WO2020252925A1 - 用户特征群中用户特征寻优方法、装置、电子设备及计算机非易失性可读存储介质 - Google Patents

用户特征群中用户特征寻优方法、装置、电子设备及计算机非易失性可读存储介质 Download PDF

Info

Publication number
WO2020252925A1
WO2020252925A1 PCT/CN2019/103437 CN2019103437W WO2020252925A1 WO 2020252925 A1 WO2020252925 A1 WO 2020252925A1 CN 2019103437 W CN2019103437 W CN 2019103437W WO 2020252925 A1 WO2020252925 A1 WO 2020252925A1
Authority
WO
WIPO (PCT)
Prior art keywords
user feature
user
score
combination
score range
Prior art date
Application number
PCT/CN2019/103437
Other languages
English (en)
French (fr)
Inventor
邓悦
金戈
徐亮
Original Assignee
平安科技(深圳)有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安科技(深圳)有限公司 filed Critical 平安科技(深圳)有限公司
Publication of WO2020252925A1 publication Critical patent/WO2020252925A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00Machine learning
    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02TCLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00Road transport of goods or passengers
    • Y02T10/10Internal combustion engine [ICE] based vehicles
    • Y02T10/40Engine management systems

Definitions

  • This application relates to the technical field of machine learning applications, and in particular to a method, device, electronic device, and computer non-volatile readable storage medium for optimizing user characteristics in a user characteristic group.
  • the optimization of user characteristics in a user characteristic group is to find a special purpose user characteristic combination that meets the needs from a user characteristic group with a large number of user characteristics, that is, to find an optimized user characteristic from the user characteristic group. For example, searching for a combination of user characteristics that meets the needs from the user characteristics group, in order to obtain a group of user combinations that can meet a predetermined purpose.
  • the optimization of user characteristics in the user characteristic group is mainly by crawling keywords from the user characteristic group, setting corresponding weight coefficients based on different keywords, and then calculating the user characteristic combinations that meet the needs through the weight coefficients.
  • the inventor of the present application realizes that in the process of optimizing user characteristics, the difference between user characteristics in all user characteristic groups cannot be considered, and the obtained user characteristic combination has the problem of unreasonable combination, and the user characteristic
  • the process of combined acquisition requires a large amount of data statistics, the algorithm load is large, and the acquisition efficiency is low.
  • an object of the present application is to provide a method, device, electronic device, and computer non-volatile readable storage medium for optimizing user characteristics in a user characteristic group.
  • a method for optimizing user characteristics in a user characteristic group includes: initializing and generating a plurality of user characteristic combinations based on the target user characteristic group; and inputting each of the user characteristic combinations and optimization targets into a pre-trained first A machine learning model to obtain an evaluation score for each of the user feature combinations; to obtain a plurality of user feature combinations corresponding to the evaluation scores in the first score range, and multiple user feature combinations corresponding to the evaluation scores in the second score range , Wherein the evaluation scores in the plurality of first score ranges are greater than a first predetermined threshold, and the evaluation scores in the plurality of second score ranges are lower than a second predetermined threshold; and each of the first score ranges is obtained
  • the user feature combination is distinguished from the user feature combination of each second score range to obtain multiple user feature differences; the multiple user feature differences and optimization goals are input into the pre-trained second machine learning model to obtain The prediction of the target user feature group optimizes the user feature combination; the user feature combination is optimized based on the prediction, and the target optimized user feature
  • a device for optimizing user characteristics in a user characteristic group is characterized by comprising: an initialization module, which is used to initialize and generate multiple user characteristic combinations based on the target user characteristic group; The user feature combination and the optimization target are respectively input to the pre-trained first machine learning model to obtain the evaluation score of each user feature combination; the first acquisition module is used to acquire the evaluation scores in a plurality of first score ranges Corresponding user feature combinations and user feature combinations corresponding to the evaluation scores in the plurality of second score ranges, wherein the evaluation scores in the plurality of first score ranges are greater than a first predetermined threshold, and the plurality of second scores The evaluation score in the range is lower than the second predetermined threshold; the second acquisition module is used to acquire the user feature difference between each user feature combination in the first score range and each user feature combination in the second score range to obtain A user feature difference; a prediction module, used to input the multiple user feature differences and optimization goals into a pre-trained second machine learning model to obtain a predicted and optimized user feature combination of the
  • a device for optimizing user characteristics in a user characteristic group includes: a processor; and
  • the memory is configured to store the user feature optimization program in the user feature group of the processor; wherein the processor is configured to execute the user feature group as described above by executing the user feature optimization program in the user feature group User characteristics optimization method in middle.
  • a computer non-volatile readable storage medium stores a user characteristic optimization program in a user characteristic group, wherein the user characteristic optimization program in the user characteristic group is executed by a processor Realize the user feature optimization method in the user feature group as described above.
  • the machine learning model is used to accurately and efficiently predict the user feature combination corresponding to the optimization target according to the distinguishing feature of the partial user feature combination, thereby ensuring the efficiency and accuracy of the target user feature combination acquisition.
  • Fig. 1 schematically shows a flowchart of a method for optimizing user characteristics in a user characteristic group.
  • FIG. 2 schematically shows an example diagram of an application scenario of a method for optimizing user characteristics in a user characteristic group.
  • Fig. 3 schematically shows a flow chart of a method for obtaining user characteristic differences.
  • Fig. 4 schematically shows a block diagram of an apparatus for optimizing user characteristics in a user characteristic group.
  • Fig. 5 shows a block diagram of an electronic device for implementing the above-mentioned method for optimizing user characteristics in a user characteristic group according to an exemplary embodiment.
  • Fig. 6 shows a schematic diagram of a computer non-volatile readable storage medium for implementing the above-mentioned method for optimizing user characteristics in a user characteristic group according to an exemplary embodiment.
  • This example embodiment first provides a method for optimizing user characteristics in a user characteristic group.
  • the method for optimizing user characteristics in a user characteristic group can be run on a server, a server cluster or a cloud server, etc.
  • the method of the present invention can also be run on other platforms according to requirements, which is not specifically limited in this exemplary embodiment.
  • the method for optimizing user characteristics in the user characteristic group may include the following steps:
  • Step S110 based on the target user feature group, initialize and generate multiple user feature combinations
  • Step S120 inputting each of the user feature combinations and optimization targets into a pre-trained first machine learning model to obtain an evaluation score of each of the user feature combinations;
  • Step S130 Obtain user feature combinations corresponding to the evaluation scores in a plurality of first score ranges, and user feature combinations corresponding to the evaluation scores in a plurality of second score ranges, wherein the evaluations in the plurality of first score ranges The score is greater than a first predetermined threshold, and the evaluation scores in the plurality of second score ranges are lower than a second predetermined threshold;
  • Step S140 Obtain user feature differences between each user feature combination in the first score range and each user feature combination in the second score range to obtain multiple user feature differences;
  • Step S150 inputting the plurality of user feature differences and optimization goals into a pre-trained second machine learning model to obtain a predicted and optimized user feature combination of the target user feature group;
  • Step S160 Based on the prediction to optimize the user feature combination, obtain a target optimized user feature combination from the target user feature group.
  • the user feature combinations corresponding to the evaluation scores in the multiple first score ranges and the user feature combinations corresponding to the evaluation scores in the multiple second score ranges, wherein the evaluation scores in the multiple first score ranges Is greater than the first predetermined threshold, the evaluation scores in the multiple second score ranges are lower than the second predetermined threshold; in this way, multiple user feature combinations can be accurately classified based on the evaluation scores, and further, by obtaining each of the first
  • the user feature combination of the score range is different from the user feature combination of each second score range, and multiple user feature differences can be obtained; the user feature difference of the series can be accurately obtained according to the comparison of the user feature combinations of different score categories, which is effective To ensure the reliability of the distinguishing feature in the subsequent steps to optimize the user feature combination analysis.
  • the multiple user feature differences and optimization goals are input into the pre-trained second machine learning model to obtain the predicted and optimized user feature combination of the target user feature group; this can be efficiently and accurately based on the user features of the series Differentiate predictions to optimize user feature combinations. Furthermore, based on the predicted and optimized user feature combination, the target optimized user feature combination can be efficiently and accurately obtained from the target user feature group. In this way, the machine learning model accurately and efficiently predicts the user feature combination corresponding to the optimization target based on the distinguishing feature of the partial user feature combination, thereby ensuring the efficiency and accuracy of the target user feature combination acquisition.
  • step S110 based on the target user feature group, a plurality of user feature combinations are initialized and generated.
  • the server 201 obtains the target user characteristic group from the platform server 202, and then initializes and generates multiple user characteristic combinations based on the target user characteristic group. In this way, the server 201 can perform the combined feature analysis of the user feature combination according to the multiple user feature combinations generated by initialization in the subsequent steps, and then can obtain the optimized target user combination feature. It can be understood that, according to requirements, the server 202 can also directly generate multiple user feature combinations based on the target user feature group. In the subsequent steps, the user feature combination feature analysis is performed according to the multiple user feature combinations generated by the initialization, and then the optimized target user feature is obtained.
  • the server 201 and the server 202 can be any devices with processing capabilities, for example, computers, microprocessors, cloud servers, etc., which are not specifically limited herein.
  • the target user characteristic group is a user characteristic group that contains a large number of user characteristics, for example, a large number of user characteristics on a shopping platform or a communication platform.
