CN112819553A

CN112819553A - Online advertisement pushing method combined with big data analysis and cloud server

Info

Publication number: CN112819553A
Application number: CN202110354811.1A
Authority: CN
Inventors: 郭举; 陈宙
Original assignee: Individual
Current assignee: Shanghai Moxing Culture Communication Co ltd
Priority date: 2021-04-01
Filing date: 2021-04-01
Publication date: 2021-05-18
Anticipated expiration: 2041-04-01
Also published as: CN112819553B

Abstract

According to the online advertisement pushing method and the cloud server combined with big data analysis, on one hand, hot advertisement information in online advertisement content is information with relatively more putting times, so that corresponding advertisement crowd data is relatively stable, and in the training process, the model has high delay performance, so that the reliability of model training is improved. On the other hand, as the training set adopted by the advertisement big data processing model training is a multidimensional advertisement delivery strategy obtained based on a plurality of online advertisement information of online advertisement contents, the model obtained by the training in the above way is used for matching the online multimedia platform of the online advertisement to be delivered, so that the condition that the online advertisement to be delivered is matched with the online multimedia platform mistakenly due to the difference of different online advertisement information can be avoided, the online advertisement to be delivered can be delivered and delivered on a proper online multimedia platform, and the waste of network resources caused by the unmatched advertisement delivery behavior is reduced.

Description

Online advertisement pushing method combined with big data analysis and cloud server

Technical Field

The disclosure relates to the technical field of big data and advertisement pushing, in particular to an online advertisement pushing method and a cloud server combining big data analysis.

Background

With the development of computer technology and the popularization of the internet, the accumulation of data information has reached a very huge step. The increase of data information is also accelerating continuously, and along with the acceleration of the construction of the internet and the internet of things, the data information is increased explosively.

Big data refers to a data set which cannot be captured, managed and processed by a conventional software tool within a certain time range, and massive, high-growth rate and diversified information assets which can have stronger decision-making power, insight discovery power and flow optimization capability only by a new processing mode are needed. At present, various enterprises can utilize related big data analysis to help the enterprises to reduce cost, improve efficiency, develop new products, make more intelligent business decisions and the like, and can also integrate and analyze data to obtain information and data relationship to be used for perceiving business trends, judging research quality, avoiding disease diffusion, or determining instant traffic road conditions and the like.

By taking advertisement pushing as an example, online advertisements can be pushed on different platforms through a big data technology, so that the coverage of advertisement pushing is improved. However, the related online advertisement push technology is difficult to realize the accurate matching between the online advertisement and the push platform, and further difficult to ensure that the online advertisement can be pushed and delivered on a proper online multimedia platform.

Disclosure of Invention

In order to solve the technical problems in the related art, the present disclosure provides an online advertisement push method and a cloud server in combination with big data analysis.

The invention provides an online advertisement pushing method combined with big data analysis, which is applied to a cloud server communicated with a plurality of online multimedia platforms, and comprises the following steps:

extracting hot advertisement information in online advertisement content of a target online multimedia platform to obtain a first hot advertisement information set; screening a set number of hot advertisement information from the first hot advertisement information set to obtain a second hot advertisement information set;

acquiring advertisement delivery data corresponding to each piece of hot advertisement information by extracting advertisement crowd data of each piece of hot advertisement information in the second hot advertisement information set so as to obtain a multi-dimensional advertisement delivery strategy; determining effective advertisement information corresponding to the same target online multimedia platform and ineffective advertisement information corresponding to different target online multimedia platforms based on the multi-dimensional advertisement putting strategy; training a pre-established advertisement big data processing model by using the effective advertisement information and the ineffective advertisement information to obtain a trained target model;

and when the online advertisement to be pushed is obtained, identifying the online multimedia platform corresponding to the online advertisement to be pushed by using the trained target model.

Preferably, the determining, based on the multidimensional advertisement delivery policy, effective advertisement information corresponding to the same target online multimedia platform and ineffective advertisement information corresponding to different target online multimedia platforms includes:

determining corresponding first effective advertisement information by using different multidimensional advertisement putting strategies of the same target online multimedia platform, updating the priority of the advertisement putting data in the multidimensional advertisement putting strategies to obtain advertisement putting updating data, and determining the advertisement putting updating data and the corresponding multidimensional advertisement putting strategies as second effective advertisement information;

and determining the corresponding invalid advertisement information by using the multidimensional advertisement putting strategies of different target online multimedia platforms.

Preferably, the determining, based on the multidimensional advertisement delivery policy, effective advertisement information corresponding to the same target online multimedia platform and ineffective advertisement information corresponding to different target online multimedia platforms includes: obtaining the first effective advertisement information, the second effective advertisement information and the ineffective advertisement information in advance;

correspondingly, the training of the pre-established advertisement big data processing model by using the effective advertisement information and the ineffective advertisement information to obtain a trained target model comprises: inputting the first effective advertisement information, the second effective advertisement information and the ineffective advertisement information into the pre-established advertisement big data processing model for training to obtain a target model for completing training.

Preferably, the effective advertisement information corresponding to the same target online multimedia platform and the ineffective advertisement information corresponding to different target online multimedia platforms are determined based on the multidimensional advertisement putting strategy; training a pre-established advertisement big data processing model by using the effective advertisement information and the ineffective advertisement information to obtain a trained target model, comprising:

obtaining the first effective advertisement information and the ineffective advertisement information in advance;

inputting the first effective advertisement information and the ineffective advertisement information into the pre-established advertisement big data processing model for training;

in the training process, when a triggering condition for training by using the second effective advertisement information is reached, selecting any one multidimensional advertisement putting strategy from the first effective advertisement information or the ineffective advertisement information;

updating the priority of the advertisement putting data in the multi-dimensional advertisement putting strategy to obtain advertisement putting updating data, and determining the advertisement putting updating data and the corresponding multi-dimensional advertisement putting strategy as the second effective advertisement information;

and training the advertisement big data processing model by using the second effective advertisement information determined currently.

