Method and device for adjusting resource putting parameter, electronic equipment and storage medium

Abstract

The present disclosure provides a resource putting parameter adjustment method and device, electronic equipment and storage medium, which relates to the technical fields of resource promotion, resource putting and artificial intelligence. The specific implementation scheme is: obtaining the putting word of a target resource, the feature information of at least two dimensions of the target resource, and the historical traffic characteristics of the target resource; based on the putting word of the target resource, the feature information of at least two dimensions of the target resource, the historical traffic characteristics of the target resource and the putting parameter adjustment model, obtaining the adjustment coefficient of the putting parameter of the target resource; based on the adjustment coefficient of the putting parameter of the target resource, adjusting the putting parameter of the target resource. The technology of the present disclosure can effectively improve the accuracy of resource putting parameter adjustment, and in turn can effectively improve the resource putting efficiency.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, specifically to the fields of resource promotion, resource allocation, and artificial intelligence, and particularly to a method, apparatus, electronic device, and storage medium for adjusting resource allocation parameters. Background Technology

[0002] In the internet age, various online resources have flourished thanks to their massive traffic. For example, advertising, as a typical resource, differs from traditional brand advertising aimed at promoting brand image; existing online advertising primarily aims to generate significant conversions. Based on this objective, various advertising tools have emerged to help advertisers personalize their campaigns.

[0003] To more effectively allocate resources, resource allocation platforms or third-party operation platforms can configure specific resource allocation strategies. These strategies are configured based on the predicted performance of the resources, which is determined by the operations personnel based on historical experience and the resource owners' needs. A resource allocation strategy can include multiple parameters, such as bidding, timing, and location. Furthermore, during the allocation process, operations personnel can adjust the resource allocation strategy based on the current performance to achieve better results. Summary of the Invention

[0004] This disclosure provides a method, apparatus, electronic device, and storage medium for adjusting resource deployment parameters.

[0005] According to one aspect of this disclosure, a method for adjusting resource deployment parameters is provided, comprising:

[0006] Obtain the target keywords, at least two dimensions of feature information of the target resource, and the historical traffic features of the target resource;

[0007] Based on the target keywords, at least two dimensions of feature information of the target resource, historical traffic features of the target resource, and the target parameter adjustment model, the adjustment coefficient of the target resource's target parameters is obtained;

[0008] The deployment parameters of the target resource are adjusted based on the adjustment coefficient of the deployment parameters of the target resource.

[0009] According to another aspect of this disclosure, an apparatus for adjusting resource deployment parameters is provided, comprising:

[0010] The feature acquisition module is used to acquire the target resource's target keywords, at least two dimensions of feature information of the target resource, and the historical traffic features of the target resource;

[0011] The coefficient acquisition module is used to acquire the adjustment coefficient of the delivery parameters of the target resource based on the target resource's delivery keywords, at least two dimensions of feature information of the target resource, the historical traffic characteristics of the target resource, and the delivery parameter adjustment model.

[0012] The adjustment module is used to adjust the deployment parameters of the target resource based on the adjustment coefficient of the deployment parameters of the target resource.

[0013] According to another aspect of this disclosure, an electronic device is provided, comprising:

[0014] At least one processor; and

[0015] A memory communicatively connected to the at least one processor; wherein,

[0016] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the methods described above and any possible implementations.

[0017] According to another aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions for causing the computer to perform the methods described above and any possible implementation thereof.

[0018] According to another aspect of this disclosure, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the aspects and any possible implementations described above.

[0019] According to the technology disclosed herein, the accuracy of resource allocation parameter adjustments can be effectively improved, thereby increasing resource allocation efficiency.

[0020] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0021] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0022] Figure 1 This is a schematic diagram based on the first embodiment of the present disclosure;

[0023] Figure 2This is a schematic diagram according to the second embodiment of the present disclosure;

[0024] Figure 3 This is a schematic diagram illustrating the working principle of the feature encoding module in this disclosure.

[0025] Figure 4 This is a schematic diagram according to the third embodiment of the present disclosure;

[0026] Figure 5 This is a schematic diagram according to the fourth embodiment of the present disclosure;

[0027] Figure 6 This is a block diagram of an electronic device used to implement the methods of the embodiments of this disclosure. Detailed Implementation

[0028] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0029] Obviously, the described embodiments are only some, not all, of the embodiments disclosed herein. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without inventive effort are within the scope of protection of this disclosure.

[0030] It should be noted that the terminal devices involved in the embodiments of this disclosure may include, but are not limited to, smart devices such as mobile phones, personal digital assistants (PDAs), wireless handheld devices, and tablet computers; the display devices may include, but are not limited to, personal computers, televisions, and other devices with display functions.

[0031] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0032] Figure 1 This is a schematic diagram based on the first embodiment of the present disclosure; as shown Figure 1 As shown, this embodiment provides a method for adjusting resource deployment parameters, which can be applied in a resource deployment platform or a resource operation platform. It allows for real-time adjustment of resource deployment parameters and specifically includes the following steps:

[0033] S101. Obtain the target keywords, at least two dimensions of feature information of the target resource, and the historical traffic features of the target resource;

[0034] In this embodiment, the target resource can be any resource whose deployment parameters need to be adjusted.

[0035] The target resource in this embodiment can be any resource that can be delivered online, such as advertisements, or news from various sections such as entertainment, sports, and finance.