  • Based on the target user feature group initialize multiple user feature combinations, that is, obtain multiple user features from the target user feature group and combine them into user feature combinations randomly or in a predetermined order, and then obtain multiple user feature combinations In this way, it is possible to quickly obtain a few user feature combinations among all possible user feature combinations for evaluation in subsequent steps.
  • the number of acquired user feature combinations can be permutation and combination according to the number of user features in the target user feature group to obtain a predetermined ratio of user feature combinations.
  • the predetermined ratio is set according to requirements, and can be, for example, 0.001%, 0.002%, etc. .
  • the initial generation of multiple user feature combinations based on the target user feature group includes:
  • a predetermined number of user features are acquired in sequence, wherein the predetermined number at each acquisition is a predetermined number more than the previous one; and the predetermined number of user features acquired are sequentially acquired , Arranged into a user characteristic string according to the permutation rule of the target user characteristic group, to obtain multiple user characteristic combinations.
  • a predetermined number of user features are acquired in sequence, wherein the predetermined number at each acquisition is a predetermined number more than the previous one, that is, for example, from a user feature group with 10,000 user features In order to obtain 100, 110..., 200 user characteristics, each time the number of user characteristics is 10 more than the previous one. Then, the obtained predetermined number of user characteristics are sequentially arranged into a user characteristic string according to the permutation rule of the target user characteristic group to obtain multiple user characteristic combinations. In this way, when the target user characteristic group has its own user characteristic arrangement rule, it can be ensured that the obtained user characteristic combination carries the user characteristic group characteristic, and the accuracy of the subsequent step analysis is guaranteed.
  • the arrangement rule may be, for example, arrangement in a predetermined order or arrangement in a predetermined table form.
  • the initial generation of multiple user feature combinations based on the target user feature group includes:
  • a predetermined number of user feature combinations are randomly initialized and generated.
  • a predetermined number of user feature combinations are generated through random initialization, which can ensure the randomness of the user feature combination acquisition, and ensure the efficiency of the user feature combination acquisition when the user feature group is not arranged in an order.
  • step S120 each of the user feature combinations and optimization targets are respectively input into the pre-trained first machine learning model to obtain an evaluation score for each of the user feature combinations.
  • each user feature combination and optimization target are respectively input into a pre-trained machine learning model, and the machine learning model can evaluate the suitability of each user feature combination and the optimization target according to the optimization target , That is, the evaluation score of each user feature combination.
  • the optimization goal is the purpose of the user feature combination, for example, whether certain user feature user feature combinations can be adapted to the user combination needs of a certain project.
  • the training method of the first machine learning model :
  • Each sample in the sample set is calibrated in advance with a corresponding evaluation score; the input data of each sample in the sample set is input into the machine learning model to obtain the corresponding Evaluation score; if the input data of the sample is input into the machine learning model, the evaluation score corresponding to the sample is inconsistent with the pre-calibrated evaluation score of the sample, then the coefficient of the learning model is adjusted until it is consistent; when the input of all samples After the data is input into the machine learning model, the obtained evaluation score corresponding to each sample is consistent with the pre-calibrated evaluation score for each sample, and the training ends.
  • User feature combinations and optimization target samples are user feature combinations and optimization target samples corresponding to a certain user feature group in history.
  • each sample is calibrated by experts in advance with a corresponding evaluation score as the output of the machine learning model. Then, the coefficients are adjusted so that when the input data of all samples is input into the machine learning model, the evaluation score corresponding to each sample is consistent with the pre-calibrated evaluation score for each of the samples.
  • the trained first is obtained.
  • a learning model can effectively ensure the accuracy of training.
  • step S130 the user feature combinations corresponding to the evaluation scores in the plurality of first score ranges and the user feature combinations corresponding to the evaluation scores in the plurality of second score ranges are obtained, wherein the plurality of first score ranges are The evaluation score of is greater than the first predetermined threshold, and the evaluation scores in the plurality of second score ranges are lower than the second predetermined threshold.
  • the first predetermined threshold is a set threshold with a higher judgment evaluation score
  • the second predetermined threshold is a set threshold with a lower judgment evaluation score.
  • the multiple first score ranges greater than the first predetermined threshold are, for example, when the first predetermined threshold is 60, the range is 60-65, 65-70, etc. multiple first score ranges, and multiple second score ranges below the second predetermined threshold
  • the score range is, for example, when the second predetermined threshold is 50, the range is 50-45, 45-40 and other multiple second score ranges.
  • the feature combination can accurately classify multiple user feature combinations according to the evaluation scores, and obtain user feature combinations of different levels corresponding to the optimization target.
  • the first predetermined threshold is equal to the second predetermined threshold as required; in another example, the first predetermined threshold is greater than the second predetermined threshold, and the score between the first predetermined threshold and the second predetermined threshold is discarded Range of user portfolio characteristics.
  • the method further includes:
  • the first machine learning model is pre-trained to obtain the evaluation score of each of the user feature combinations; among all the evaluation scores, the user feature combinations corresponding to the evaluation scores in multiple first score ranges that are greater than the first predetermined threshold are obtained , And user feature combinations corresponding to evaluation scores in multiple second score ranges that are lower than the second predetermined threshold.
  • the first predetermined threshold is 60
  • the range is 60-65, 65-70 and other multiple first score ranges, where the 60-65 range does not obtain the user feature combination.
  • step S140 the user feature differences between each user feature combination in the first score range and each user feature combination in the second score range are obtained to obtain multiple user feature differences.
  • each first score range corresponds to the second score range, which is a second score range pre-associated with each first score range.
  • multiple first score ranges greater than the first predetermined threshold are, for example, when the first predetermined threshold is 60, the range is 60-65, 65-70, etc. multiple first score ranges, and multiple first score ranges below the second predetermined threshold
  • the second score range is, for example, when the second predetermined threshold is 50, the range is 50-45, 45-40 and other multiple second score ranges.
  • the first score range of 60-65 corresponds to the second score range of 50-45.
  • Score range the first score range of 65-70 corresponds to the second score range of 45-40.
  • the user feature difference may be a user feature feature mapping vector combined from a user feature with a high rating or a conversion vector of a user feature feature combined with a user feature with a low rating or a difference between a converted character string.
  • the mapping vector of user feature features is obtained by vector mapping of user feature features. For example, a high-scoring user feature feature is 50 males and 120 females, and the mapping is (1, 50, 21, 120, 22), where 1 is the gender dimension vector , 50 and 120 are the number dimensions, 21 and 22 are the gender attribute value dimensions.
  • the feature vector of a low-rated user is (1, 10, 21, 160, 22)
  • their difference is expressed as (1, 50, 21, 120, 22)-(1, 10, 21, 160, 22).
  • the different technical features are serialized based on the evaluation of each user feature combination, which effectively guarantees the accuracy of the user feature combination in the subsequent steps based on the user feature differentiation prediction.
  • the obtaining the user feature difference between each user feature combination of the first score range and each user feature combination of the second score range to obtain multiple user feature differences includes: Randomly acquiring a user characteristic combination from each user characteristic combination in the first score range; randomly acquiring a user characteristic combination from each user characteristic combination in the second score range;
  • each user characteristic combination in each score range can represent the user characteristic combination characteristics of the range. Furthermore, by randomly obtaining a user feature combination from each user feature combination in the first score range; randomly obtaining a user feature combination from each user feature combination in the second score range; each user feature combination can be obtained efficiently.
  • a user feature combination in the first score range is distinguished from a user feature combination in the second score range corresponding to each first score range to obtain multiple user feature differences.
  • the acquisition of the user feature difference between each user feature combination of the first score range and each user feature combination of the second score range is obtained, and multiple users are obtained
  • the feature distinction includes: step S310, acquiring a first user feature combination element of each user feature combination in the first score range;
  • Step S320 Acquire the second user feature combination element of the user feature combination of the second score range corresponding to each first score range;
  • Step S330 Acquire each first user feature combination element and each first user feature combination element 2. Distinguishing features of the user feature combination elements, and obtaining user feature differences between each user feature combination in the first score range and each user feature combination in the second score range.
  • User feature combination elements include at least the number of user features in the user feature combination, preset feature tags of user features, and attribute values of multiple attributes of user features, etc.
  • the preset feature tags of user features are, for example, the identity of a user, user features
  • the attribute values of the multiple attributes are, for example, the gender of the user, male and female, age, and so on.
  • step S150 the plurality of user feature differences and optimization targets are input into a pre-trained second machine learning model to obtain a predicted and optimized user feature combination of the target user feature group.
  • the multiple user feature differences are the user feature differences in the series range, which can accurately reflect the user feature combination score change and the user feature combination feature change, and furthermore, the multiple user feature distinction and optimization target input are pre-trained
  • a good second machine learning model can efficiently and accurately obtain the target user feature group based on the differentiation analysis of the user characteristics from the optimization target to predict and optimize the user feature combination adapted to the optimization target, ensuring that the target is obtained from the user feature group in the subsequent steps The efficiency and accuracy of user feature combinations.