Preferably, the screening out a set number of the popular advertisement information from the first popular advertisement information set to obtain a second popular advertisement information set includes:

determining a first advertisement putting frequency corresponding to the popular advertisement information in the first popular advertisement information set based on a preset advertisement popular degree;

dividing the popular advertisement information corresponding to the same first advertisement putting frequency into the same popular advertisement information subset;

determining any hot advertisement information subset according to a preset instruction; screening a set number of hot advertisement information from the determined hot advertisement information subsets according to a preset indication, or screening a set number of hot advertisement information from the determined hot advertisement information subsets and the hot advertisement information subsets of which the first advertisement putting frequency is higher than the historical putting frequency of the hot advertisement information subsets according to a preset indication, so as to obtain a second hot advertisement information set;

correspondingly, the obtaining of the advertisement delivery data corresponding to each of the popular advertisement information by extracting the advertisement crowd data of each of the popular advertisement information in the second popular advertisement information set to obtain the multidimensional advertisement delivery strategy includes:

extracting advertisement crowd data of each piece of hot advertisement information in the second hot advertisement information set;

converting the delivery times of each advertisement crowd data into the first advertisement delivery frequency corresponding to the current popular advertisement information subset to obtain first advertisement delivery data corresponding to each popular advertisement information;

and integrating the first advertisement putting data according to a preset instruction, and determining the first advertisement putting data as the corresponding multi-dimensional advertisement putting strategy.

screening a set number of hot advertisement information from the first hot advertisement information set according to a preset indication to obtain a second hot advertisement information set;

converting the putting times of each advertisement crowd data into a second advertisement putting frequency to obtain second advertisement putting data corresponding to each popular advertisement information;

and integrating the second advertisement delivery data according to a preset instruction, and determining the second advertisement delivery data as the corresponding multi-dimensional advertisement delivery strategy.

Preferably, before the step of converting the number of impressions of each advertisement crowd data into the second advertisement impression frequency, the method further includes:

determining the effective delivery times of the first popular advertisement information set of the current target online multimedia platform, and determining the effective delivery times as the second advertisement delivery frequency;

or determining the effective delivery times of the first popular advertisement information sets of all the target online multimedia platforms, and determining the effective delivery times as the second advertisement delivery frequency.

Preferably, before the screening out a set number of the popular advertisement information from the first popular advertisement information set to obtain a second popular advertisement information set, the method further includes:

determining first target hot advertisement information corresponding to a first set advertisement label from the first hot advertisement information set, wherein the frequency of putting the first target hot advertisement information is higher than the frequency of putting other hot advertisement information in the first hot advertisement information set;

determining second target hot advertisement information corresponding to a second set advertisement label from the first hot advertisement information set, wherein the releasing times of the second target hot advertisement information are less than the releasing times of other hot advertisement information in the first hot advertisement information set;

removing the first targeted trending advertising information and the second targeted trending advertising information from the first set of trending advertising information.

Preferably, when the online advertisement to be pushed is obtained, identifying the online multimedia platform corresponding to the online advertisement to be pushed by using the trained target model includes:

when the online advertisement to be pushed is obtained, selecting advertisement platform matching data of the online advertisement to be pushed by using the trained target model to obtain target advertisement platform matching data;

matching the target advertisement platform matching data with a pre-established advertisement platform matching data set to determine an online multimedia platform corresponding to the target advertisement platform matching data; the advertisement platform matching data set comprises advertisement platform matching data obtained by selecting advertisement content data of a popular advertisement content set by using the trained target model and corresponding online multimedia platform identification;

when the online advertisement to be pushed is obtained, the trained target model is used for selecting the advertisement platform matching data of the online advertisement to be pushed so as to obtain the target advertisement platform matching data, and the method comprises the following steps:

acquiring the multidimensional advertisement putting strategy corresponding to the online advertisement to be pushed to obtain a target multidimensional advertisement putting strategy and/or updating the priority of the advertisement putting data in the target multidimensional advertisement putting strategy to obtain target advertisement putting updating data;

selecting a plurality of target multi-dimensional advertisement putting strategies and/or advertisement platform matching data of a plurality of target multi-advertisement putting updating data by using the trained target model;

and determining the data meeting the timeliness requirement of the selected advertisement platform matching data to obtain the target advertisement platform matching data.

The invention also provides a cloud server, which comprises a processor and a memory; the processor is connected with the memory in communication, and the processor is used for reading the computer program from the memory and executing the computer program to realize the method.

The technical scheme provided by the embodiment of the disclosure can have the following beneficial effects.

It can be understood that, in the present application, hot advertisement information in online advertisement content of a target online multimedia platform is extracted to obtain a first hot advertisement information set, a set number of hot advertisement information is screened from the first hot advertisement information set to obtain a second hot advertisement information set, then advertisement crowd data corresponding to each hot advertisement information in the second hot advertisement information set is extracted to obtain advertisement delivery data corresponding to each hot advertisement information to obtain a multidimensional advertisement delivery strategy, valid advertisement information corresponding to the same target online multimedia platform and invalid advertisement information corresponding to different target online multimedia platforms are determined based on the multidimensional advertisement delivery strategy, and then a pre-established advertisement big data processing model is trained to obtain a trained target model, and when the online advertisement to be pushed is obtained, identifying the online multimedia platform corresponding to the online advertisement to be pushed by using the trained target model.

That is, hot advertisement information of online advertisement content of a target online multimedia platform is extracted first, then a set number of hot advertisement information is screened out, and then advertisement putting data corresponding to each hot advertisement information is obtained by extracting advertisement crowd data of each screened out hot advertisement information, so that a multi-dimensional advertisement putting strategy is obtained, and training of an advertisement big data processing model is performed based on the multi-dimensional advertisement putting strategy. Therefore, on one hand, on the one hand, the hot advertisement information in the online advertisement content is the information with relatively more putting times, so that the corresponding advertisement crowd data is relatively stable, and in the training process, the delayed use performance of the model is relatively high, thereby improving the reliability of model training. On the other hand, as the training set adopted by the advertisement big data processing model training is a multidimensional advertisement delivery strategy obtained based on a plurality of online advertisement information of online advertisement contents, the model obtained by the training in the above way is used for matching the online multimedia platform of the online advertisement to be delivered, so that the condition that the online advertisement to be delivered is matched with the online multimedia platform mistakenly due to the difference of different online advertisement information can be avoided, the online advertisement to be delivered can be delivered and delivered on a proper online multimedia platform, and the waste of network resources caused by the unmatched advertisement delivery behavior is reduced.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and together with the description, serve to explain the principles of the invention.

Fig. 1 is a schematic diagram of a hardware structure of a cloud server according to an embodiment of the present invention.

Fig. 2 is a schematic communication architecture diagram of an online advertisement push system incorporating big data analysis according to an embodiment of the present invention.