[0036] In this embodiment, the target resource's keywords can be the resource's keywords or thematic terms. During resource promotion, the resource promotion platform can more effectively promote the target resource based on these keywords. For example, if the target resource is an advertisement, the advertisement can be placed and promoted based on these keywords. Furthermore, advertisers can also bid on advertisements based on these keywords.

[0037] The target resource in this embodiment may also include feature information in at least two dimensions, such as text and image dimensions; it may also include video and text dimensions, where the video itself includes a series of images, and the processing method can be the same as for the image dimension. It may also include audio and image dimensions, where audio can be converted into text through speech recognition and processed in the same way as the text dimension. Of course, it may also include three or more of the above-mentioned dimensions, which will not be elaborated further here.

[0038] The historical traffic characteristics of the target resource in this embodiment refer to the traffic characteristics of the target resource within a preset time period before the current moment. The traffic characteristics within the preset time period may include at least one of the following: the number of impressions within the preset time period, the historical return on investment (ROI) within the preset time period, the historical ranking within the preset time period, and the number of clicks within the preset time period. The historical ranking within the preset time period can be based on the average of all historical rankings, or it can be the ranking with the most historical impressions, or it can be determined using other mathematical calculations; no limitation is made here. The preset time period can be set according to needs or experience.

[0039] S102. Based on the target resource's keywords, at least two dimensions of the target resource's feature information, the target resource's historical traffic characteristics, and the target resource's target parameter adjustment model, obtain the adjustment coefficient of the target resource's target resource's target parameter.

[0040] The delivery parameter adjustment model in this embodiment is an end-to-end neural network model. In use, the target keywords, at least two dimensions of feature information of the target resource, and historical traffic characteristics of the target resource are input into the delivery parameter adjustment model. Based on the input information, the model can predict the adjustment coefficient of the delivery parameters for the target resource.

[0041] S103. Adjust the deployment parameters of the target resources based on the adjustment coefficient of the deployment parameters of the target resources.

[0042] The resource delivery parameter adjustment method in this embodiment employs a delivery parameter adjustment model. Based on the target resource's keywords, at least two dimensions of feature information, and historical traffic characteristics, it adjusts the target resource's delivery parameters in real time. This overcomes the technical problem in existing technologies where manual configuration and adjustment of resource delivery parameters by operators leads to poor accuracy. Therefore, the technical solution in this embodiment effectively improves the accuracy of resource delivery parameter adjustments, thereby significantly enhancing resource delivery performance.

[0043] Figure 2 This is a schematic diagram based on the second embodiment of the present disclosure; as shown Figure 2 As shown, the method for adjusting resource deployment parameters in this embodiment is based on the above... Figure 1 Based on the technical solutions of the illustrated embodiments, the technical solutions of this disclosure will be further described in more detail. For example... Figure 2 As shown, the method for adjusting resource deployment parameters in this embodiment may specifically include the following steps:

[0044] S201. Obtain the target keywords, at least two dimensions of feature information of the target resource, and the historical traffic features of the target resource;

[0045] The feature information in this embodiment, which has at least two dimensions, can be exemplified by including image-level feature information and text-level feature information.

[0046] For example, if the target resource is an advertisement, the corresponding advertisement should be an image-text ad. The text-level feature information can include the ad's theme. If the advertiser hasn't specified an ad theme, a phrase including the target keywords can be obtained from the ad summary as the ad theme. The image-level feature information can refer to the key image specified by the advertiser. If the advertiser hasn't specified one, a cover image can be selected. Alternatively, if no cover image is specified, content recognition can be performed on multiple images included in the ad content to obtain the image that best matches the ad's target keywords, serving as the ad's image-level feature information.

[0047] S202. Using the feature encoding module in the delivery parameter adjustment model, based on the delivery words of the target resource, the feature information of the target resource in at least two dimensions, and the historical traffic characteristics of the target resource, the feature expression of the target resource is obtained.

[0048] S203. Using the coefficient adjustment module in the deployment parameter adjustment model, based on the feature expression of the target resource, predict the adjustment coefficient of the deployment parameter of the target resource;

[0049] For example, if the resource is an advertisement, the resource delivery parameter in this embodiment can be the bid for the advertisement.

[0050] The parameter adjustment model in this embodiment can be divided into two main parts: a feature encoding module and a coefficient adjustment module. Each of the feature encoding module and the coefficient adjustment module is also a neural network model.

[0051] For example, in one embodiment of this disclosure, step S201 may specifically include the following steps:

[0052] (1) The historical traffic characteristics of the target resource are encoded using a feature encoding module to obtain the first feature expression;

[0053] For example, before this step, the historical traffic characteristics of the target resource can be binned to avoid an excessively large range of feature distribution, making the feature distribution more balanced and reasonable, thereby effectively improving the accuracy of feature representation.

[0054] For example, some resources might have single-digit impressions or clicks, while others might have millions, a significant difference that can lead to inaccurate feature representation. Therefore, features can be binned, distributing all feature values ​​within a reasonable range. In practical applications, this could be done by dividing the values ​​into bins of 1-10. Each bin corresponds to a range of impressions or clicks, and the percentage of each bin's range can be the same or different. The value of the corresponding feature is then represented by the identifier of its corresponding bin.

[0055] When there are multiple historical traffic features, the first encoding unit can encode each historical traffic feature of the target resource to obtain the corresponding feature representation. The process of obtaining this feature representation can be understood as a vector representation process. Finally, the feature representations of all historical traffic features are concatenated to flatten all historical traffic feature representations into a one-hot vector, which is the first feature representation of the target resource.