  • the training method of the second machine learning model collect user feature differences and optimization target sample sets, each sample in the sample set is calibrated in advance with a corresponding optimized user feature combination; The input data of each sample in the sample set is input into the machine learning model, and the optimized user feature combination corresponding to each sample is obtained; if there is input data of the sample, after inputting the machine learning model, the optimized user feature combination and pair corresponding to the sample are obtained.
  • the optimized user feature combinations calibrated in advance by the samples are inconsistent, adjust the coefficients of the learning model until they are consistent; when the input data of all samples is input into the machine learning model, the optimized user feature combinations corresponding to each sample obtained are compared with each The optimized user feature combinations calibrated in advance by the samples are consistent, and the training ends.
  • the user characteristic distinction and optimization target sample are the user characteristic distinction and optimization target samples corresponding to a certain user characteristic group in history.
  • each sample is calibrated by experts in advance with the corresponding optimized user feature combination as the output of the machine learning model. Then, the coefficients are adjusted so that when the input data of all samples is input into the machine learning model, the optimized user feature combination corresponding to each sample is consistent with the optimized user feature combination calibrated in advance for each sample.
  • the training ends, and the obtained The trained second learning model can effectively ensure the accuracy of training.
  • step S160 based on the predicted optimized user feature combination, a target optimized user feature combination is obtained from the target user feature group.
  • the user feature combination is optimized based on prediction, and the target user feature group can be accurately obtained from the target user feature group through the user feature feature tag or user feature attribute name, etc. from the target user feature group.
  • this application uses a machine learning model to accurately and efficiently predict the user feature combination corresponding to the optimization target based on the distinguishing features of some user feature combinations, thereby ensuring the efficiency and accuracy of the target user feature combination acquisition.
  • the user characteristic optimization device in the user characteristic group may include a user characteristic initialization module 410, an evaluation module 420, a first acquisition module 430, a second acquisition module 440, a prediction module 450 and an optimized characteristic acquisition module 460.
  • the initialization module 410 can be used to initialize and generate multiple user feature combinations based on the target user feature group; the evaluation module 420 can be used to input each of the user feature combinations and optimization targets into the pre-trained first machine learning Model to obtain the evaluation score of each of the user feature combinations; the first obtaining module 430 can be used to obtain the user feature combinations corresponding to the evaluation scores in the first score range, and the evaluation scores in the second score range The corresponding user feature combination, wherein the evaluation scores in the plurality of first score ranges are greater than a first predetermined threshold, and the evaluation scores in the plurality of second score ranges are lower than a second predetermined threshold; the second acquisition module 440 It can be used to obtain the user feature difference between each user feature combination in the first score range and the user feature combination in each second score range to obtain multiple user feature differences; the prediction module 450 can be used to combine the multiple User feature distinction and optimization target input the pre-trained second machine learning model to obtain the predicted optimized user feature combination of the target user feature group; the optimized feature acquisition module
  • modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory.
  • the features and functions of two or more modules or units described above may be embodied in one module or unit.
  • the features and functions of a module or unit described above can be further divided into multiple modules or units to be embodied.
  • the technical solution according to the embodiments of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (can be a CD-ROM, U disk, mobile hard disk, etc.) or on the network , Including several instructions to make a computing device (which can be a personal computer, server, mobile terminal, or network device, etc.) execute the method according to the embodiment of the present application.
  • a non-volatile storage medium can be a CD-ROM, U disk, mobile hard disk, etc.
  • a computing device which can be a personal computer, server, mobile terminal, or network device, etc.
  • the electronic device 500 according to this embodiment of the present invention will be described below with reference to FIG. 5.
  • the electronic device 500 shown in FIG. 5 is only an example, and should not bring any limitation to the function and application scope of the embodiment of the present invention.
  • the electronic device 500 is represented in the form of a general-purpose computing device.
  • the components of the electronic device 500 may include, but are not limited to: the aforementioned at least one processing unit 510, the aforementioned at least one storage unit 520, and a bus 530 connecting different system components (including the storage unit 520 and the processing unit 510).
  • the storage unit stores program code, and the program code can be executed by the processing unit 510, so that the processing unit 510 executes the various exemplary methods described in the "Exemplary Methods" section of this specification. Implementation steps.
  • the processing unit 510 may perform step S110 as shown in FIG.
  • Step S120 based on the target user feature group, initialize and generate a plurality of user feature combinations; S120: input each of the user feature combinations and optimization targets into the pre- The trained first machine learning model obtains the evaluation score of each of the user feature combinations; Step S130: Obtain the user feature combinations corresponding to the evaluation scores in the multiple first score ranges, and the user feature combinations in the multiple second score ranges The user feature combination corresponding to the evaluation score, wherein the evaluation scores in the plurality of first score ranges are greater than a first predetermined threshold, and the evaluation scores in the plurality of second score ranges are lower than a second predetermined threshold; step S140: Obtain the user feature differences between each user feature combination in the first score range and each user feature combination in the second score range to obtain multiple user feature differences; step S150: distinguish and optimize the multiple user features The second machine learning model pre-trained by the target input is used to obtain the predicted optimized user feature combination of the target user feature group; step S160: based on the prediction and optimized user feature combination, the target optimized user is obtained
  • the storage unit 520 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 5201 and/or a cache storage unit 5202, and may further include a read-only storage unit (ROM) 5203.
  • the storage unit 520 may also include a program/utility tool 5204 having a set (at least one) program module 5205.
  • program module 5205 includes but is not limited to: an operating system, one or more application programs, other program modules, and program data, Each of these examples or some combination may include the implementation of a network environment.
  • the bus 530 may represent one or more of several types of bus structures, including a storage unit bus or a storage unit controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local area using any bus structure among multiple bus structures. bus.
  • the electronic device 500 may also communicate with one or more external devices 700 (such as keyboards, pointing devices, Bluetooth devices, etc.), and may also communicate with one or more devices that enable customers to interact with the electronic device 500, and/or communicate with Any device (such as a router, modem, etc.) that enables the electronic device 500 to communicate with one or more other computing devices. This communication can be performed through an input/output (I/O) interface 550.
  • the electronic device 500 may also communicate with one or more networks (for example, a local area network (LAN), a wide area network (WAN), and/or a public network, such as the Internet) through the network adapter 560. As shown in the figure, the network adapter 560 communicates with other modules of the electronic device 500 through the bus 530.
  • LAN local area network
  • WAN wide area network
  • public network such as the Internet
  • the technical solution according to the embodiments of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (can be a CD-ROM, U disk, mobile hard disk, etc.) or on the network , Including several instructions to make a computing device (which may be a personal computer, server, terminal device, or network device, etc.) execute the method according to the embodiment of the present application.
  • a non-volatile storage medium can be a CD-ROM, U disk, mobile hard disk, etc.
  • a computing device which may be a personal computer, server, terminal device, or network device, etc.
  • a computer-readable storage medium is also provided, on which is stored a program product capable of implementing the foregoing method of this specification.
  • various aspects of the present invention may also be implemented in the form of a program product, which includes program code, and when the program product runs on a terminal device, the program code is used to make the The terminal device executes the steps according to various exemplary embodiments of the present invention described in the above "Exemplary Method" section of this specification. Referring to FIG. 6, a program product 600 for implementing the above method according to an embodiment of the present invention is described.
  • the readable storage medium can be any tangible medium that contains or stores a program, and the program can be used by or in combination with an instruction execution system, device, or device.
  • the program product can use any combination of one or more readable media.
  • the readable medium may be a readable signal medium or a readable storage medium.
  • the readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or a combination of any of the above.
  • readable storage media include: electrical connections with one or more wires, portable disks, hard disks, random access memory (RAM), read only memory (ROM), erasable Type programmable read only memory (EPROM or flash memory), optical fiber, portable compact disk read only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above.
  • the computer-readable signal medium may include a data signal propagated in baseband or as a part of a carrier wave, and readable program code is carried therein. This propagated data signal can take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing.
  • the readable signal medium may also be any readable medium other than a readable storage medium, and the readable medium may send, propagate, or transmit a program for use by or in combination with the instruction execution system, apparatus, or device.
  • the program code contained on the readable medium can be transmitted by any suitable medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the foregoing.
  • the program code used to perform the operations of the present invention can be written in any combination of one or more programming languages.
  • the programming languages include object-oriented programming languages—such as Java, C++, etc., as well as conventional procedural styles. Programming language-such as "C" language or similar programming language.
  • the program code can be executed entirely on the client computing device, partly executed on the client device, executed as a stand-alone software package, partly executed on the client computing device and partly executed on the remote computing device, or entirely on the remote computing device or server Executed on.
  • the remote computing device can be connected to the client computing device through any kind of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (for example, using Internet service providers) Business to connect via the Internet).
  • LAN local area network
  • WAN wide area network
  • an external computing device for example, using Internet service providers