Fig. 3 is a flowchart of an online advertisement pushing method in conjunction with big data analysis according to an embodiment of the present invention.

Fig. 4 is a block diagram of an online advertisement push device incorporating big data analysis according to an embodiment of the present invention.

Detailed Description

Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order.

The method provided by the embodiment of the application can be executed in a cloud server, a computer device or a similar operation device. Taking an operation on a cloud server as an example, fig. 1 is a hardware structure block diagram of a cloud server implementing an online advertisement push method combined with big data analysis according to an embodiment of the present invention. As shown in fig. 1, the cloud server 10 may include one or more processors 102 (only one is shown in fig. 1) (the processor 102 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, and optionally, may further include a transmission device 106 for communication functions. It will be understood by those of ordinary skill in the art that the structure shown in fig. 1 is merely illustrative and is not intended to limit the structure of the cloud server described above. For example, cloud server 10 may also include more or fewer components than shown in FIG. 1, or have a different configuration than shown in FIG. 1.

The memory 104 may be used to store computer programs, for example, software programs and modules of application software, such as a computer program corresponding to the online advertisement delivery method combined with big data analysis in the embodiment of the present invention, and the processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, so as to implement the above-mentioned method. The memory 104 may include high speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, memory 104 may further include memory located remotely from processor 102, which may be connected to cloud server 10 via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

The transmission device 106 is used for receiving or transmitting data via a network. The specific example of the network described above may include a wireless network provided by a communication provider of the cloud server 10. In one example, the transmission device 106 includes a Network adapter (NIC), which can be connected to other Network devices through a base station so as to communicate with the internet. In one example, the transmission device 106 may be a Radio Frequency (RF) module, which is used for communicating with the internet in a wireless manner.

Currently, the advertisement push matching of the online multimedia platform based on the advertisement big data processing generally extracts online advertisement information with relatively more advertisement content in online advertisement content, then selects data such as key words of the online advertisement information, and inputs the selected data such as the key words into the advertisement big data processing network for training and learning, so as to obtain a trained target model for performing the online multimedia platform matching of the online advertisement to be pushed, however, the training and learning are performed based on the online advertisement information with relatively more advertisement content in the online advertisement content, on one hand, the data noise of the data such as the key words for training is more, which results in the efficiency of data characteristic processing and matching being reduced, thereby affecting the delayed performance of the model, on the other hand, the trained model is in the process of the advertisement push matching of the multimedia platform, the differences between different information of online advertising content make it difficult to ensure accurate pairing of ad push and multimedia platforms. Therefore, the online advertisement pushing method combining big data analysis is provided, the delay performance of the model can be improved, and the online advertisement to be pushed can be pushed and launched on a proper online multimedia platform.

In the online advertisement push method combining big data analysis of the present application, a system framework adopted may specifically refer to fig. 2, and the system framework may specifically include: a cloud server 10 and several online multimedia platforms 20 establishing a communication connection with the cloud server 10. The online multimedia platform 20 includes, but is not limited to, a streaming media platform or a digital media platform, and is not limited thereto.

In the present application, the cloud server 10 executes the steps of the online advertisement push method in combination with big data analysis, including extracting the hot advertisement information in the online advertisement content of the target online multimedia platform to obtain a first hot advertisement information set; screening a set number of hot advertisement information from the first hot advertisement information set to obtain a second hot advertisement information set; acquiring advertisement delivery data corresponding to each piece of hot advertisement information by extracting advertisement crowd data of each piece of hot advertisement information in the second hot advertisement information set so as to obtain a multi-dimensional advertisement delivery strategy; determining effective advertisement information corresponding to the same target online multimedia platform and ineffective advertisement information corresponding to different target online multimedia platforms based on the multi-dimensional advertisement putting strategy; training a pre-established advertisement big data processing model by using the effective advertisement information and the ineffective advertisement information to obtain a trained target model; and when the online advertisement to be pushed is obtained, identifying the online multimedia platform corresponding to the online advertisement to be pushed by using the trained target model.

It can be understood that the online advertisement push method combined with big data analysis provided by the application can be particularly applied to an online multimedia platform matching scene of an advertisement push application, the advertisement push application generally has an online multimedia platform selection function, and the selection of the online multimedia platform can help an advertisement push party to match to a proper online multimedia platform more quickly and accurately. Because the number of the online advertisements is relatively large, in the process of selecting the online multimedia platform for the online advertisements to be pushed, the cloud server can execute the steps of the online advertisement pushing method combined with big data analysis, perform model training by using part of the online advertisement content, and then perform online multimedia platform matching of the online advertisements to be pushed on the rest of the online advertisements by using the trained target model, so that the matching of the online advertisements and the online multimedia platform is realized, and the matching result can be output or recorded in the online multimedia platform. The advertisement presenter can check the matching result of the online advertisement to be pushed and the online multimedia platform through the corresponding interaction device, such as checking the "ad 1 of the online advertisement to be pushed", thereby entering the statistical list corresponding to the online multimedia platform matched with the "ad 1 of the online advertisement to be pushed".

Referring to fig. 3, an embodiment of the present application discloses an online advertisement push method combining big data analysis, which specifically includes:

s11, hot advertisement information in the online advertisement content of the target online multimedia platform is extracted to obtain a first hot advertisement information set.

For some possible embodiments, the hot advertisement information can be extracted through a classifier model, a deep learning strategy based on historical extraction results, or algorithms such as to-be-processed advertisement information selected according to manual marking, so as to obtain a first hot advertisement information set of the target online multimedia platform.

Further, in this embodiment, first target hot advertisement information corresponding to a first set advertisement tag may be determined from the first hot advertisement information set, and the frequency of delivering the first target hot advertisement information is higher than the frequency of delivering other hot advertisement information in the first hot advertisement information set; determining second target hot advertisement information corresponding to a second set advertisement label from the first hot advertisement information set, wherein the releasing times of the second target hot advertisement information are less than the releasing times of other hot advertisement information in the first hot advertisement information set; and removing the first target hot advertisement information and the second target hot advertisement information from the first hot advertisement information set.