[0056] (2) Using a feature encoding module, the text dimension feature information in the target resource's delivery words and at least two dimensions of feature information is encoded using a self-attention mechanism to obtain the first self-attention feature expression;

[0057] (3) Using a feature encoding module, the image dimension feature information in at least two dimensions of feature information is encoded by a self-attention mechanism to obtain the second self-attention feature expression;

[0058] (4) Using a feature encoding module, the cross-attention mechanism is used to encode the first self-attention feature expression and the second self-attention feature expression to obtain the second feature expression corresponding to the text dimension and the third feature expression corresponding to the image dimension, respectively.

[0059] (5) Using a feature encoding module, based on the first feature expression, the second feature expression and the third feature expression, the feature expression of the target resource is obtained.

[0060] For example, Figure 3 This is a schematic diagram illustrating the working principle of the feature encoding module in this disclosure. For example... Figure 3 As shown, the leftmost part illustrates the process of the feature encoding module encoding the historical traffic features of the target resource. For example, historical traffic features can include ROI, RANK, etc., for a preset time period prior to the current moment. By encoding the aforementioned historical traffic features such as ROI and RANK of the target resource separately, feature representations of each historical traffic feature are obtained and concatenated to obtain the first feature representation of the target resource. The middle part illustrates the process of the feature encoding module encoding the text-dimensional feature information. In this encoding process, the input text-dimensional feature information includes the concatenation of the target words and the title. First, a Long Short-Term Memory (LSTM) network model is used to represent the input text feature information as a vector. Then, a self-attention mechanism is applied to obtain the first self-attention feature representation. Encoding through the self-attention mechanism allows the first self-attention feature representation to notice the correlation between the target words and the title, making the feature representation more accurate.

[0061] The right side illustrates the process by which the feature encoding module encodes the image's dimensional features. In this encoding process, the input is an image. First, a Convolutional Neural Network (CNN) is used to represent the image's features. Then, self-attention encoding is performed to obtain a second self-attention feature representation. This self-attention mechanism allows the second self-attention feature representation to consider the correlations between information at different locations within the image, resulting in more accurate feature representation.

[0062] Next, in the cross-attention stage, the first self-attention feature representation of the text dimension and the second self-attention feature representation of the image dimension are encoded using a cross-attention mechanism. This allows the features of the text dimension to notice the features of the image dimension, and vice versa, thereby making the second and third feature representations of the target resource more accurate.

[0063] Finally, the first, second, and third feature representations of the target resource are concatenated to obtain the feature representation of the target resource. The encoding methods of the Self-Attention and Cross-Attention mechanisms used can be found in existing technologies and will not be elaborated upon here.

[0064] The above methods can accurately and efficiently obtain the feature representation of the target resource.

[0065] The coefficient adjustment module in this embodiment can adopt the structure of a deep neural network (DNN) model.

[0066] When in use, the feature representation of the target resource obtained above is input into the coefficient adjustment module. The coefficient adjustment module can predict and output the adjustment coefficient of the deployment parameters of the target resource based on the input information.

[0067] By employing the feature encoding module and coefficient adjustment module in the deployment parameter adjustment model, the adjustment coefficients of the deployment parameters of the target resource can be obtained efficiently and accurately.

[0068] S204. Adjust the deployment parameters of the target resources based on the adjustment coefficient of the deployment parameters of the target resources;

[0069] For example, in this embodiment, the adjustment coefficient can be in the range of [0, 2]. When using it, the target resource's delivery parameter is multiplied by the adjustment coefficient of the delivery parameter to obtain the adjusted delivery parameter.

[0070] S205. Collect the traffic characteristics of each reference resource within a preset time period after the deployment parameters of each reference resource are adjusted.

[0071] The deployment parameters for multiple reference resources were adjusted using steps S201-S205 described above. Furthermore, all reference resources used the same deployment parameter adjustment model; that is, the model parameters were identical.

[0072] To facilitate use in real-world applications, the adjustment times of the delivery parameters for multiple reference resources can all fall within a preset time window. In this case, the target resource can be included among the multiple reference resources. Then, the traffic characteristics of each reference resource within a preset time period after the delivery parameter adjustment are taken from the multiple reference resources corresponding to this preset time window. These traffic characteristics can include those included in the historical traffic characteristics mentioned above, and of course, can also include other traffic characteristics, which will not be elaborated upon here.

[0073] Since it takes time to evaluate the effect of adjusting resource deployment parameters, in this embodiment, the aforementioned traffic characteristics can only be collected after a preset time interval following the parameter adjustment. The preset time interval can be set according to the specific resource deployment situation, combined with experience or needs; for example, it can be 1 day, 1 week, or other durations, and is not limited here.

[0074] S206. Based on the traffic characteristics of each reference resource within a preset time period, obtain the values ​​of the consideration parameters for each reference resource.

[0075] The parameters considered in this embodiment are those relevant to the campaign objective. For example, if the campaign objective is to maximize revenue, the corresponding parameter could be revenue. If the campaign objective is to maximize clicks, the corresponding parameter could be clicks. If the campaign objective is to maximize impressions, the corresponding parameter could be impressions. If the campaign objective is to minimize conversion costs, the corresponding parameter could be conversion costs.