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • Data Mining & Analysis (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Artificial Intelligence (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Machine Translation (AREA)
  • Image Analysis (AREA)

Abstract

一种用户特征群中用户特征寻优方法、装置、电子设备及计算机非易失性可读存储介质,属于机器学习应用技术领域,该方法包括:初始化生成多个用户特征组合;将每个用户特征组合及寻优目标输入第一机器学习模型,得到评价分数;获取多个第一分数范围的用户特征组合,及多个第二分数范围的用户特征组合;获取每个第一分数范围的用户特征组合与第二分数范围的用户特征组合的用户特征区别;将多个用户特征区别及寻优目标输入第二机器学习模型,得到预测优化用户特征组合;获取目标优化用户特征组合。上述方法通过机器学习模型根据部分用户特征组合的区别特征预测寻优目标对应的用户特征组合,进而保证目标用户特征组合获取的高效性和准确性。

Description

用户特征群中用户特征寻优方法、装置、电子设备及计算机非易失性可读存储介质 技术领域
本申请要求2019年06月19日递交、发明名称为“用户特征群中用户特征寻优方法、装置、介质及电子设备”的中国专利申请201910533729.8的优先权,在此通过引用将其全部内容合并于此。
本申请涉及机器学习应用技术领域,尤其涉及一种用户特征群中用户特征寻优方法、装置、电子设备及计算机非易失性可读存储介质。
背景技术
用户特征群中用户特征寻优就是从拥有大量用户特征的用户特征群中寻找满足需求的特殊用途的用户特征组合,即从用户特征群中寻找优化用户特征。例如,从用户特征群中寻找满足需求的用户特征组合,以求得到可以满足预定目的的一群用户组合。
技术问题
目前,进行用户特征群中用户特征寻优的主要是通过从用户特征群中爬取关键字,基于不同的关键字设置对应的权重系数,然后通过权重系数计算出满足需求的用户特征组合。这样,本申请的发明人意识到,在用户特征寻优过程中,不能够考虑到所有用户特征群中用户特征之间的区别,获取到的用户特征组合存在组合不合理的问题,且用户特征组合获取的过程需要进行大量数据的统计,算法负荷大,获取效率低。
技术解决方案
为了解决上述技术问题,本申请的一个目的在于提供一种用户特征群中用户特征寻优方法、装置、电子设备及计算机非易失性可读存储介质。
其中,本申请所采用的技术方案为:
一方面,一种用户特征群中用户特征寻优方法,包括:基于目标用户特征群,初始化生成多个用户特征组合;将每个所述用户特征组合及寻优目标分别输入预先训练好的第一机器学习模型,得到每个所述用户特征组合的评价分数;获取多个第一分数范围中的评价分数对应的用户特征组合,及多个第二分数范围中的评价分数对应的用户特征组合,其中,所述多个第一分数范围中的评价分数大于第一预定阈值,所述多个第二分数范围中的评价分数低于第二预定阈值;获取每个所述第一分数范围的用户特征组合与每个第二分数范围的用户特征组合的用户特征区别,得到多个用户特征区别;将所述多个用户特征区别及寻优目标输入预先训练好的第二机器学习模型,得到所述目标用户特征群的预测优化用户特征组合;基于所述预测优化用户特征组合,从所述目标用户特征群中获取目标优化用户特征组合。
另一方面,一种用户特征群中用户特征寻优装置,其特征在于,包括:初始化模块,用于基于目标用户特征群,初始化生成多个用户特征组合;评价模块,用于将每个所述用户特征组合及寻优目标分别输入预先训练好的第一机器学习模型,得到每个所述用户特征组合的评价分数;第一获取模块,用于获取多个第一分数范围中的评价分数对应的用户特征组合,及多个第二分数范围中的评价分数对应的用户特征组合,其中,所述多个第一分数范围中的评价分数大于第一预定阈值,所述多个第二分数范围中的评价分数低于第二预定阈值;第二获取模块,用于获取每个所述第一分数范围的用户特征组合与每个第二分数范围的用户特征组合的用户特征区别,得到多个用户特征区别;预测模块,用于将所述多个用户特征区别及寻优目标输入预先训练好的第二机器学习模型,得到所述目标用户特征群的预测优化用户特征组合;优化特征获取模块,用于基于所述预测优化用户特征组合,从所述目标用户特征群中获取目标优化用户特征组合。
另一方面,一种用户特征群中用户特征寻优装置,包括:处理器;以及
存储器,用于存储所述处理器的用户特征群中用户特征寻优程序;其中,所述处理器配置为经由执行所述用户特征群中用户特征寻优程序来执行如上所述的用户特征群中用户特征寻优方法。
另一方面,一种计算机非易失性可读存储介质,其上存储有用户特征群中用户特征寻优程序,其特征在于,所述用户特征群中用户特征寻优程序被处理器执行时实现如上所述的用户特征群中用户特征寻优方法。
有益效果
在上述技术方案中,通过机器学习模型根据部分用户特征组合的区别特征准确、高效预测寻优目标对应的用户特征组合,进而保证目标用户特征组合获取的高效性和准确性。
应当理解的是,以上的一般描述和后文的细节描述仅是示例性和解释性的,并不能限制本申请。
附图说明
此处的附图被并入说明书中并构成本说明书的一部分,示出了符合本申请的实施例,并于说明书一起用于解释本申请的原理。
图1示意性示出一种用户特征群中用户特征寻优方法的流程图。
图2示意性示出一种用户特征群中用户特征寻优方法的应用场景示例图。
图3示意性示出一种获取用户特征区别的方法流程图。
图4示意性示出一种用户特征群中用户特征寻优装置的方框图。
图5示出根据示例性实施例的用于实现上述用户特征群中用户特征寻优方法的电子设备的框图。
图6示出根据示例性实施例的用于实现上述用户特征群中用户特征寻优方法的计算机非易失性可读存储介质的示意图。
通过上述附图,已示出本申请明确的实施例,后文中将有更详细的描述,这些附图和文字描述并不是为了通过任何方式限制本申请构思的范围,而是通过参考特定实施例为本领域技术人员说明本申请的概念。
本发明的实施方式
这里将详细地对示例性实施例执行说明,其示例表示在附图中。下面的描述涉及附图时,除非另有表示,不同附图中的相同数字表示相同或相似的要素。以下示例性实施例中所描述的实施方式并不代表与本申请相一致的所有实施方式。相反,它们仅是与如所附权利要求书中所详述的、本申请的一些方面相一致的装置和方法的例子。
现在将参考附图更全面地描述示例实施方式。然而,示例实施方式能够以多种形式实施,且不应被理解为限于在此阐述的范例;相反,提供这些实施方式使得本申请将更加全面和完整,并将示例实施方式的构思全面地传达给本领域的技术人员。所描述的特征、结构或特性可以以任何合适的方式结合在一个或更多实施方式中。
本示例实施方式中首先提供了用户特征群中用户特征寻优方法,该用户特征群中用户特征寻优方法可以运行于服务器,也可以运行于服务器集群或云服务器等,当然,本领域技术人员也可以根据需求在其他平台运行本发明的方法,本示例性实施例中对此不做特殊限定。参考图1所示,该用户特征群中用户特征寻优方法可以包括以下步骤:
步骤S110,基于目标用户特征群,初始化生成多个用户特征组合;
步骤S120,将每个所述用户特征组合及寻优目标分别输入预先训练好的第一机器学习模型,得到每个所述用户特征组合的评价分数;
步骤S130,获取多个第一分数范围中的评价分数对应的用户特征组合,及多个第二分数范围中的评价分数对应的用户特征组合,其中,所述多个第一分数范围中的评价分数大于第一预定阈值,所述多个第二分数范围中的评价分数低于第二预定阈值;
步骤S140,获取每个所述第一分数范围的用户特征组合与每个第二分数范围的用户特征组合的用户特征区别,得到多个用户特征区别;
步骤S150,将所述多个用户特征区别及寻优目标输入预先训练好的第二机器学习模型,得到所述目标用户特征群的预测优化用户特征组合;
步骤S160,基于所述预测优化用户特征组合,从所述目标用户特征群中获取目标优化用户特征组合。
上述用户特征群中用户特征寻优方法中,首先,基于目标用户特征群,初始化生成多个用户特征组合;这样可以快速的获取少数几个用户特征组合,进而,将每个所述用户特征组合及寻优目标分别输入预先训练好的第一机器学习模型,得到每个所述用户特征组合的评价分数;可以准确地对每个用户特征组合与寻优目标的适配性进行准确评估。然后,获取多个第一分数范围中的评价分数对应的用户特征组合,及多个第二分数范围中的评价分数对应的用户特征组合,其中,所述多个第一分数范围中的评价分数大于第一预定阈值,所述多个第二分数范围中的评价分数低于第二预定阈值;这样可以根据评价分数准确地将多个用户特征组合分类,进而,通过获取每个所述第一分数范围的用户特征组合与每个第二分数范围的用户特征组合的用户特征区别,得到多个用户特征区别;可以根据不同分数类别的用户特征组合的对比准确得到系列范围的用户特征区别,有效保证区别特征在后续步骤中进行优化用户特征组合分析的可靠性。然后,将所述多个用户特征区别及寻优目标输入预先训练好的第二机器学习模型,得到所述目标用户特征群的预测优化用户特征组合;这样可以高效准确地基于系列范围的用户特征区别预测到优化用户特征组合。进而,基于所述预测优化用户特征组合,可以高效、准确地从所述目标用户特征群中获取目标优化用户特征组合。以这种方式,通过机器学习模型根据部分用户特征组合的区别特征准确、高效预测寻优目标对应的用户特征组合,进而保证目标用户特征组合获取的高效性和准确性。
下面,将结合附图对本示例实施方式中上述用户特征群中用户特征寻优方法中的各步骤进行详细的解释以及说明。
在步骤S110中,基于目标用户特征群,初始化生成多个用户特征组合。
本示例的实施方式中,参考图2所示,服务器201从平台服务器202上获取目标用户特征群,然后基于目标用户特征群,初始化生成多个用户特征组合。这样可以由服务器201在后续步骤中根据初始化生成的多个用户特征组合进行用户特征组合的组合特征分析,进而可以获取到优化目标用户组合特征。可以理解,根据需求,也可以直接由服务器202基于目标用户特征群,初始化生成多个用户特征组合。在后续步骤中根据初始化生成的多个用户特征组合进行用户特征组合特征分析,进而获取到优化目标用户特征。其中,服务器201和服务器202可以是任何具有处理能力的设备,例如,电脑、微处理器、云服务器等,在此不做特殊限定。