It should be noted that, in order to correct the information noise and the repeated advertisement information introduced by the trending advertisement information extraction algorithm, the present embodiment may perform placement frequency statistical analysis on the first trending advertisement information set of the target online multimedia platform, and generally, the placement frequency of the trending advertisement information of each online multimedia platform approximately follows gaussian distribution. It can be understood that relatively more or less hot advertisement information is extracted and obtained probably due to information noise of a hot advertisement information extraction algorithm, and in the process of repeated advertisement information screening, because the hot advertisement information with shorter time is reserved, if excessive reservation operation is performed, the information noise is introduced to influence the delay performance of the model, and more hot advertisement information is the hot advertisement to be verified. Therefore, in order to correct information noise introduced by the hot advertisement information extraction algorithm and avoid excessive repeated advertisement query operation in the repeated advertisement information screening process, the hot advertisement information with longer time corresponding to the first set advertisement tag in the first hot advertisement information set and the hot advertisement information with shorter time corresponding to the second set advertisement tag in the second hot advertisement information set can be filtered. For example, the hot advertisement information with the filtering and placing times exceeding x% hot advertisement information in the first hot advertisement information set and the hot advertisement information with the filtering and placing times shorter than y% hot advertisement information in the first hot advertisement information set.

S12, screening out a set number of hot advertisement information from the first hot advertisement information set to obtain a second hot advertisement information set.

S13, obtaining advertisement putting data corresponding to each hot advertisement information by extracting advertisement crowd data of each hot advertisement information in the second hot advertisement information set so as to obtain a multi-dimensional advertisement putting strategy.

In deep learning, the hot advertisement information of the classification matching process has a concept of "intention to place". Therefore, the embodiment can combine the popular advertisement information into the multidimensional advertisement delivery strategy based on different advertisement delivery intentions, that is, the advertisement crowd data of a plurality of popular advertisement information forms the multidimensional advertisement delivery strategy for training.

For some possible embodiments, a set number of hot advertisement information may be directly screened out from the first hot advertisement information set according to a preset instruction to obtain the second hot advertisement information set, then advertisement crowd data of each hot advertisement information in the second hot advertisement information set is extracted, the delivery times of each advertisement crowd data are converted into a second advertisement delivery frequency to obtain second advertisement delivery data corresponding to each hot advertisement information, and then the second advertisement delivery data are integrated according to the preset instruction to determine the corresponding multidimensional advertisement delivery strategy.

Wherein, the specific process of converting the delivery times of each advertisement crowd data into a second advertisement delivery frequency comprises: if the advertisement putting frequency corresponding to the advertisement crowd is larger than the second advertisement putting frequency, extracting data of the second advertisement putting frequency from the advertisement crowd data, and if the putting frequency corresponding to the advertisement crowd is smaller than the second advertisement putting frequency, performing retention operation on the advertisement crowd data to obtain data of the second advertisement putting frequency.

In an alternative embodiment, the number of effective impressions of the first hot advertisement information set of the current target online multimedia platform may be determined, and the number of effective impressions may be determined as the second advertisement impression frequency.

In some other embodiments, the number of effective impressions of the first hot advertisement information set of all the target online multimedia platforms may be determined, and the number of effective impressions may be determined as the second advertisement impression frequency.

That is, the present embodiment may perform the operations of S11 to S13 on the online advertisement content of a plurality of target online multimedia platforms to determine the multidimensional advertisement delivery policy corresponding to each target online multimedia platform. And moreover, the corresponding multidimensional advertisement putting strategy can be determined by utilizing a plurality of online advertisement contents of each target online multimedia platform, so that the credibility of the multidimensional advertisement putting strategy as a decision basis in the matching process of the online advertisement to be pushed and the online multimedia platform can be ensured.

For example, the effective delivery times of a first hot advertisement information set of all target online multimedia platforms are determined, the effective delivery times are determined as a second advertisement delivery frequency F, that is, the advertisement delivery times within a specific time duration, k hot advertisement information are directly screened out from the first hot advertisement information set of the target online multimedia platform of which the multidimensional advertisement delivery strategy is to be determined currently according to a preset instruction to obtain a second hot advertisement information set, then an advertisement characteristic segment character of each hot advertisement information in the second hot advertisement information set is extracted, the number of characteristic classifiers Classifier corresponding to the advertisement characteristic segment character is set as M, the delivery times of each advertisement characteristic segment character are converted into the second advertisement delivery frequency F, then the advertisement characteristic segments characters of the k hot advertisement information are integrated according to the preset instruction, and the characteristic matrix is determined as (M, f, k) of the multi-dimensional advertisement delivery strategy.

In an alternative embodiment, the advertisement delivery data corresponding to each piece of popular advertisement information may be obtained by extracting click rate data of each piece of popular advertisement information, so as to obtain the multidimensional advertisement delivery policy.

It should be noted that, in the embodiment of the present application, a set number of pieces of hot advertisement information are screened according to a preset indication to determine a corresponding multidimensional advertisement delivery policy, so that the hot advertisement information of the same hot advertisement content can be combined according to the preset indication for multiple times, which is convenient for effective expansion of advertisement resources, and further improves the processing efficiency of the hot advertisement information.

S14, determining effective advertisement information corresponding to the same target online multimedia platform and ineffective advertisement information corresponding to different target online multimedia platforms based on the multidimensional advertisement putting strategy.

For example, the effective advertisement information may be understood as advertisement information in which a click amount, a subsequent browsing operation, or a consultation operation exists in a set period, and the ineffective advertisement information may be understood as advertisement information in which a click amount does not exist in a set period.

And S15, training a pre-established advertisement big data processing model by using the effective advertisement information and the ineffective advertisement information to obtain a trained target model.

It should be noted that, in this embodiment, a method of tag processing of advertisement big data may be adopted to perform advertisement platform matching data selection. It can be understood that the advertisement information from the same online multimedia platform is marked as 1, the advertisement information from different online multimedia platforms is marked as 0, and the target advertisement information which can be used as the basis for judging the advertisement matching of the online multimedia platforms is selected by learning the similarity of the same online multimedia platform and the difference of the matching data of different advertisement platforms.

For some possible embodiments, the corresponding first effective advertisement information may be determined by using different multidimensional advertisement delivery strategies of the same target online multimedia platform, the corresponding ineffective advertisement information may be determined by using different multidimensional advertisement delivery strategies of the different target online multimedia platforms, and then the pre-established advertisement big data processing model may be trained by using the first effective advertisement information and the ineffective advertisement information to obtain a trained target model.

