Resource delivery effect prediction method and device, electronic equipment and storage medium

Abstract

The invention relates to a resource delivery effect prediction method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a resource material of a to-be-predicted resource, and determining multi-modal information corresponding to the resource material; according to the multi-modal information, determining the similarity between the to-be-predicted resource and each historical resource in a plurality of historical resources; wherein the putting effect of the historical resources meets a preset putting effect condition; and inputting the plurality of similarities into a pre-trained resource effect prediction model to obtain a resource delivery prediction effect of the to-be-predicted resource output by the resource effect prediction model. Therefore, the accuracy of resource delivery effect prediction can be improved, and the user experience is improved.

Description

Technical Field

[0001] This application relates to the technical field of resource placement, and in particular, to a method, device, electronic device, and storage medium for predicting the effect of resource placement. Background Art

[0002] In the field of resource placement, personalized recommendation of resources is of great significance for improving user experience and resource conversion rate. Among them, accurate prediction of the resource placement effect (such as resource click-through rate) can not only improve the efficiency of resource placement, but also display more resource content that meets the interests and needs of users, thereby enhancing the overall user experience.

[0003] In the prior art, when predicting the placement effect of resources, it is generally based on the historical feedback behavior of users on resources in a historical time period, such as the click-through rate of users on resources in a historical time period. However, this method depends on the historical placement data of resources. When the historical placement data of resource creativity is small, the accuracy of resource prediction will be reduced, affecting the user experience. Summary of the Invention

[0004] This application provides a method, device, electronic device, and storage medium for predicting the effect of resource placement, so as to solve the technical problem in the prior art that when predicting the placement effect of resources, it depends on the historical placement data of resources. When the historical placement data of resource creativity is small, the accuracy of resource prediction will be reduced, affecting the user experience.

[0005] In a first aspect, this application provides a method for predicting the effect of resource placement, and the method includes:

[0006] Obtain the resource material of the resource to be predicted, and determine the multi-modal information corresponding to the resource material;

[0007] According to the multi-modal information, determine the similarity between the resource to be predicted and each of the multiple historical resources, where the placement effect of the historical resources meets a preset placement effect condition;

[0008] Input the multiple similarities into a pre-trained resource effect prediction model to obtain the resource placement prediction effect of the resource to be predicted output by the resource effect prediction model.

[0009] As an optional implementation, the resource material includes a creative picture and a creative text;

[0010] The determining the multi-modal information corresponding to the resource material includes:

[0011] Preprocess the creative picture to obtain a processed picture;

[0012] Process the processed picture through a pre-trained first processing model to obtain an image vector corresponding to the creative picture;

[0013] Perform word segmentation on the creative text to obtain a segmented text;

[0014] Process the segmented text through a pre-trained second processing model to obtain a text vector corresponding to the creative text;

[0015] Perform fusion processing on the image vector and the text vector to obtain multimodal information corresponding to the resource material.

[0016] As an optional implementation manner, the performing fusion processing on the image vector and the text vector to obtain multimodal information corresponding to the resource material includes:

[0017] Concatenate the image vector and the text vector to obtain multimodal information corresponding to the resource material;

[0018] Or,

[0019] Determine an image weight corresponding to the image vector;

[0020] Perform weighted fusion on the image vector and the text vector according to the image weight to obtain multimodal information corresponding to the resource material.

[0021] As an optional implementation manner, the determining, according to the multimodal information, a similarity between the to-be-predicted resource and each historical resource among a plurality of historical resources includes:

[0022] Obtain a unique identifier of each historical resource among a plurality of historical resources stored in a preset first database;

[0023] According to the unique identifier, obtain historical multimodal information of each historical resource from a preset second database;

[0024] Use a preset similarity algorithm to determine a similarity between the multimodal information and each historical multimodal information.