目标用户特征群就是包含有大量用户特征的用户特征群,例如,某个如购物平台或者交流平台上大量的用户特征。基于目标用户特征群,初始化生成多个用户特征组合,就是从目标用户特征群中,进行获取多个用户特征后随机或者按照预定顺序组合成用户特征组合,进而多个获取得到多个用户特征组合,这样可以快速的获取所有可能的用户特征组合中少数几个用户特征组合,用于进行后续步骤中进行评价。其中,获取的用户特征组合的数目可以是将目标用户特征群中按照用户特征个数进行排列组合,获取预定比例的用户特征组合,预定比例根据需求设定,可以是例如0.001%、0.002%等。
本示例的一种实施方式中,所述基于目标用户特征群,初始化生成多个用户特征组合,包括:
从所述目标用户特征群中,依次分别获取预定数目个用户特征,其中,每次获取时的所述预定数目比前一次多预定个数个;依次分别将获取的所述预定数目个用户特征,按照所述目标用户特征群的排列规则排列成用户特征串,得到多个用户特征组合。
从所述目标用户特征群中,依次分别获取预定数目个用户特征,其中,每次获取时的所述预定数目比前一次多预定个数个,就是例如从拥有10000个用户特征的用户特征群中依次获取100个、110个......、200个用户特征,每次获取时用户特征个数比前一次多10个。然后,依次分别将获取的所述预定数目个用户特征,按照所述目标用户特征群的排列规则排列成用户特征串,得到多个用户特征组合。这样可以在目标用户特征群拥有自己地用户特征排列规则时,保证获取到的用户特征组合携带有用户特征群特征,保证后续步骤分析的准确性。其中,排列规则可以是例如按照预定顺序排列或者按照预定表格形式排列等。
本示例的一种实施方式中,所述基于目标用户特征群,初始化生成多个用户特征组合,包括:
基于目标用户特征群,随机初始化生成预定数目个用户特征组合。
这样通过随机初始化生成预定数目个用户特征组合,可以保证用户特征组合获取的随机性,在用户特征群没有排列顺序时,保证用户特征组合获取的效率。
在步骤S120中, 将每个所述用户特征组合及寻优目标分别输入预先训练好的第一机器学习模型,得到每个所述用户特征组合的评价分数。
本示例的实施方式中,将每个用户特征组合及寻优目标分别输入预先训练好的机器学习模型,可以由机器学习模型根据寻优目标评价每个用户特征组合与寻优目标的适配性,即每个所述用户特征组合的评价分数。其中,寻优目标就是用户特征组合的用途,例如某些用户特征用户特征组合在是否可以适配用于某个项目的用户组合需求。
本示例的一种实施方式中,所述第一机器学习模型的训练方法:
收集用户特征组合及寻优目标样本集,所述样本集中每个样本事先标定对应的评价分数;将所述样本集中每个样本的输入数据分别输入机器学习模型,得到每个所述样本对应的评价分数;如果存在有样本的输入数据输入机器学习模型后,得到样本对应的评价分数与对所述样本事先标定的评价分数不一致,则调整学习模型的系数,直到一致;当所有的样本的输入数据输入机器学习模型后,得到的每个样本对应的评价分数与对每个所述样本事先标定的评价分数一致,训练结束。
用户特征组合及寻优目标样本就是历史上对应于某个用户特征群的用户特征组合及寻优目标样本。通过收集用户特征组合及寻优目标样本集作为机器学习模型的输入,每个样本事先由专家标定对应的评价分数,作为机器学习模型的输出。然后,通过调整系数使得当所有的样本的输入数据输入机器学习模型后,得到的每个样本对应的评价分数与对每个所述样本事先标定的评价分数一致,训练结束,得到训练好的第一学习模型,可以有效保证训练准确性。
在步骤S130中,获取多个第一分数范围中的评价分数对应的用户特征组合,及多个第二分数范围中的评价分数对应的用户特征组合,其中,所述多个第一分数范围中的评价分数大于第一预定阈值,所述多个第二分数范围中的评价分数低于第二预定阈值。
本示例的实施方式中,第一预定阈值为设置的判断评价分数较高的阈值,第二预定阈值为设置的判断评价分数较低的阈值。大于第一预定阈值的多个第一分数范围就是例如第一预定阈值为60时,范围为60-65,65-70等多个第一分数范围,低于第二预定阈值的多个第二分数范围,就是例如第二预定阈值为50时,范围为50-45,45-40等多个第二分数范围。这样获取所有所述评价分数中,大于第一预定阈值的多个第一分数范围的评价分数对应的用户特征组合,及低于第二预定阈值的多个第二分数范围的评价分数对应的用户特征组合,可以根据评价分数准确地将多个用户特征组合分类,得到与寻优目标对应的不同级别的用户特征组合。其中,一种示例中,根据需要第一预定阈值与第二预定阈值相等;另一个示例性中,第一预定阈值大于第二预定阈值,舍弃第一预定阈值与第二预定阈值之间的分数范围的用户组合特征。
本示例的一种实施方式中,在所述获取多个第一分数范围中的评价分数对应的用户特征组合,及多个第二分数范围中的评价分数对应的用户特征组合,其中,所述多个第一分数范围中的评价分数大于第一预定阈值,所述多个第二分数范围中的评价分数低于第二预定阈值之后,所述方法还包括:
如果有所述第一分数范围及所述第二分数范围没有获取到对应的用户特征组合,基于所述目标用户特征群,初始化生成预定组数个用户特征组合;将所述用户特征组合分别输入预先训练好的第一机器学习模型,得到每个所述用户特征组合的评价分数;获取所有所述评价分数中,大于第一预定阈值的多个第一分数范围的评价分数对应的用户特征组合,及低于第二预定阈值的多个第二分数范围的评价分数对应的用户特征组合当所有所述第一分数范围及所述第二分数范围有获取到对应的用户特征组合,获取结束。
以这种方式,如果有所述第一分数范围及所述第二分数范围没有获取到对应的用户特征组合,例如大于第一预定阈值的多个第一分数范围为例如第一预定阈值为60时,范围为60-65,65-70等多个第一分数范围,其中60-65的范围,没有获取到用户特征组合。通过基于所述目标用户特征群,初始化生成预定组数个用户特征组合,重新进行评分、分类,直到当所有所述第一分数范围及所述第二分数范围有获取到对应的用户特征组合,获取结束。可以保证每个第一分数范围及第二分数范围有对应的用户特征组合,保证后续步骤进行分析的可靠性。
在步骤S140中, 获取每个所述第一分数范围的用户特征组合与每个第二分数范围的用户特征组合的用户特征区别,得到多个用户特征区别。
本示例的实施方式中,每个第一分数范围对应一个所述第二分数范围,就是对每个第一分数范围预先关联的一个第二分数范围。例如,大于第一预定阈值的多个第一分数范围就是例如第一预定阈值为60时,范围为60-65,65-70等多个第一分数范围,低于第二预定阈值的多个第二分数范围,就是例如第二预定阈值为50时,范围为50-45,45-40等多个第二分数范围,此时,60-65的第一分数范围对应50-45的第二分数范围,65-70的第一分数范围对应45-40的第二分数范围。这样获取每个所述第一分数范围的用户特征组合与每个第一分数范围对应的所述第二分数范围的用户特征组合的用户特征区别,可以得到系列范围的多个用户特征区别。其中,用户特征区别可以是由评分高的用户特征组合的用户特征特征的映射向量或者转化得到的字符串与评分低的用户特征组合的用户特征特征的映射向量或者转化得到字符串的差。其中用户特征特征的映射向量通过用户特征特征进行向量映射得到,例如,一个高评分用户特征特征为50男120女,映射为(1,50,21,120,22),其中1为性别维度向量,50、120为个数维度,21、22为性别属性值维度,此时如果一个低评分用户特征特征向量为(1,10,21,160,22),则他们的差表示为(1,50,21,120,22)-(1,10,21,160,22)。这样使得区别技术特征基于每个用户特征组合的评价系列化存在,有效保证后续步骤基于用户特征区别预测优化用户特征组合准确性。
本示例的一种实施方式中,所述获取每个所述第一分数范围的用户特征组合与每个第二分数范围的用户特征组合的用户特征区别,得到多个用户特征区别,包括:从每个所述第一分数范围的用户特征组合中随机获取一个用户特征组合;从每个所述第二分数范围的用户特征组合中随机获取一个用户特征组合;
获取每个所述第一分数范围的所述随机获取的一个用户特征组合与每个所述第二分数范围的所述随机获取的一个用户特征组合的用户特征区别,得到多个用户特征区别。
每个第一分数范围及每个第二分数范围中可能具有多个用户特征组合。但每个分数范围中一个用户特征组合就可以代表该范围的用户特征组合特征。进而,通过从每个所述第一分数范围的用户特征组合中随机获取一个用户特征组合;从每个所述第二分数范围的用户特征组合中随机获取一个用户特征组合;可以高效的获取每个所述第一分数范围的一个用户特征组合与每个第一分数范围对应的所述第二分数范围的一个用户特征组合的用户特征区别,得到多个用户特征区别。
本示例的一种实施方式中,参考图3所示,所述获取每个所述第一分数范围的用户特征组合与每个第二分数范围的用户特征组合的用户特征区别,得到多个用户特征区别,包括:步骤S310,获取每个所述第一分数范围的用户特征组合的第一用户特征组合要素;
步骤S320,获取每个第一分数范围对应的所述第二分数范围的用户特征组合的第二用户特征组合要素;步骤S330,获取每个所述第一用户特征组合要素与每个所述第二用户特征组合要素的区别特征,得到每个所述第一分数范围的用户特征组合与每个第二分数范围的用户特征组合的用户特征区别。
用户特征组合要素至少包括用户特征组合中用户特征的数量、用户特征的预设特征标签、用户特征多个属性的属性值等,用户特征的预设特征标签就是例如某个用户的身份,用户特征多个属性的属性值就是例如用户的性别的男女,年龄的大小等。这样通过获取每个第一分数范围的用户特征组合的第一用户特征组合要素,及获取每个第一分数范围对应的第二分数范围的用户特征组合的第二用户特征组合要素,然后,获取每个第一用户特征组合要素及与第一用户特征组合要素对应的第二用户特征组合要素的区别特征,可以准确地获取到每个用户特征组合的区别特征。有效保证后续步骤中分析的准确性。
在步骤S150中, 将所述多个用户特征区别及寻优目标输入预先训练好的第二机器学习模型,得到所述目标用户特征群的预测优化用户特征组合。
本示例的实施方式中,多个用户特征区别为系列范围的用户特征区别,可以准确反映用户特征组合评分变化与用户特征组合特征变化,进而,将多个用户特征区别及寻优目标输入预先训练好的第二机器学习模型,可以由寻优目标高效准确地基于用户特征区别分析得到目标用户特征群的与寻优目标适应的预测优化用户特征组合,保证后续步骤中从用户特征群中获取目标用户特征组合的效率和准确性。
本示例的一种实施方式中,所述第二机器学习模型的训练方法:收集用户特征区别及寻优目标样本集,所述样本集中每个样本事先标定对应的优化用户特征组合;将所述样本集中每个样本的输入数据分别输入机器学习模型,得到每个所述样本对应的优化用户特征组合;如果存在有样本的输入数据输入机器学习模型后,得到样本对应的优化用户特征组合与对所述样本事先标定的优化用户特征组合不一致,则调整学习模型的系数,直到一致;当所有的样本的输入数据输入机器学习模型后,得到的每个样本对应的优化用户特征组合与对每个所述样本事先标定的优化用户特征组合一致,训练结束。