Further, in this embodiment, based on a heat map analysis or a user behavior analysis (knowledge graph) training method in the advertisement big data tag processing technology, taking the heat map analysis training method as an example, the corresponding first Effective advertisement information is determined by using two different multidimensional advertisement delivery strategies of the same target online multimedia platform, the corresponding ineffective advertisement information Invalid advertisement information is determined by using two multidimensional advertisement delivery strategies of different target online multimedia platforms, the Effective advertisement information is marked as 1, the Invalid advertisement information is marked as 0, and training is performed to obtain a trained target model.

In the embodiment, a feedforward neural network can be used as a basic structure, and the advertisement big data processing model established in advance can be understood that the matrix description value of the feature matrix corresponding to the hot advertisement information is small, so that the subsequent advertisement information processing efficiency is high, and the model training rate can be further improved.

And S16, when the online advertisement to be pushed is obtained, identifying the online multimedia platform corresponding to the online advertisement to be pushed by using the trained target model.

For some possible embodiments, the trained target model can be used to select advertisement platform matching data corresponding to the online advertisement to be pushed, and then the corresponding online multimedia platform can be identified.

It can be understood that, in the embodiment of the present application, hot advertisement information in online advertisement content of a target online multimedia platform is extracted to obtain a first hot advertisement information set, a set number of hot advertisement information is screened from the first hot advertisement information set to obtain a second hot advertisement information set, then advertisement crowd data corresponding to each hot advertisement information in the second hot advertisement information set is extracted to obtain advertisement putting data corresponding to each hot advertisement information to obtain a multidimensional advertisement putting strategy, valid advertisement information corresponding to the same target online multimedia platform and invalid advertisement information corresponding to different target online multimedia platforms are determined based on the multidimensional advertisement putting strategy, and then a pre-established advertisement big data processing model is trained to obtain a trained target model, and when the online advertisement to be pushed is obtained, identifying the online multimedia platform corresponding to the online advertisement to be pushed by using the trained target model.

That is, in the embodiment of the application, hot advertisement information of online advertisement content of a target online multimedia platform is extracted first, then a set number of hot advertisement information is screened out, and then advertisement delivery data corresponding to each hot advertisement information is obtained by extracting advertisement crowd data of each screened out hot advertisement information, so as to obtain a multidimensional advertisement delivery strategy, and training of an advertisement big data processing model is performed based on the multidimensional advertisement delivery strategy.

Therefore, on one hand, on the one hand, the hot advertisement information in the online advertisement content is the information with relatively more putting times, so that the corresponding advertisement crowd data is relatively stable, and in the training process, the delayed use performance of the model is relatively high, thereby improving the reliability of model training.

On the other hand, as the training set adopted by the advertisement big data processing model training is a multidimensional advertisement delivery strategy obtained based on a plurality of online advertisement information of online advertisement contents, the model obtained by the training in the above way is used for matching the online multimedia platform of the online advertisement to be delivered, so that the condition that the online advertisement to be delivered is matched with the online multimedia platform mistakenly due to the difference of different online advertisement information can be avoided, the online advertisement to be delivered can be delivered and delivered on a proper online multimedia platform, and the waste of network resources caused by the unmatched advertisement delivery behavior is reduced.

Based on the same inventive concept, the embodiment of the application discloses a specific online advertisement pushing method combined with big data analysis, which specifically comprises the following steps:

s21, hot advertisement information in the online advertisement content of the target online multimedia platform is extracted to obtain a first hot advertisement information set.

S22, determining a first advertisement putting frequency corresponding to the popular advertisement information in the first popular advertisement information set based on a preset advertisement popular degree.

S23, dividing the hot advertisement information corresponding to the same first advertisement putting frequency into the same hot advertisement information subset.

And S24, determining any hot advertisement information subset according to preset instructions.

S25, screening a set number of hot advertisement information from the determined hot advertisement information subsets according to a preset instruction, or screening a set number of hot advertisement information from the determined hot advertisement information subsets and the hot advertisement information subsets with the first advertisement delivery frequency higher than the historical delivery frequency of the hot advertisement information subsets according to a preset instruction, so as to obtain a second hot advertisement information set.

For some possible embodiments, the present application may determine the first advertisement delivery frequency and the advertisement hot degree of the delivery times of the hot advertisement information based on the delivery times of all the hot advertisement information in the first hot advertisement information set, and then determine the first advertisement delivery frequency corresponding to the hot advertisement information in the first hot advertisement information set based on the determined advertisement hot degree. For example, according to the distribution of the number of impressions of all the hot advertisement information in a first hot advertisement information set, several impressions weights are determined to obtain corresponding first advertisement impression frequencies, such as f1, f2, f3, and the hot advertisement information corresponding to the same first advertisement impression frequency is divided into the same hot advertisement information subsets. For example, the hot advertisement information with the impression frequency n1 and n2 corresponds to the first advertisement impression frequency f1, and the hot advertisement information with the impression frequency n1 and n2 is divided into the same hot advertisement information subset, and the first advertisement impression frequency corresponding to the hot advertisement information subset is f 1. Accordingly, the hot advertisement information subsets corresponding to other first advertisement putting frequencies can be obtained.

That is, according to the distribution situation of the number of times of putting the hot advertisement information, the number of times of putting corresponding to the hot advertisement information can be quantized into a plurality of batches, and accordingly the hot advertisement information is classified into different sets of times of putting so as to obtain different hot advertisement information subsets. And selecting a delivery time array according to a preset indication, namely selecting a hot advertisement information subset. And screening a set number of hot advertisement information from the putting number array or the putting number array and a group with the putting number higher than the putting number array according to a preset indication to obtain a second hot advertisement information set.

S26, obtaining advertisement putting data corresponding to each hot advertisement information by extracting advertisement crowd data of each hot advertisement information in the second hot advertisement information set so as to obtain a multi-dimensional advertisement putting strategy.

For some possible embodiments, this embodiment may extract advertisement crowd data of each of the popular advertisement information in the second popular advertisement information set; converting the delivery times of each advertisement crowd data into the first advertisement delivery frequency corresponding to the current popular advertisement information subset to obtain first advertisement delivery data corresponding to each popular advertisement information; and integrating the first advertisement putting data according to a preset instruction, and determining the first advertisement putting data as the corresponding multi-dimensional advertisement putting strategy.