[0076] The values ​​of the consideration parameters for each reference resource can be obtained directly based on the traffic characteristics of the corresponding reference resource within a preset time period, or they can be calculated by combining the basic information of the reference resource. For example, if the consideration parameter is the number of impressions, it can be obtained directly from the traffic characteristics within the preset time period. If the consideration parameter is revenue, the revenue and cost of the reference resource within the preset time period can be obtained based on the above information, and the revenue is obtained by subtracting the cost from the revenue.

[0077] S207. Taking the deployment targets of multiple reference resources as the objective, and based on the values ​​of the consideration parameters of each reference resource, adjust the parameters of the deployment parameter adjustment model.

[0078] In this embodiment, the deployment targets of multiple reference resources must be the same.

[0079] For example, in practice, the value of the global consideration parameter can be obtained first based on the values ​​of the consideration parameters of each reference resource among multiple reference resources; then, with the delivery target as the objective, the parameters of the delivery parameter adjustment model can be adjusted based on the value of the global consideration parameter.

[0080] The global consideration parameter can be the average of the values ​​of multiple reference resources. Alternatively, it can be the average of the maximum and minimum values, or it can be obtained through other mathematical calculations. No specific limitation is imposed here.

[0081] Specifically, the parameters of the deployment parameter adjustment model can be adjusted using an evolutionary algorithm based on the values ​​of globally considered parameters, with the deployment target as the objective. For example, the evolutionary algorithm of OpenAI's Evolution Strategies (ES) can be used to adjust the parameters of the deployment parameter adjustment model.

[0082] Using the above method, the parameters of the delivery parameter adjustment model can be adjusted accurately and efficiently based on the values ​​of the reference resource consideration parameters, making the delivery parameter adjustment model more and more accurate during use.

[0083] Optionally, in one embodiment of this disclosure, for any resource, after the parameter adjustment model predicts the adjustment coefficient of the resource in step S203 and before step S204, soft constraint processing can be performed to further improve the accuracy of the adjustment coefficient.

[0084] In one embodiment of this disclosure, soft constraint processing can be performed by referring to the historical ROI of the resource. If the historical ROI of the resource within a preset time period before the current moment is greater than or equal to a preset ROI threshold, it is considered a high ROI; otherwise, it is considered a low ROI. The preset ROI threshold can be set based on experience or requirements. For high ROI resources, the adjustment coefficient can be constrained to between 0.9 and 2 during soft constraint processing. For low ROI resources, the adjustment coefficient can be constrained to between 0 and 1.1. For example, the adjustment coefficient after soft constraint processing for high ROI can be equal to min(max(adjustment coefficient predicted by the deployment parameter adjustment model, 0.9), 2); the adjustment coefficient after soft constraint processing for low ROI can be equal to min(adjustment coefficient predicted by the deployment parameter adjustment model, 1.1).

[0085] For example, in a scenario where the resource is advertising, the investment parameter is the advertising bid, and the adjustment coefficient is the advertising bid coefficient, the ad delivery parameter adjustment model predicts an ad bidding frequency of 1.8. Under soft constraint handling, if the historical ROI of the ad is low, the adjustment frequency needs to be constrained to between 0 and 1.1, so the bidding frequency can be min(1.8, 1.1), which equals 1.1. Conversely, if the historical ROI of the ad is high, the adjustment frequency needs to be constrained to between 0.9 and 2, so the bidding frequency can be min(max(1.8, 0.9), 2), which equals 1.8.

[0086] The following example uses resources as advertisements and the goal of maximizing revenue to explain in detail the principle of parameter adjustment in the advertising parameter adjustment model.

[0087] Specifically, the model's parameters can be adjusted by updating the delivery parameters based on the reward, aligning them with the goal of maximizing the reward and thus allowing the model to obtain more rewards. This reward is represented by the value of the parameter corresponding to the delivery objective. In this scenario, the reward is the revenue.

[0088] During modeling, the observation (state) can be represented by the features of the target resource. By inputting the feature representation of the target resource into the coefficient adjustment module of the DNN structure, the bid adjustment coefficient (rate) can be determined. Then, the reward is calculated, for example, the reward can be calculated using the following formula: As a hyperparameter, cash represents revenue and cost of conversion, which can be used to define the relationship between revenue and conversion cost. Based on this approach, the model can be guided to dynamically adjust its learning objectives and update its parameters according to the campaign goals.

[0089] Steps S205-S207 above adjust the parameters of the delivery parameter adjustment model based on the values ​​of consideration parameters for multiple reference resources. In practical applications, the parameters of the delivery parameter adjustment model can also be adjusted based on the values ​​of consideration parameters for each resource, which increases the number of adjustments and reduces efficiency compared to adjusting based on multiple reference resources in the above scenario.

[0090] For example, after step S204, the following steps may also be included:

[0091] (a) Collect traffic characteristics of the target resource within a preset time period after the deployment parameters are adjusted;

[0092] (b) Based on the traffic characteristics of the target resource within a preset time period, obtain the values ​​of the consideration parameters of the target resource;

[0093] (c) Based on the target resource delivery target, adjust the parameters of the delivery parameter adjustment model according to the value of the consideration parameters.

[0094] The adjustment principle is the same as that of steps S205-S207 above. Using steps (a)-(c), the parameters of the deployment parameter adjustment model can also be accurately adjusted.

[0095] As described in the above embodiments, the deployment parameter adjustment model needs to be adjusted continuously during online use.