用户特征区别及寻优目标样本就是历史上对应于某个用户特征群的用户特征区别及寻优目标样本。通过收集用户特征区别及寻优目标样本集作为机器学习模型的输入,每个样本事先由专家标定对应的优化用户特征组合,作为机器学习模型的输出。然后,通过调整系数使得当所有的样本的输入数据输入机器学习模型后,得到的每个样本对应的优化用户特征组合与对每个所述样本事先标定的优化用户特征组合一致,训练结束,得到训练好的第二学习模型,可以有效保证训练准确性。
在步骤S160中, 基于所述预测优化用户特征组合,从所述目标用户特征群中获取目标优化用户特征组合。
本示例的实施方式中,基于预测优化用户特征组合,可以从目标用户特征群中通过用户特征特征标签或者用户特征属性名等逐个对比,准确地从目标用户特征群中获取目标优化用户特征组合。
最后,本申请通过机器学习模型根据部分用户特征组合的区别特征准确、高效预测寻优目标对应的用户特征组合,进而保证目标用户特征组合获取的高效性和准确性。
本申请还提供了一种用户特征群中用户特征寻优装置。参考图4所示,该用户特征群中用户特征寻优装置可以包括用户特征初始化模块410、评价模块420、第一获取模块430、第二获取模块440、预测模块450及优化特征获取模块460。其中:初始化模块410可以用于基于目标用户特征群,初始化生成多个用户特征组合;评价模块420可以用于将每个所述用户特征组合及寻优目标分别输入预先训练好的第一机器学习模型,得到每个所述用户特征组合的评价分数;第一获取模块430可以用于获取多个第一分数范围中的评价分数对应的用户特征组合,及多个第二分数范围中的评价分数对应的用户特征组合,其中,所述多个第一分数范围中的评价分数大于第一预定阈值,所述多个第二分数范围中的评价分数低于第二预定阈值;第二获取模块440可以用于获取每个所述第一分数范围的用户特征组合与每个第二分数范围的用户特征组合的用户特征区别,得到多个用户特征区别;预测模块450可以用于将所述多个用户特征区别及寻优目标输入预先训练好的第二机器学习模型,得到所述目标用户特征群的预测优化用户特征组合;优化特征获取模块460可以用于基于所述预测优化用户特征组合,从目标用户特征群中获取目标优化用户特征组合。
上述用户特征群中用户特征寻优装置中各模块的具体细节已经在对应的用户特征群中用户特征寻优方法中进行了详细的描述,因此此处不再赘述。
应当注意,尽管在上文详细描述中提及了用于动作执行的设备的若干模块或者单元,但是这种划分并非强制性的。实际上,根据本申请的实施方式,上文描述的两个或更多模块或者单元的特征和功能可以在一个模块或者单元中具体化。反之,上文描述的一个模块或者单元的特征和功能可以进一步划分为由多个模块或者单元来具体化。
此外,尽管在附图中以特定顺序描述了本申请中方法的各个步骤,但是,这并非要求或者暗示必须按照该特定顺序来执行这些步骤,或是必须执行全部所示的步骤才能实现期望的结果。附加的或备选的,可以省略某些步骤,将多个步骤合并为一个步骤执行,以及/或者将一个步骤分解为多个步骤执行等。通过以上的实施方式的描述,本领域的技术人员易于理解,这里描述的示例实施方式可以通过软件实现,也可以通过软件结合必要的硬件的方式来实现。因此,根据本申请实施方式的技术方案可以以软件产品的形式体现出来,该软件产品可以存储在一个非易失性存储介质(可以是CD-ROM,U盘,移动硬盘等)中或网络上,包括若干指令以使得一台计算设备(可以是个人计算机、服务器、移动终端、或者网络设备等)执行根据本申请实施方式的方法。
在本申请的示例性实施例中,还提供了一种能够实现上述方法的电子设备。
所属技术领域的技术人员能够理解,本发明的各个方面可以实现为系统、方法或程序产品。因此,本发明的各个方面可以具体实现为以下形式,即:完全的硬件实施方式、完全的软件实施方式(包括固件、微代码等),或硬件和软件方面结合的实施方式,这里可以统称为“电路”、“模块”或“系统”。
下面参照图5来描述根据本发明的这种实施方式的电子设备500。图5显示的电子设备500仅仅是一个示例,不应对本发明实施例的功能和使用范围带来任何限制。
如图5所示,电子设备500以通用计算设备的形式表现。电子设备500的组件可以包括但不限于:上述至少一个处理单元510、上述至少一个存储单元520、连接不同系统组件(包括存储单元520和处理单元510)的总线530。
其中,所述存储单元存储有程序代码,所述程序代码可以被所述处理单元510执行,使得所述处理单元510执行本说明书上述“示例性方法”部分中描述的根据本发明各种示例性实施方式的步骤。例如,所述处理单元510可以执行如图1中所示的步骤S110:基于目标用户特征群,初始化生成多个用户特征组合;S120:将每个所述用户特征组合及寻优目标分别输入预先训练好的第一机器学习模型,得到每个所述用户特征组合的评价分数;步骤S130:获取多个第一分数范围中的评价分数对应的用户特征组合,及多个第二分数范围中的评价分数对应的用户特征组合,其中,所述多个第一分数范围中的评价分数大于第一预定阈值,所述多个第二分数范围中的评价分数低于第二预定阈值;步骤S140:获取每个所述第一分数范围的用户特征组合与每个第二分数范围的用户特征组合的用户特征区别,得到多个用户特征区别;步骤S150:将所述多个用户特征区别及寻优目标输入预先训练好的第二机器学习模型,得到所述目标用户特征群的预测优化用户特征组合;步骤S160:基于所述预测优化用户特征组合,从所述目标用户特征群中获取目标优化用户特征组合。存储单元520可以包括易失性存储单元形式的可读介质,例如随机存取存储单元(RAM)5201和/或高速缓存存储单元5202,还可以进一步包括只读存储单元(ROM)5203。存储单元520还可以包括具有一组(至少一个)程序模块5205的程序/实用工具5204,这样的程序模块5205包括但不限于:操作系统、一个或者多个应用程序、其它程序模块以及程序数据,这些示例中的每一个或某种组合中可能包括网络环境的实现。总线530可以为表示几类总线结构中的一种或多种,包括存储单元总线或者存储单元控制器、外围总线、图形加速端口、处理单元或者使用多种总线结构中的任意总线结构的局域总线。
电子设备500也可以与一个或多个外部设备700(例如键盘、指向设备、蓝牙设备等)通信,还可与一个或者多个使得客户能与该电子设备500交互的设备通信,和/或与使得该电子设备500能与一个或多个其它计算设备进行通信的任何设备(例如路由器、调制解调器等等)通信。这种通信可以通过输入/输出(I/O)接口550进行。并且,电子设备500还可以通过网络适配器560与一个或者多个网络(例如局域网(LAN),广域网(WAN)和/或公共网络,例如因特网)通信。如图所示,网络适配器560通过总线530与电子设备500的其它模块通信。应当明白,尽管图中未示出,可以结合电子设备500使用其它硬件和/或软件模块,包括但不限于:微代码、设备驱动器、冗余处理单元、外部磁盘驱动阵列、RAID系统、磁带驱动器以及数据备份存储系统等。通过以上的实施方式的描述,本领域的技术人员易于理解,这里描述的示例实施方式可以通过软件实现,也可以通过软件结合必要的硬件的方式来实现。因此,根据本申请实施方式的技术方案可以以软件产品的形式体现出来,该软件产品可以存储在一个非易失性存储介质(可以是CD-ROM,U盘,移动硬盘等)中或网络上,包括若干指令以使得一台计算设备(可以是个人计算机、服务器、终端装置、或者网络设备等)执行根据本申请实施方式的方法。
在本申请的示例性实施例中,还提供了一种计算机可读存储介质,其上存储有能够实现本说明书上述方法的程序产品。在一些可能的实施方式中,本发明的各个方面还可以实现为一种程序产品的形式,其包括程序代码,当所述程序产品在终端设备上运行时,所述程序代码用于使所述终端设备执行本说明书上述“示例性方法”部分中描述的根据本发明各种示例性实施方式的步骤。参考图6所示,描述了根据本发明的实施方式的用于实现上述方法的程序产品600,其可以采用便携式紧凑盘只读存储器(CD-ROM)并包括程序代码,并可以在终端设备,例如个人电脑上运行。然而,本发明的程序产品不限于此,在本文件中,可读存储介质可以是任何包含或存储程序的有形介质,该程序可以被指令执行系统、装置或者器件使用或者与其结合使用。所述程序产品可以采用一个或多个可读介质的任意组合。可读介质可以是可读信号介质或者可读存储介质。可读存储介质例如可以为但不限于电、磁、光、电磁、红外线、或半导体的系统、装置或器件,或者任意以上的组合。可读存储介质的更具体的例子(非穷举的列表)包括:具有一个或多个导线的电连接、便携式盘、硬盘、随机存取存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、光纤、便携式紧凑盘只读存储器(CD-ROM)、光存储器件、磁存储器件、或者上述的任意合适的组合。计算机可读信号介质可以包括在基带中或者作为载波一部分传播的数据信号,其中承载了可读程序代码。这种传播的数据信号可以采用多种形式,包括但不限于电磁信号、光信号或上述的任意合适的组合。可读信号介质还可以是可读存储介质以外的任何可读介质,该可读介质可以发送、传播或者传输用于由指令执行系统、装置或者器件使用或者与其结合使用的程序。可读介质上包含的程序代码可以用任何适当的介质传输,包括但不限于无线、有线、光缆、RF等等,或者上述的任意合适的组合。可以以一种或多种程序设计语言的任意组合来编写用于执行本发明操作的程序代码,所述程序设计语言包括面向对象的程序设计语言—诸如Java、C++等,还包括常规的过程式程序设计语言—诸如“C”语言或类似的程序设计语言。程序代码可以完全地在客户计算设备上执行、部分地在客户设备上执行、作为一个独立的软件包执行、部分在客户计算设备上部分在远程计算设备上执行、或者完全在远程计算设备或服务器上执行。在涉及远程计算设备的情形中,远程计算设备可以通过任意种类的网络,包括局域网(LAN)或广域网(WAN),连接到客户计算设备,或者,可以连接到外部计算设备(例如利用因特网服务提供商来通过因特网连接)。
此外,上述附图仅是根据本发明示例性实施例的方法所包括的处理的示意性说明,而不是限制目的。易于理解,上述附图所示的处理并不表明或限制这些处理的时间顺序。另外,也易于理解,这些处理可以是例如在多个模块中同步或异步执行的。本领域技术人员在考虑说明书及实践这里公开的发明后,将容易想到本申请的其他实施例。本申请旨在涵盖本申请的任何变型、用途或者适应性变化,这些变型、用途或者适应性变化遵循本申请的一般性原理并包括本申请未公开的本技术领域中的公知常识或惯用技术手段。说明书和实施例仅被视为示例性的,本申请的真正范围和精神由权利要求指出。