For example, k pieces of popular advertisement information are screened out from the determined popular advertisement information subsets according to preset instructions to obtain a second popular advertisement information set, then an advertisement feature segment character of each piece of popular advertisement information in the second popular advertisement information set is extracted, the number of feature classifiers Classifier corresponding to the advertisement feature segment character is set to be M, the number of delivery times of each advertisement feature segment character is converted into a first advertisement delivery frequency A corresponding to the current popular advertisement information subset, then the advertisement feature segments characters of the k pieces of popular advertisement information are integrated according to the preset instructions, and the multidimensional advertisement delivery strategy with the feature matrix of (M, A, k) is determined.

S27, determining effective advertisement information corresponding to the same target online multimedia platform and ineffective advertisement information corresponding to different target online multimedia platforms based on the multidimensional advertisement putting strategy.

And S28, training a pre-established advertisement big data processing model by using the effective advertisement information and the ineffective advertisement information to obtain a trained target model.

And S29, when the online advertisement to be pushed is obtained, identifying the online multimedia platform corresponding to the online advertisement to be pushed by using the trained target model.

It can be understood that, in the embodiment of the present application, hot advertisement information corresponding to the same first advertisement delivery frequency is divided into the same hot advertisement information subsets, when the hot advertisement information is screened, any one of the hot advertisement information subsets is determined according to a preset indication, then a set number of hot advertisement information is screened from the determined hot advertisement information subsets according to the preset indication to obtain a second hot advertisement information set, correspondingly, in the process of classifying the advertisement crowd delivery times, the delivery times of each advertisement crowd data are converted into the first advertisement delivery frequency corresponding to the current hot advertisement information subset, because the hot advertisement information is divided into the same hot advertisement information subsets according to the corresponding first advertisement delivery frequency, the processing efficiency of the hot advertisement information is improved, and the occurrence of errors in the hot advertisement information division is reduced, and further the service-extending performance of the model is improved. And screening a set number of hot advertisement information from the determined hot advertisement information subset and the hot advertisement information subset with the first advertisement putting frequency higher than the historical putting frequency of the hot advertisement information subset according to a preset indication to obtain a second hot advertisement information set, wherein in the process of classifying the putting times of the advertisement crowd, the putting times of each advertisement crowd data are converted into the first advertisement putting frequency corresponding to the hot advertisement information subset screened according to the preset indication, so that the error condition in the process of dividing the hot advertisement information can be further reduced.

For some possible embodiments, the present application discloses a specific online advertisement pushing method combining big data analysis, including:

s31, hot advertisement information in the online advertisement content of the target online multimedia platform is extracted to obtain a first hot advertisement information set.

S32, screening out a set number of hot advertisement information from the first hot advertisement information set to obtain a second hot advertisement information set.

S33, obtaining advertisement putting data corresponding to each hot advertisement information by extracting advertisement crowd data of each hot advertisement information in the second hot advertisement information set so as to obtain a multi-dimensional advertisement putting strategy.

S34, determining corresponding first effective advertisement information by using different multidimensional advertisement putting strategies of the same target online multimedia platform, updating the priority of the advertisement putting data in the multidimensional advertisement putting strategies to obtain advertisement putting updating data, determining the advertisement putting updating data and the corresponding multidimensional advertisement putting strategies as second effective advertisement information, and determining corresponding ineffective advertisement information by using the multidimensional advertisement putting strategies of different target online multimedia platforms.

S35, training a pre-established advertisement big data processing model by utilizing the first effective advertisement information, the second effective advertisement information and the ineffective advertisement information to obtain a trained target model.

In a specific implementation manner, the embodiment may obtain the first effective advertisement information, the second effective advertisement information, and the ineffective advertisement information in advance, and then input the first effective advertisement information, the second effective advertisement information, and the ineffective advertisement information into the advertisement big data processing model established in advance for training, so as to obtain a target model for completing training.

For example, based on the heat map analysis training method, two different multidimensional advertisement delivery strategies of the same target online multimedia platform are used to determine corresponding first Effective advertisement information Effective advertisement, two multidimensional advertisement delivery strategies of different target online multimedia platforms are used to determine corresponding Invalid advertisement information Invalid advertisement, a multidimensional advertisement delivery strategy to be updated for advertisement delivery is selected, the priority of advertisement delivery data in the selected multidimensional advertisement delivery strategy is updated to obtain advertisement delivery update data, and the advertisement delivery update data and the corresponding multidimensional advertisement delivery strategy are determined as second Effective advertisement information AD.

In another specific implementation, the embodiment may obtain the first effective advertisement information and the ineffective advertisement information in advance, and then input the first effective advertisement information and the ineffective advertisement information into the advertisement big data processing model established in advance for training; in the training process, when a triggering condition for training by using the second effective advertisement information is reached, selecting any one multidimensional advertisement putting strategy from the first effective advertisement information or the ineffective advertisement information; updating the priority of the advertisement putting data in the multi-dimensional advertisement putting strategy to obtain advertisement putting updating data, and determining the advertisement putting updating data and the corresponding multi-dimensional advertisement putting strategy as the second effective advertisement information; and training the advertisement big data processing model by using the second effective advertisement information determined currently.

For some possible embodiments, the advertisement information used for training uses the first effective advertisement information and the second effective advertisement information in turn, and when the second effective advertisement information is needed for training, the corresponding second effective advertisement information is determined. That is, for some possible embodiments, the second effective advertisement information including the advertisement placement update data may be determined before training, or the second effective advertisement information including the advertisement placement update data may be determined during training.

And S36, when the online advertisement to be pushed is obtained, identifying the online multimedia platform corresponding to the online advertisement to be pushed by using the trained target model.

It can be understood that the effective advertisement information established in the embodiment of the present application includes not only the first effective advertisement information determined by different multidimensional advertisement putting strategies of the same target online multimedia platform, but also updating the priority of the advertisement putting data in the multidimensional advertisement putting strategy to obtain advertisement putting updating data, and training and learning by using the advertisement putting updating data and the corresponding second effective advertisement information determined by the multidimensional advertisement putting strategy, and by using such effective advertisement information and the corresponding ineffective advertisement information of different target online multimedia platforms, model learning can be divided into two parts of online multimedia platform correlation training and learning of advertisement putting updating and pushing feedback training and learning of advertisement big data of the online multimedia platforms. Therefore, the model learns the advertisement contents/information of the same batch, and can disturb the randomness corresponding to the advertisement putting sequence and the difference between the same advertisement platform matching data and different advertisement platform matching data according to the preset instructions. Therefore, overfitting between the obtained advertisement platform matching data of the trained target model and the advertisement putting priority of the multidimensional advertisement putting strategy is avoided, and effective expansion of advertisement resources can be realized by carrying out advertisement putting updating processing, the delayed use performance of the model is further improved, and the accuracy of advertisement pushing matching of the online multimedia platform is improved.