[0096] Optionally, in practical applications, after step S206 and before step S207, the process may further include: detecting whether the delivery parameter adjustment model meets the preset requirements based on the traffic characteristics of each reference resource within a preset time period and a preset evaluation threshold; if not, proceeding to step S207. If the requirements are met, step S207 can be omitted, indicating that the delivery parameter adjustment model is mature and the predictions are accurate, and the parameters no longer need adjustment. The delivery target may change, or the usage environment may change, at which point the delivery parameter adjustment model can be adjusted again according to the method described in the above embodiment.

[0097] The preset evaluation threshold can be determined based on the campaign objective. For example, if the objective is to maximize revenue, the evaluation threshold could be an ROI of 1.3. If the objective is to maximize clicks, the corresponding evaluation threshold could be a preset time period followed by a preset number of clicks. This preset number of clicks can be set based on actual needs or experience. The preset requirements can specify that either a value greater than the preset evaluation threshold or a value less than the preset evaluation threshold is acceptable.

[0098] The parameters corresponding to the evaluation threshold, such as ROI or click count mentioned above, can also be obtained based on the traffic characteristics of each reference resource within a preset time period. Specifically, based on the traffic characteristics of each reference resource within a preset time period, the values ​​of the parameters corresponding to the evaluation threshold for each reference resource within that preset time period can be obtained. Then, referring to the calculation process of the global consideration parameters, a global value for the parameter corresponding to an evaluation threshold can be obtained. This global value is compared with the preset evaluation threshold to determine whether the delivery parameter adjustment model meets the preset requirements.

[0099] Optionally, in the scenario where the parameters of the delivery parameter adjustment model are adjusted based on the traffic characteristics of a single resource in steps (a)-(c) above, the above method can also be used to detect and evaluate whether the delivery parameter adjustment model meets the preset requirements.

[0100] Based on the above, it can be seen that the delivery parameter adjustment model in this embodiment can be used in a cold start mode when deployed online, adjusting as it is used, and becoming more accurate the longer it is used. Of course, the delivery parameter adjustment model can be continuously adjusted during use. Alternatively, it can be stopped once the preset requirements are met, reducing processes and improving efficiency. However, if the environment or delivery target changes, the parameters of the delivery parameter adjustment model need to be adjusted again. For example, in an insurance advertising scenario, the delivery parameter adjustment model has already met the preset requirements and no further parameter adjustments are needed. However, if the delivery parameter adjustment model is directly applied to a food advertising scenario, it will still need to be adjusted continuously; that is, the delivery parameter adjustment model needs to continue learning the characteristics of food data to more accurately adjust the resource delivery parameters.

[0101] For example, in advertising application scenarios, existing advertising management platforms cannot quantify bidding decision factors for advertising parameters such as bid, leading to inaccurate bidding and consequently poor advertising performance. By adopting the technical solution of the above embodiments of this disclosure, a delivery parameter adjustment model is designed to encode the feature information of advertisements in the bidding environment, obtaining the feature representation of the advertisement. This requires considering not only traffic indicators such as keywords, historical ROI, and ad display position, but also textual features such as the relevance between keywords and ad titles, as well as visual features such as ad images. Therefore, this disclosure uses a multimodal fusion approach in the feature encoding module of the delivery parameter adjustment model to represent the bidding environment of the advertisement, obtaining a highly accurate feature representation of the advertisement. Furthermore, based on an evolutionary algorithm, a reward adjustment model is used to adjust the parameters, causing the model parameters to adjust towards higher rewards. This makes the delivery parameter adjustment model increasingly accurate, and the predicted adjustment coefficients of the delivery parameters also become more accurate, thereby improving the efficiency of advertising delivery in the bidding environment. For example, experimental verification shows that in advertising application scenarios, after adjusting the bids for advertising using the solution disclosed herein, revenue can be increased by 33.7% and ROI can be increased by more than 9.5%.

[0102] The resource deployment parameter adjustment method in this embodiment can not only accurately and efficiently adjust the resource deployment parameters, but also adjust the parameters of the deployment parameter adjustment model, making the deployment parameter adjustment model more accurate the more it is used, effectively improving the accuracy of the deployment parameter adjustment model, thereby further improving the accuracy of resource deployment parameter adjustment and effectively improving the efficiency of resource deployment.

[0103] Figure 4 This is a schematic diagram based on the third embodiment of this disclosure; as shown Figure 4As shown, this embodiment provides a resource deployment parameter adjustment device 400, including:

[0104] The feature acquisition module 401 is used to acquire the target resource's target keywords, at least two dimensions of feature information of the target resource, and the historical traffic features of the target resource;

[0105] The coefficient acquisition module 402 is used to acquire the adjustment coefficient of the delivery parameters of the target resource based on the delivery keywords of the target resource, the feature information of the target resource in at least two dimensions, the historical traffic features of the target resource, and the delivery parameter adjustment model.

[0106] The adjustment module 403 is used to adjust the deployment parameters of the target resource based on the adjustment coefficient of the deployment parameters of the target resource.

[0107] The resource deployment parameter adjustment device 400 in this embodiment achieves the same implementation principle and technical effect as the above-mentioned related method embodiments by using the above-mentioned modules. For details, please refer to the description of the above-mentioned related method embodiments, which will not be repeated here.

[0108] Figure 5 This is a schematic diagram based on the fourth embodiment of the present disclosure; as shown Figure 5 As shown, this embodiment provides a resource deployment parameter adjustment device 500, including a... Figure 4 The modules with the same name and function shown are: feature acquisition module 501, coefficient acquisition module 502, and adjustment module 503.

