialization parameter recommendation device of an algorithm, comprising:

the acquisition module is used for responding to a recommendation request received from a user and acquiring input information corresponding to the recommendation request, wherein the input information comprises a target algorithm and a first input model feature;

the first determining module is used for determining the numerical vector of the target historical second input model feature matched with the numerical vector of the first input model feature according to the first historical data corresponding to the target algorithm, wherein the first historical data comprises the numerical vectors of a plurality of historical second input model features, and obtaining a target first initialization parameter value corresponding to the target historical second input model feature;

The second determining module is used for determining a target model training representation value according to second historical data corresponding to the target algorithm, wherein the second historical data comprises a plurality of historical second initialization parameter values and model training representation values corresponding to each historical second initialization parameter value respectively, and obtaining a target second initialization parameter value corresponding to the target model training representation value; and

and the third determining module is used for determining an initialization parameter recommended value of the target algorithm according to the target first initialization parameter value and the target second initialization parameter value and recommending the initialization parameter recommended value to the user.

7. The apparatus of claim 6, wherein the second historical data further comprises a model evaluation AUC index corresponding to each historical second initial parameter value, a degree of elevation index corresponding to each historical second initial parameter value, and a degree of differentiation KS index corresponding to each historical second initial parameter value, the apparatus comprising:

and the first calculation module is used for determining the model training representation values corresponding to each historical second initial parameter value according to the model evaluation AUC index, the lifting degree index and the differentiation degree KS index.

8. An electronic device, comprising:

one or more processors;

storage means for storing one or more programs,

wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to perform the method of any of claims 1-5.

9. A computer readable storage medium having stored thereon executable instructions which, when executed by a processor, cause the processor to perform the method according to any of claims 1-5.

10. A computer program product comprising a computer program which, when executed by a processor, implements the method according to any one of claims 1 to 5.