On the basis of the above content, the embodiment of the application discloses a specific online advertisement push method combining big data analysis, which includes:

s41, hot advertisement information in the online advertisement content of the target online multimedia platform is extracted to obtain a first hot advertisement information set.

S42, screening out a set number of hot advertisement information from the first hot advertisement information set to obtain a second hot advertisement information set.

S43, obtaining advertisement putting data corresponding to each hot advertisement information by extracting advertisement crowd data of each hot advertisement information in the second hot advertisement information set so as to obtain a multi-dimensional advertisement putting strategy.

S44, determining effective advertisement information corresponding to the same target online multimedia platform and ineffective advertisement information corresponding to different target online multimedia platforms based on the multidimensional advertisement putting strategy.

And S45, training a pre-established advertisement big data processing model by using the effective advertisement information and the ineffective advertisement information to obtain a trained target model.

And S46, when the online advertisement to be pushed is obtained, selecting the advertisement platform matching data of the online advertisement to be pushed by using the trained target model to obtain the target advertisement platform matching data.

For some possible embodiments, the present embodiment may obtain the multidimensional advertisement delivery policy corresponding to the online advertisement to be delivered to obtain a target multidimensional advertisement delivery policy and/or update the priority of the advertisement delivery data in the target multidimensional advertisement delivery policy to obtain target advertisement delivery update data; selecting a plurality of target multi-dimensional advertisement putting strategies and/or advertisement platform matching data of a plurality of target multi-advertisement putting updating data by using the trained target model; and determining the data meeting the timeliness requirement of the selected advertisement platform matching data to obtain the target advertisement platform matching data.

S47, matching the target advertisement platform matching data with a pre-established advertisement platform matching data set to determine an online multimedia platform corresponding to the target advertisement platform matching data; the advertisement platform matching data set comprises advertisement platform matching data obtained by selecting advertisement content data of the popular advertisement content set by using the trained target model and corresponding online multimedia platform identification.

That is, in the embodiment of the present application, in the model application stage, the multidimensional advertisement delivery policy of the online advertisement to be delivered may be obtained according to the contents of the foregoing S41 to S43, or advertisement delivery update may be further performed to obtain advertisement delivery update data, and then the trained target model of the present embodiment is used to select the advertisement platform matching data and then the data meeting timeliness is taken as the target advertisement platform matching data. And matching the target data with the online multimedia platform of the online advertisement to be pushed by comparing the similarity, such as Euclidean distance, of the matching data of the existing advertisement platform in the preset database.

It can be understood that, in the embodiment of the application, in the model application stage, a plurality of multidimensional advertisement delivery strategies and/or a plurality of multi-dimensional advertisement delivery strategies to be delivered to the online advertisement are/is utilized by the trained target model, the target multi-advertisement delivery updating data is used for selecting the advertisement platform matching data, then the data meeting the timeliness requirement is determined to obtain the target advertisement platform matching data, the matching accuracy of the target advertisement platform matching data and the pre-established advertisement platform matching data set can be improved, the advertisement platform matching data in the advertisement platform matching data set is utilized, the advertisement platform matching data obtained by selecting the advertisement content data from the hot advertisement content set by the trained target model, and the advertisement delivery matching accuracy of the online multimedia platform is further improved.

That is, in the embodiment of the application, hot advertisement information of online advertisement content of a target online multimedia platform is extracted first, then a set number of hot advertisement information is screened out, and then advertisement delivery data corresponding to each hot advertisement information is obtained by extracting advertisement crowd data of each screened out hot advertisement information, so as to obtain a multidimensional advertisement delivery strategy, and training of an advertisement big data processing model is performed based on the multidimensional advertisement delivery strategy. Therefore, on one hand, on the one hand, the hot advertisement information in the online advertisement content is the information with relatively more putting times, so that the corresponding advertisement crowd data is relatively stable, and in the training process, the delayed use performance of the model is relatively high, thereby improving the reliability of model training. On the other hand, as the training set adopted by the advertisement big data processing model training is a multidimensional advertisement delivery strategy obtained based on a plurality of online advertisement information of online advertisement contents, the model obtained by the training in the above way is used for matching the online multimedia platform of the online advertisement to be delivered, so that the condition that the online advertisement to be delivered is matched with the online multimedia platform mistakenly due to the difference of different online advertisement information can be avoided, the online advertisement to be delivered can be delivered and delivered on a proper online multimedia platform, and the waste of network resources caused by the unmatched advertisement delivery behavior is reduced.

In some optional embodiments, after the step "identifying the online multimedia platform corresponding to the online advertisement to be pushed by using the trained target model", the method may further include mining and analyzing advertisement feedback data of the online advertisement to be pushed, so as to achieve optimization processing of the advertisement content of the online advertisement to be pushed, thereby improving subsequent advertisement pushing efficiency, avoiding that too many communication resources are occupied due to frequent advertisement pushing through the online multimedia platform, and also avoiding that the user end device frequently receives advertisements to affect handling of normal services.

On the basis of the above contents, after the step "identifying the online multimedia platform corresponding to the online advertisement to be pushed by using the trained target model", the method may further include the following steps: and sending the online advertisement to be pushed to a corresponding online multimedia platform, receiving advertisement feedback data uploaded by the online multimedia platform corresponding to the online advertisement to be pushed, mining user intention on the advertisement feedback data to obtain a user intention mining result, and optimizing the advertisement content of the online advertisement to be pushed according to the user intention mining result.