[0109] like Figure 5 As shown, in the resource deployment parameter adjustment device 500 of this embodiment, the coefficient acquisition module 502 includes:

[0110] The feature expression acquisition unit 5021 is used to acquire the feature expression of the target resource based on the target resource's target words, at least two dimensions of the target resource's feature information, and the target resource's historical traffic features, using the feature encoding module in the delivery parameter adjustment model.

[0111] The coefficient prediction unit 5022 is used to predict the adjustment coefficient of the deployment parameters of the target resource based on the feature expression of the target resource by employing the coefficient adjustment module in the deployment parameter adjustment model.

[0112] Furthermore, in one embodiment of this disclosure, the feature representation acquisition unit 5021 is used for:

[0113] The feature encoding module is used to encode the historical traffic features of the target resource to obtain a first feature representation;

[0114] The feature encoding module is used to encode the target resource's delivery words and the text dimension feature information in the at least two dimensions of feature information using a self-attention mechanism to obtain a first self-attention feature expression;

[0115] The feature encoding module is used to encode the image dimension feature information in the at least two dimensions of feature information using a self-attention mechanism to obtain a second self-attention feature representation.

[0116] Using the feature encoding module, cross-attention mechanism encoding is performed based on the first self-attention feature expression and the second self-attention feature expression to obtain the second feature expression corresponding to the text dimension and the third feature expression corresponding to the image dimension, respectively;

[0117] The feature encoding module is used to obtain the feature expression of the target resource based on the first feature expression, the second feature expression, and the third feature expression.

[0118] Furthermore, in one embodiment of this disclosure, the feature representation acquisition unit 5021 is also used for:

[0119] The historical traffic characteristics of the target resource are divided into buckets.

[0120] Furthermore, such as Figure 5 As shown, in one embodiment of this disclosure, the resource deployment parameter adjustment device 500 further includes:

[0121] The first acquisition module 504 is used to acquire the traffic characteristics of the target resource within a preset time period after the delivery parameters are adjusted.

[0122] The first acquisition module 505 is used to acquire the value of the consideration parameters of the target resource based on the traffic characteristics of the target resource within the preset time length; the consideration parameters are the parameters to be considered for the target resource's deployment.

[0123] The adjustment module 503 is also used to adjust the parameters of the delivery parameter adjustment model based on the values ​​of the consideration parameters, with the delivery target of the target resource as the objective.

[0124] Alternatively, such as Figure 5 As shown, in one embodiment of this disclosure, the resource deployment parameter adjustment device 500 further includes:

[0125] The second acquisition module 506 is used to acquire the traffic characteristics of each reference resource within a preset time period after the delivery parameters of the multiple reference resources are adjusted; the delivery parameter adjustment times of the multiple reference resources are all within a preset time window; the multiple reference resources include the target resource;

[0126] The second acquisition module 507 is used to acquire the values ​​of the consideration parameters of each of the multiple reference resources based on the traffic characteristics of each of the reference resources within the preset time length; the consideration parameters are the parameters to be considered for the deployment target of the target resource;

[0127] The adjustment module 503 is further configured to adjust the parameters of the deployment parameter adjustment model based on the values ​​of the consideration parameters of each of the multiple reference resources, with the deployment target of the multiple reference resources as the target; the multiple reference resources have the same deployment target.

[0128] Furthermore, in one embodiment of this disclosure, the adjustment module 503 is configured to:

[0129] Based on the values ​​of the consideration parameters of each of the plurality of reference resources, the value of the global consideration parameter is obtained;

[0130] With the target of the deployment as the objective, the parameters of the deployment parameter adjustment model are adjusted based on the values ​​of the global consideration parameters.

[0131] Furthermore, in one embodiment of this disclosure, the adjustment module 503 is configured to:

[0132] Using the aforementioned delivery target as the objective and based on the values ​​of the global consideration parameters, an evolutionary algorithm is employed to adjust the parameters of the delivery parameter adjustment model.

[0133] Furthermore, such as Figure 5 As shown, in one embodiment of this disclosure, the resource deployment parameter adjustment device 500 further includes:

[0134] The detection module 508 is used to detect and determine that the delivery parameter adjustment model has not met the preset requirements based on the traffic characteristics of each of the plurality of reference resources within the preset time length and the preset evaluation threshold.

[0135] Furthermore, such as Figure 5 As shown, in one embodiment of this disclosure, the resource deployment parameter adjustment device 500 further includes:

[0136] The soft constraint processing module 509 is used to perform soft constraint processing on the adjustment coefficients of the deployment parameters of the target resource. For details, please refer to the relevant descriptions in the above method embodiments.

[0137] The resource deployment parameter adjustment device 500 in this embodiment achieves the same implementation principle and technical effect as the above-mentioned related method embodiments by using the above-mentioned modules. For details, please refer to the description of the above-mentioned related method embodiments, which will not be repeated here.

[0138] The acquisition, storage, and application of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0139] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0140] Figure 6 A schematic block diagram of an example electronic device 600 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0141] like Figure 6 As shown, device 600 includes a computing unit 601, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 602 or a computer program loaded from storage unit 608 into random access memory (RAM) 603. RAM 603 may also store various programs and data required for the operation of device 600. The computing unit 601, ROM 602, and RAM 603 are interconnected via bus 604. Input / output (I / O) interface 605 is also connected to bus 604.