Claims (20)

  1. 一种用户特征群中用户特征寻优方法,其特征在于,包括:
    基于目标用户特征群,初始化生成多个用户特征组合;
    将每个所述用户特征组合及寻优目标分别输入预先训练好的第一机器学习模型,得到每个所述用户特征组合的评价分数;
    获取多个第一分数范围中的评价分数对应的用户特征组合,及多个第二分数范围中的评价分数对应的用户特征组合,其中,所述多个第一分数范围中的评价分数大于第一预定阈值,所述多个第二分数范围中的评价分数低于第二预定阈值;
    获取每个所述第一分数范围的用户特征组合与每个第二分数范围的用户特征组合的用户特征区别,得到多个用户特征区别;
    将所述多个用户特征区别及寻优目标输入预先训练好的第二机器学习模型,得到所述目标用户特征群的预测优化用户特征组合;
    基于所述预测优化用户特征组合,从所述目标用户特征群中获取目标优化用户特征组合。
  2. 根据权利要求1所述的方法,其特征在于,所述基于目标用户特征群,初始化生成多个用户特征组合,包括:
    从所述目标用户特征群中,依次分别获取预定数目个用户特征,其中,每次获取时的所述预定数目比前一次多预定个数个;
    依次分别将获取的所述预定数目个用户特征,按照所述目标用户特征群的排列规则排列成用户特征串,得到多个用户特征组合。
  3. 根据权利要求1所述的方法,其特征在于,所述基于目标用户特征群,初始化生成多个用户特征组合,包括:
    基于目标用户特征群,随机初始化生成预定数目个用户特征组合。
  4. 根据权利要求1所述的方法,其特征在于,在获取多个第一分数范围中的评价分数对应的用户特征组合,及多个第二分数范围中的评价分数对应的用户特征组合,其中,所述多个第一分数范围中的评价分数大于第一预定阈值,所述多个第二分数范围中的评价分数低于第二预定阈值之后,所述方法还包括:
    如果有所述第一分数范围及所述第二分数范围没有获取到对应的用户特征组合,基于所述目标用户特征群,初始化生成预定组数个用户特征组合;
    将所述用户特征组合分别输入预先训练好的第一机器学习模型,得到每个所述用户特征组合的评价分数;
    获取所有所述评价分数中,大于第一预定阈值的多个第一分数范围的评价分数对应的用户特征组合,及低于第二预定阈值的多个第二分数范围的评价分数对应的用户特征组合;
    当所有所述第一分数范围及所述第二分数范围有获取到对应的用户特征组合,获取结束。
  5. 根据权利要求1所述的方法,其特征在于,所述获取每个所述第一分数范围的用户特征组合与每个第二分数范围的用户特征组合的用户特征区别,得到多个用户特征区别,包括:
    从每个所述第一分数范围的用户特征组合中随机获取一个用户特征组合;
    从每个所述第二分数范围的用户特征组合中随机获取一个用户特征组合;
    获取每个所述第一分数范围的所述随机获取的一个用户特征组合与每个所述第二分数范围的所述随机获取的一个用户特征组合的用户特征区别,得到多个用户特征区别。
  6. 根据权利要求1所述的方法,其特征在于,所述获取每个所述第一分数范围的用户特征组合与每个第二分数范围的用户特征组合的用户特征区别,得到多个用户特征区别,包括:
    获取每个所述第一分数范围的用户特征组合的第一用户特征组合要素;
    获取每个所述第二分数范围的用户特征组合的第二用户特征组合要素;
    获取每个所述第一用户特征组合要素与每个所述第二用户特征组合要素的区别特征,得到每个所述第一分数范围的用户特征组合与每个第二分数范围的用户特征组合的用户特征区别。
  7. 根据权利要求1所述的方法,其特征在于,所述第一机器学习模型的训练方法:
    收集用户特征组合及寻优目标样本集,所述样本集中每个样本事先标定对应的评价分数;
    将所述样本集中每个样本的输入数据分别输入机器学习模型,得到每个所述样本对应的评价分数;
    如果存在有样本的输入数据输入机器学习模型后,得到样本对应的评价分数与对所述样本事先标定的评价分数不一致,则调整学习模型的系数,直到一致;
    当所有的样本的输入数据输入机器学习模型后,得到的每个样本对应的评价分数与对每个所述样本事先标定的评价分数一致,训练结束。
  8. 一种用户特征群中用户特征寻优装置,其特征在于,包括:
    初始化模块,用于基于目标用户特征群,初始化生成多个用户特征组合;
    评价模块,用于将每个所述用户特征组合及寻优目标分别输入预先训练好的第一机器学习模型,得到每个所述用户特征组合的评价分数;
    第一获取模块,用于获取多个第一分数范围中的评价分数对应的用户特征组合,及多个第二分数范围中的评价分数对应的用户特征组合,其中,所述多个第一分数范围中的评价分数大于第一预定阈值,所述多个第二分数范围中的评价分数低于第二预定阈值;
    第二获取模块,用于获取每个所述第一分数范围的用户特征组合与每个第二分数范围的用户特征组合的用户特征区别,得到多个用户特征区别;
    预测模块,用于将所述多个用户特征区别及寻优目标输入预先训练好的第二机器学习模型,得到所述目标用户特征群的预测优化用户特征组合;
    优化特征获取模块,用于基于所述预测优化用户特征组合,从所述目标用户特征群中获取目标优化用户特征组合。
  9. 根据权利要求8所述的装置,所述初始化模块被配置为:
    从所述目标用户特征群中,依次分别获取预定数目个用户特征,其中,每次获取时的所述预定数目比前一次多预定个数个;
    依次分别将获取的所述预定数目个用户特征,按照所述目标用户特征群的排列规则排列成用户特征串,得到多个用户特征组合。
  10. 根据权利要求8所述的装置,所述第二获取模块被配置为:
    如果有所述第一分数范围及所述第二分数范围没有获取到对应的用户特征组合,基于所述目标用户特征群,初始化生成预定组数个用户特征组合;
    将所述用户特征组合分别输入预先训练好的第一机器学习模型,得到每个所述用户特征组合的评价分数;
    获取所有所述评价分数中,大于第一预定阈值的多个第一分数范围的评价分数对应的用户特征组合,及低于第二预定阈值的多个第二分数范围的评价分数对应的用户特征组合;
    当所有所述第一分数范围及所述第二分数范围有获取到对应的用户特征组合,获取结束。
  11. 根据权利要求8所述的装置,所述第二获取模块被配置为:
    从每个所述第一分数范围的用户特征组合中随机获取一个用户特征组合;
    从每个所述第二分数范围的用户特征组合中随机获取一个用户特征组合;
    获取每个所述第一分数范围的所述随机获取的一个用户特征组合与每个所述第二分数范围的所述随机获取的一个用户特征组合的用户特征区别,得到多个用户特征区别。
  12. 根据权利要求8所述的装置,所述第二获取模块被配置为:
    获取每个所述第一分数范围的用户特征组合的第一用户特征组合要素;
    获取每个所述第二分数范围的用户特征组合的第二用户特征组合要素;
    获取每个所述第一用户特征组合要素与每个所述第二用户特征组合要素的区别特征,得到每个所述第一分数范围的用户特征组合与每个第二分数范围的用户特征组合的用户特征区别。
  13. 根据权利要求8所述的装置,所述评价模块被配置为:
    收集用户特征组合及寻优目标样本集,所述样本集中每个样本事先标定对应的评价分数;
    将所述样本集中每个样本的输入数据分别输入机器学习模型,得到每个所述样本对应的评价分数;
    如果存在有样本的输入数据输入机器学习模型后,得到样本对应的评价分数与对所述样本事先标定的评价分数不一致,则调整学习模型的系数,直到一致;
    当所有的样本的输入数据输入机器学习模型后,得到的每个样本对应的评价分数与对每个所述样本事先标定的评价分数一致,训练结束。
  14. 一种电子设备,其特征在于,包括:处理器;以及存储器,用于存储所述处理器的用户特征群中用户特征寻优程序;其中,所述处理器配置为经由执行所述用户特征群中用户特征寻优程序来执行以下处理:
    基于目标用户特征群,初始化生成多个用户特征组合;
    将每个所述用户特征组合及寻优目标分别输入预先训练好的第一机器学习模型,得到每个所述用户特征组合的评价分数;
    获取多个第一分数范围中的评价分数对应的用户特征组合,及多个第二分数范围中的评价分数对应的用户特征组合,其中,所述多个第一分数范围中的评价分数大于第一预定阈值,所述多个第二分数范围中的评价分数低于第二预定阈值;
    获取每个所述第一分数范围的用户特征组合与每个第二分数范围的用户特征组合的用户特征区别,得到多个用户特征区别;
    将所述多个用户特征区别及寻优目标输入预先训练好的第二机器学习模型,得到所述目标用户特征群的预测优化用户特征组合;
    基于所述预测优化用户特征组合,从所述目标用户特征群中获取目标优化用户特征组合。
  15. 根据权利要求14所述的电子设备,其特征在于,所述基于目标用户特征群,初始化生成多个用户特征组合,包括:
    从所述目标用户特征群中,依次分别获取预定数目个用户特征,其中,每次获取时的所述预定数目比前一次多预定个数个;
    依次分别将获取的所述预定数目个用户特征,按照所述目标用户特征群的排列规则排列成用户特征串,得到多个用户特征组合。
  16. 根据权利要求14所述的电子设备,其特征在于,还包括:
    如果有所述第一分数范围及所述第二分数范围没有获取到对应的用户特征组合,基于所述目标用户特征群,初始化生成预定组数个用户特征组合;
    将所述用户特征组合分别输入预先训练好的第一机器学习模型,得到每个所述用户特征组合的评价分数;
    获取所有所述评价分数中,大于第一预定阈值的多个第一分数范围的评价分数对应的用户特征组合,及低于第二预定阈值的多个第二分数范围的评价分数对应的用户特征组合;
    当所有所述第一分数范围及所述第二分数范围有获取到对应的用户特征组合,获取结束。
  17. 根据权利要求14所述的电子设备,其特征在于,所述获取每个所述第一分数范围的用户特征组合与每个第二分数范围的用户特征组合的用户特征区别,得到多个用户特征区别,包括:
    从每个所述第一分数范围的用户特征组合中随机获取一个用户特征组合;
    从每个所述第二分数范围的用户特征组合中随机获取一个用户特征组合;
    获取每个所述第一分数范围的所述随机获取的一个用户特征组合与每个所述第二分数范围的所述随机获取的一个用户特征组合的用户特征区别,得到多个用户特征区别。
  18. 根据权利要求14所述的电子设备,其特征在于,所述获取每个所述第一分数范围的用户特征组合与每个第二分数范围的用户特征组合的用户特征区别,得到多个用户特征区别,包括:
    获取每个所述第一分数范围的用户特征组合的第一用户特征组合要素;
    获取每个所述第二分数范围的用户特征组合的第二用户特征组合要素;
    获取每个所述第一用户特征组合要素与每个所述第二用户特征组合要素的区别特征,得到每个所述第一分数范围的用户特征组合与每个第二分数范围的用户特征组合的用户特征区别。
  19. 根据权利要求14所述的电子设备,其特征在于,还包括:
    收集用户特征组合及寻优目标样本集,所述样本集中每个样本事先标定对应的评价分数;
    将所述样本集中每个样本的输入数据分别输入机器学习模型,得到每个所述样本对应的评价分数;
    如果存在有样本的输入数据输入机器学习模型后,得到样本对应的评价分数与对所述样本事先标定的评价分数不一致,则调整学习模型的系数,直到一致;
    当所有的样本的输入数据输入机器学习模型后,得到的每个样本对应的评价分数与对每个所述样本事先标定的评价分数一致,训练结束。
  20. 一种计算机非易失性可读存储介质,其上存储有用户特征群中用户特征寻优程序,其特征在于,所述用户特征群中用户特征寻优程序被处理器执行时执行以下处理:
    基于目标用户特征群,初始化生成多个用户特征组合;
    将每个所述用户特征组合及寻优目标分别输入预先训练好的第一机器学习模型,得到每个所述用户特征组合的评价分数;
    获取多个第一分数范围中的评价分数对应的用户特征组合,及多个第二分数范围中的评价分数对应的用户特征组合,其中,所述多个第一分数范围中的评价分数大于第一预定阈值,所述多个第二分数范围中的评价分数低于第二预定阈值;
    获取每个所述第一分数范围的用户特征组合与每个第二分数范围的用户特征组合的用户特征区别,得到多个用户特征区别;
    将所述多个用户特征区别及寻优目标输入预先训练好的第二机器学习模型,得到所述目标用户特征群的预测优化用户特征组合;
    基于所述预测优化用户特征组合,从所述目标用户特征群中获取目标优化用户特征组合。
PCT/CN2019/103437 2019-06-19 2019-08-29 用户特征群中用户特征寻优方法、装置、电子设备及计算机非易失性可读存储介质 WO2020252925A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910533729.8A CN110348581B (zh) 2019-06-19 2019-06-19 用户特征群中用户特征寻优方法、装置、介质及电子设备
CN201910533729.8 2019-06-19