In some optional embodiments, the "mining user intention on the advertisement feedback data to obtain user intention mining results" described in the above steps may include the following:

s51, acquiring a plurality of user behavior portrait information to be mined according to the advertisement feedback data; the user behavior portrait information is access depth information of an advertisement message acquired by a service user side;

s52, respectively acquiring time sequence type intention tendency between each user behavior portrait information and the historical mining portrait information; wherein, the time sequence type intention tendency is used for showing the relativity of the corresponding portrait of every two portrait information in a time sequence level; the historical mining image information is image information which is mined;

s53, based on the obtained time sequence type intention tendency, obtaining a mining result whether each user behavior portrait information and the historical mining portrait information belong to the same advertisement message by using an advertisement group intention mining network obtained by pre-training; the advertisement group intention mining network is a model obtained by utilizing a plurality of samples with classification marks, wherein the samples comprise two sample portrait information, and the classification marks are used for indicating whether the two sample portrait information belong to the same sample advertisement message;

in an alternative embodiment, the plurality of samples with the classification mark further includes: and service interaction information corresponding to the sample portrait information. Before the mining result of whether each user behavior portrait information and the historical mining portrait information belong to the same advertisement message is obtained by using an advertisement group intention mining network obtained through pre-training based on the obtained time sequence type intention tendency, the method may further include: respectively acquiring the service interaction intention tendency of each user behavior portrait information and the historical mining portrait information; the service interaction type intention tendency is used for indicating the correlation of advertisement messages in service interaction information corresponding to every two portrait information respectively on a service interaction layer. The method for acquiring mining results of whether each user behavior portrait information and the historical mining portrait information belong to the same advertisement message or not by utilizing an advertisement group intention mining network obtained by pre-training based on the acquired time sequence type intention tendency comprises the following steps: and inputting the acquired time sequence type intention tendency and service interaction type intention tendency into an advertisement group intention mining network obtained by pre-training, and acquiring a mining result of whether each user behavior portrait information and the historical mining portrait information belong to the same advertisement message.

In an alternative embodiment, the advertisement group intention mining network is trained in advance by adopting the following steps: acquiring access depth information of a plurality of sample advertisement messages collected by a plurality of service clients; taking every two portrait information and every two portrait information in the access depth information of the sample advertisement messages as a sample, and marking the classification mark for the sample; acquiring a time sequence type intention tendency and a service interaction type intention tendency of each sample; and training a preset model by utilizing the time sequence type intention tendency and the service interaction type intention tendency of each sample and the classification mark of each sample to obtain the advertisement group intention mining network.

And S54, based on the mining result, mining the user intention based on the time sequence level of the access depth information belonging to the same advertisement message in the plurality of user behavior image information and the historical mining image information, and obtaining the user intention mining result based on the time sequence level.

Through the method described in the above steps S51-S54, the following beneficial effects can be achieved: the method comprises the steps of firstly determining user behavior portrait information according to advertisement feedback data, secondly determining whether each piece of user behavior portrait information and historical mining portrait information belong to a mining result of the same advertisement message or not based on a time sequence type intention tendency between the obtained user behavior portrait information and the historical mining portrait information, and then mining access depth information of the same advertisement message based on a time sequence level user intention to obtain a user intention mining result. Therefore, whether each user behavior image information and the historical mining image information belong to the mining result of the same advertisement message is determined, so that an effective basis can be provided for subsequent user intention mining, and then the advertisement feedback data is subjected to user intention mining in a time sequence level, so that the accuracy of the user intention mining result can be effectively ensured.

On the basis of the above, please refer to fig. 4, the present invention further provides a block diagram of an online advertisement delivery apparatus 400 combining big data analysis, which includes the following functional modules.

The first information determining module 410 is configured to extract hot advertisement information in online advertisement content of the target online multimedia platform to obtain a first hot advertisement information set.

A second information determining module 420, configured to screen a set number of hot advertisement information from the first hot advertisement information set to obtain a second hot advertisement information set.

And an advertisement strategy generating module 430, configured to obtain advertisement delivery data corresponding to each piece of the popular advertisement information by extracting advertisement crowd data of each piece of the popular advertisement information in the second popular advertisement information set, so as to obtain a multidimensional advertisement delivery strategy.

An advertisement information obtaining module 440, configured to determine, based on the multidimensional advertisement delivery policy, valid advertisement information corresponding to the same target online multimedia platform and invalid advertisement information corresponding to different target online multimedia platforms.

And the network model training module 450 is configured to train a pre-established advertisement big data processing model by using the effective advertisement information and the ineffective advertisement information to obtain a trained target model.

And an advertisement push matching module 460, configured to, when an online advertisement to be pushed is obtained, identify an online multimedia platform corresponding to the online advertisement to be pushed by using the trained target model.

Further, a readable storage medium is provided, on which a program is stored, which when executed by a processor implements the method described above.

It will be understood that the invention is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the invention is limited only by the appended claims.

Claims

1. An online advertisement pushing method combined with big data analysis is applied to a cloud server communicating with a plurality of online multimedia platforms, and the method comprises the following steps:

2. The method of claim 1, wherein the determining effective advertisement information corresponding to the same target online multimedia platform and ineffective advertisement information corresponding to different target online multimedia platforms based on the multidimensional advertisement delivery strategy comprises:

3. The method of claim 2, wherein the determining effective advertisement information corresponding to the same target online multimedia platform and ineffective advertisement information corresponding to different target online multimedia platforms based on the multidimensional advertisement delivery policy comprises: obtaining the first effective advertisement information, the second effective advertisement information and the ineffective advertisement information in advance;

4. The online advertisement push method combined with big data analysis according to claim 2, wherein the effective advertisement information corresponding to the same target online multimedia platform and the ineffective advertisement information corresponding to different target online multimedia platforms are determined based on the multidimensional advertisement delivery strategy; training a pre-established advertisement big data processing model by using the effective advertisement information and the ineffective advertisement information to obtain a trained target model, comprising:

5. The online advertisement push method combining big data analysis according to claim 1, wherein the screening out a set number of the popular advertisement information from the first popular advertisement information set to obtain a second popular advertisement information set comprises:

6. The online advertisement push method combining big data analysis according to claim 1, wherein the screening out a set number of the popular advertisement information from the first popular advertisement information set to obtain a second popular advertisement information set comprises:

7. The method of claim 6, wherein before converting the number of impressions of each advertisement crowd into a second advertisement impression frequency, the method further comprises:

8. The online advertisement push method combining big data analysis according to claim 1, wherein before the step of screening out a set number of the popular advertisement information from the first popular advertisement information set to obtain a second popular advertisement information set, the method further comprises:

9. The online advertisement push method combining big data analysis according to any one of claims 1 to 8, wherein when the online advertisement to be pushed is obtained, the online multimedia platform corresponding to the online advertisement to be pushed is identified by using the trained target model, and the method includes:

10. A cloud server comprising a processor and a memory; the processor is connected in communication with the memory, and the processor is configured to read the computer program from the memory and execute the computer program to implement the method of any one of claims 1 to 9.