[0142] Multiple components in device 600 are connected to I / O interface 605, including: input unit 606, such as keyboard, mouse, etc.; output unit 607, such as various types of monitors, speakers, etc.; storage unit 608, such as disk, optical disk, etc.; and communication unit 609, such as network card, modem, wireless transceiver, etc. Communication unit 609 allows device 600 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0143] The computing unit 601 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above, such as the methods of this disclosure. For example, in some embodiments, the methods of this disclosure may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 608. In some embodiments, part or all of the computer program may be loaded and / or installed on device 600 via ROM 602 and / or communication unit 609. When the computer program is loaded into RAM 603 and executed by the computing unit 601, one or more steps of the methods of this disclosure described above may be performed. Alternatively, in other embodiments, the computing unit 601 may be configured to perform the methods of this disclosure by any other suitable means (e.g., by means of firmware).

[0144] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0145] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0146] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0147] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0148] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0149] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0150] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0151] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for adjusting resource deployment parameters, comprising: Obtain the target keywords, at least two dimensions of feature information of the target resource, and the historical traffic features of the target resource; The at least two dimensions include text and images; Based on the target resource's keywords, at least two dimensions of the target resource's feature information, the target resource's historical traffic features, and the target resource's target parameter adjustment model, the adjustment coefficient of the target resource's target resource's target parameter is obtained; the target resource's historical traffic features refer to the target resource's traffic features within a preset time period before the current moment. Based on the adjustment coefficient of the target resource's deployment parameters, the deployment parameters of the target resource are adjusted; Based on the target resource's keywords, at least two dimensions of feature information of the target resource, historical traffic characteristics of the target resource, and a targeting parameter adjustment model, the adjustment coefficients of the targeting resource's targeting parameters are obtained, including: The feature encoding module in the delivery parameter adjustment model is used to encode the combination information of the delivery words of the target resource and the text dimension feature information in the at least two dimensions of feature information, as well as the image dimension feature information in the at least two dimensions of feature information, using a self-attention mechanism to obtain the first self-attention feature expression and the second self-attention feature expression. Using the feature encoding module, cross-attention mechanism encoding is performed based on the first self-attention feature expression and the second self-attention feature expression to obtain the second feature expression corresponding to the text dimension and the third feature expression corresponding to the image dimension, respectively; Using the feature encoding module, the feature expression of the target resource is obtained by encoding the first feature expression, the second feature expression, and the third feature expression based on the historical traffic features of the target resource.

2. The method according to claim 1, wherein, Based on the target resource's keywords, at least two dimensions of feature information of the target resource, historical traffic characteristics of the target resource, and a targeting parameter adjustment model, the adjustment coefficient of the targeting resource's targeting parameters is obtained, further including: The coefficient adjustment module in the aforementioned deployment parameter adjustment model is used to predict the adjustment coefficient of the deployment parameters of the target resource based on the feature representation of the target resource.

3. The method according to claim 2, wherein, Before obtaining the feature representation of the target resource using the feature encoding module based on the first feature representation, the second feature representation, and the third feature representation obtained by encoding the historical traffic features of the target resource, the method further includes: The feature encoding module is used to encode the historical traffic features of the target resource to obtain a first feature representation.

4. The method according to claim 3, before encoding the historical traffic features of the target resource using the feature encoding module to obtain the first feature expression, the method further includes: The historical traffic characteristics of the target resource are divided into buckets.

5. The method according to any one of claims 1-4, wherein, After adjusting the deployment parameters of the target resource based on the adjustment coefficient of the deployment parameters of the target resource, the method further includes: Collect traffic characteristics of the target resource within a preset time period after the deployment parameters are adjusted; Based on the traffic characteristics of the target resource within the preset time period, the values ​​of the consideration parameters of the target resource are obtained; the consideration parameters are the parameters to be considered for the target resource's deployment objectives. With the target resource deployment objective as the objective, the parameters of the deployment parameter adjustment model are adjusted based on the values ​​of the consideration parameters.

6. The method according to any one of claims 1-4, wherein, After adjusting the deployment parameters of the target resource based on the adjustment coefficient of the deployment parameters of the target resource, the method further includes: The traffic characteristics of each reference resource within a preset time period after the delivery parameters of the multiple reference resources are adjusted are collected; the delivery parameter adjustment times of the multiple reference resources are all within a preset time window; the multiple reference resources include the target resource; Based on the traffic characteristics of each of the multiple reference resources within the preset time length, the values ​​of the consideration parameters of each reference resource are obtained; the consideration parameters are the parameters to be considered for the target resource's deployment objective. With the target of the multiple reference resources as the objective, the parameters of the deployment parameter adjustment model are adjusted based on the values ​​of the consideration parameters of each of the multiple reference resources; the deployment targets of the multiple reference resources are the same.

7. The method according to claim 6, wherein, Taking the deployment target of the plurality of reference resources as the objective, and based on the values ​​of the consideration parameters of each of the plurality of reference resources, the parameters of the deployment parameter adjustment model are adjusted, including: Based on the values ​​of the consideration parameters of each of the plurality of reference resources, the value of the global consideration parameter is obtained; With the target of the deployment as the objective, the parameters of the deployment parameter adjustment model are adjusted based on the values ​​of the global consideration parameters.

8. The method according to claim 7, wherein, Using the aforementioned delivery target as the objective, and based on the values ​​of the global consideration parameters, the parameters of the delivery parameter adjustment model are adjusted, including: Using the aforementioned delivery target as the objective and based on the values ​​of the global consideration parameters, an evolutionary algorithm is employed to adjust the parameters of the delivery parameter adjustment model.