Publications (1)

Publication Number Publication Date
WO2020252925A1 true WO2020252925A1 (zh) 2020-12-24

Family

ID=68182439

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/103437 WO2020252925A1 (zh) 2019-06-19 2019-08-29 用户特征群中用户特征寻优方法、装置、电子设备及计算机非易失性可读存储介质

Country Status (2)

Country Link
CN (1) CN110348581B (zh)
WO (1) WO2020252925A1 (zh)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111737319B (zh) * 2020-06-24 2024-03-01 平安科技(深圳)有限公司 用户集群的预测方法、装置、计算机设备和存储介质
CN112669982B (zh) * 2020-12-31 2023-02-21 南方科技大学 密切接触者的确定方法、装置、设备和存储介质

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104239351A (zh) * 2013-06-20 2014-12-24 阿里巴巴集团控股有限公司 一种用户行为的机器学习模型的训练方法及装置
CN105260414A (zh) * 2015-09-24 2016-01-20 精硕世纪科技(北京)有限公司 用户行为相似性计算方法及装置
US20170262445A1 (en) * 2016-03-08 2017-09-14 Facebook, Inc. Statistical feature engineering of user attributes
CN107480713A (zh) * 2017-08-08 2017-12-15 云南大学 含主导特征的co‑location模式挖掘方法

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8266078B2 (en) * 2009-02-06 2012-09-11 Microsoft Corporation Platform for learning based recognition research
CN106548210B (zh) * 2016-10-31 2021-02-05 腾讯科技(深圳)有限公司 基于机器学习模型训练的信贷用户分类方法及装置
CN108154430A (zh) * 2017-12-28 2018-06-12 上海氪信信息技术有限公司 一种基于机器学习和大数据技术的信用评分构建方法
CN109858974A (zh) * 2019-02-18 2019-06-07 重庆邮电大学 已购车用户识别模型构建方法及识别方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104239351A (zh) * 2013-06-20 2014-12-24 阿里巴巴集团控股有限公司 一种用户行为的机器学习模型的训练方法及装置
CN105260414A (zh) * 2015-09-24 2016-01-20 精硕世纪科技(北京)有限公司 用户行为相似性计算方法及装置
US20170262445A1 (en) * 2016-03-08 2017-09-14 Facebook, Inc. Statistical feature engineering of user attributes
CN107480713A (zh) * 2017-08-08 2017-12-15 云南大学 含主导特征的co‑location模式挖掘方法

Also Published As

Publication number Publication date
CN110348581B (zh) 2023-08-18
CN110348581A (zh) 2019-10-18

Similar Documents

Publication Publication Date Title
WO2020182122A1 (zh) 用于生成文本匹配模型的方法和装置
US11062215B2 (en) Using different data sources for a predictive model
WO2021068513A1 (zh) 异常对象识别方法、装置、介质及电子设备
CN108520470B (zh) 用于生成用户属性信息的方法和装置
WO2021120677A1 (zh) 一种仓储模型训练方法、装置、计算机设备及存储介质
WO2020244065A1 (zh) 基于人工智能的字向量定义方法、装置、设备及存储介质
WO2023065859A1 (zh) 物品推荐方法、装置及存储介质
WO2019034087A1 (zh) 用户偏好确定方法、装置、设备及存储介质
CN113254716B (zh) 视频片段检索方法、装置、电子设备和可读存储介质
WO2022083093A1 (zh) 图谱中的概率计算方法、装置、计算机设备及存储介质
WO2021196935A1 (zh) 数据校验方法、装置、电子设备和存储介质
WO2021184547A1 (zh) 对话机器人意图语料生成方法、装置、介质及电子设备
WO2023217127A1 (zh) 一种因果关系确定方法及相关设备
CN115841366B (zh) 物品推荐模型训练方法、装置、电子设备及存储介质
WO2023226947A1 (zh) 端云协同推荐系统、方法以及电子设备
WO2020252925A1 (zh) 用户特征群中用户特征寻优方法、装置、电子设备及计算机非易失性可读存储介质
WO2021143016A1 (zh) 近似数据处理方法、装置、介质及电子设备
CN116684330A (zh) 基于人工智能的流量预测方法、装置、设备及存储介质
US20230041339A1 (en) Method, device, and computer program product for user behavior prediction
CN113344647B (zh) 一种信息推荐的方法及装置
CN111582649B (zh) 基于用户app独热编码的风险评估方法、装置和电子设备
WO2023236588A1 (zh) 基于客群偏差平滑优化的用户分类方法及装置
WO2021072864A1 (zh) 文本相似度获取方法、装置、电子设备及计算机可读存储介质
CN116401372A (zh) 知识图谱表示学习方法、装置、电子设备及可读存储介质
CN114357180A (zh) 知识图谱的更新方法及电子设备

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19934101

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19934101

Country of ref document: EP

Kind code of ref document: A1