9. The method according to claim 6, wherein, Before adjusting the parameters of the deployment parameter adjustment model based on the values ​​of the consideration parameters of each of the plurality of reference resources, with the deployment target as the objective, the method further includes: Based on the traffic characteristics of each of the multiple reference resources within the preset time length and the preset evaluation threshold, it is detected and determined that the delivery parameter adjustment model has not met the preset requirements.

10. The method according to any one of claims 1-4 and 7-9, wherein, After obtaining the adjustment coefficients of the target resource's delivery parameters based on the target resource's keywords, at least two dimensions of feature information of the target resource, historical traffic characteristics of the target resource, and a delivery parameter adjustment model, and before adjusting the target resource's delivery parameters based on the adjustment coefficients of the target resource's delivery parameters, the method further includes: The adjustment coefficients of the deployment parameters of the target resource are subject to soft constraints.

11. A device for adjusting resource deployment parameters, comprising: The feature acquisition module is used to acquire the target resource's target keywords, at least two dimensions of feature information of the target resource, and the historical traffic features of the target resource; The at least two dimensions include text and images; The coefficient acquisition module is used to acquire the adjustment coefficient of the target resource's delivery parameters based on the target resource's delivery keywords, at least two dimensions of the target resource's feature information, the target resource's historical traffic features, and the delivery parameter adjustment model; the target resource's historical traffic features refer to the target resource's traffic features within a preset time period before the current moment; The adjustment module is used to adjust the deployment parameters of the target resource based on the adjustment coefficient of the deployment parameters of the target resource; The coefficient acquisition module includes: a feature representation acquisition unit, used for: The feature encoding module in the delivery parameter adjustment model is used to encode the combination information of the delivery words of the target resource and the text dimension feature information in the at least two dimensions of feature information, as well as the image dimension feature information in the at least two dimensions of feature information, using a self-attention mechanism to obtain the first self-attention feature expression and the second self-attention feature expression. Using the feature encoding module, cross-attention mechanism encoding is performed based on the first self-attention feature expression and the second self-attention feature expression to obtain the second feature expression corresponding to the text dimension and the third feature expression corresponding to the image dimension, respectively; Using the feature encoding module, the feature expression of the target resource is obtained by encoding the first feature expression, the second feature expression, and the third feature expression based on the historical traffic features of the target resource.

12. The apparatus according to claim 11, wherein, The coefficient acquisition module further includes: The coefficient prediction unit is used to predict the adjustment coefficient of the deployment parameters of the target resource based on the feature expression of the target resource, using the coefficient adjustment module in the deployment parameter adjustment model.

13. The apparatus according to claim 12, wherein, The feature representation acquisition unit is further configured to: The feature encoding module is used to encode the historical traffic features of the target resource to obtain a first feature representation.

14. The apparatus according to claim 12, wherein the feature representation acquisition unit is further configured to: The historical traffic characteristics of the target resource are divided into buckets.

15. The apparatus according to any one of claims 11-14, wherein, The device further includes: The first acquisition module is used to acquire the traffic characteristics of the target resource within a preset time period after the deployment parameters are adjusted. The first acquisition module is used to acquire the values ​​of the consideration parameters of the target resource based on the traffic characteristics of the target resource within the preset time length; the consideration parameters are the parameters to be considered for the target resource's deployment objectives. The adjustment module is further configured to adjust the parameters of the deployment parameter adjustment model based on the values ​​of the consideration parameters, with the deployment target of the target resource as the objective.

16. The apparatus according to any one of claims 11-14, wherein, The device further includes: The second acquisition module is used to acquire the traffic characteristics of each reference resource within a preset time period after the delivery parameters of the multiple reference resources are adjusted; the delivery parameter adjustment times of the multiple reference resources are all within a preset time window; the multiple reference resources include the target resource; The second acquisition module is used to acquire the values ​​of the consideration parameters of each of the multiple reference resources based on the traffic characteristics of each of the reference resources within the preset time length; the consideration parameters are the parameters to be considered for the target resource's deployment target; The adjustment module is further configured to adjust the parameters of the deployment parameter adjustment model based on the values ​​of the consideration parameters of each of the multiple reference resources, with the deployment target of the multiple reference resources as the objective; the multiple reference resources have the same deployment target.

17. The apparatus according to claim 16, wherein, The adjustment module is used for: Based on the values ​​of the consideration parameters of each of the plurality of reference resources, the value of the global consideration parameter is obtained; With the target of the deployment as the objective, the parameters of the deployment parameter adjustment model are adjusted based on the values ​​of the global consideration parameters.

18. The apparatus according to claim 17, wherein, The adjustment module is used for: Using the aforementioned delivery target as the objective and based on the values ​​of the global consideration parameters, an evolutionary algorithm is employed to adjust the parameters of the delivery parameter adjustment model.

19. The apparatus according to claim 16, wherein, The device further includes: The detection module is used to detect and determine that the delivery parameter adjustment model has not met the preset requirements based on the traffic characteristics of each of the plurality of reference resources within the preset time length and the preset evaluation threshold.

20. The apparatus according to any one of claims 11-14 and 17-19, wherein, The device further includes: The soft constraint processing module is used to perform soft constraint processing on the adjustment coefficients of the deployment parameters of the target resource.

21. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-10.

22. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-10.

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

Images