[0025] As an optional implementation manner, the delivery effect includes a delivery click-through rate, and the first database pre-stores a historical resource sequence, and the historical resource sequence is arranged in descending order of the delivery click-through rate;

[0026] The obtaining a unique identifier of each historical resource among a plurality of historical resources stored in a preset first database includes:

[0027] Obtain the unique identifiers of the top N historical resources in the historical resource sequence stored in the first database; N is a positive integer.

[0028] As an alternative implementation, the step of inputting the multiple similarities into a pre-trained resource effect prediction model to obtain the resource placement prediction effect of the to-be-predicted resource output by the resource effect prediction model includes:

[0029] Process the multiple similarities according to a preset processing algorithm to obtain similarity features;

[0030] Obtain the placement features of the resource material;

[0031] Input the similarity features, the multimodal information, and the placement features into a pre-trained resource effect prediction model to obtain the resource placement prediction effect of the to-be-predicted resource output by the resource effect prediction model.

[0032] As an alternative implementation, the step of processing the multiple similarities according to a preset processing algorithm to obtain similarity features includes:

[0033] Perform bucketing processing on the multiple similarities to obtain bucketed data;

[0034] Perform vector mapping on the bucketed data to obtain the similarity features.

[0035] As an alternative implementation, the step of processing the multiple similarities according to a preset processing algorithm to obtain similarity features includes:

[0036] Perform binary encoding on the multiple similarities to obtain a binary sequence;

[0037] Perform vector mapping on the binary sequence to obtain the similarity features.

[0038] As an alternative implementation, the step of obtaining the placement features of the resource material includes:

[0039] Obtain the historical placement effect features, historical placement user features, and resource placement strategy features corresponding to the resource material;

[0040] The step of inputting the similarity features, the multimodal information, and the placement features into a pre-trained resource effect prediction model to obtain the resource placement prediction effect of the to-be-predicted resource output by the resource effect prediction model includes:

[0041] Fuse the similarity features, the multimodal information, and the historical placement effect features to obtain fused features;

[0042] Input the fusion feature, the historical delivery user feature, and the resource delivery strategy feature into the resource effect prediction model to obtain the resource delivery prediction effect of the to-be-predicted resource output by the resource effect prediction model.

[0043] In a second aspect, the present application provides a resource delivery effect prediction device, and the device includes:

[0044] An acquisition module, configured to acquire resource materials of a to-be-predicted resource and determine multimodal information corresponding to the resource materials;

[0045] A determination module, configured to determine the similarity between the to-be-predicted resource and each of multiple historical resources according to the multimodal information; wherein the delivery effect of the historical resources meets a preset delivery effect condition;

[0046] A data processing module, configured to input multiple similarities into a pre-trained resource effect prediction model to obtain the resource delivery prediction effect of the to-be-predicted resource output by the resource effect prediction model.

[0047] As an optional implementation manner, the resource materials include a creative picture and creative text;

[0048] The acquisition module includes:

[0049] A preprocessing sub-module, configured to preprocess the creative picture to obtain a processed picture;

[0050] A picture processing sub-module, configured to process the processed picture through a pre-trained first processing model to obtain an image vector corresponding to the creative picture;

[0051] A word segmentation processing sub-module, configured to perform word segmentation processing on the creative text to obtain a segmented text;

[0052] A text processing sub-module, configured to process the segmented text through a pre-trained second processing model to obtain a text vector corresponding to the creative text;

[0053] A fusion processing sub-module, configured to perform fusion processing on the image vector and the text vector to obtain multimodal information corresponding to the resource materials.

[0054] As an optional implementation manner, the fusion processing sub-module is specifically configured to:

[0055] Concatenate the image vector and the text vector to obtain multimodal information corresponding to the resource materials;

[0056] Or,

[0057] Determine an image weight corresponding to the image vector;

[0058] According to the image weight, the image vector and the text vector are weighted and fused to obtain the multimodal information corresponding to the resource material.

[0059] As an optional implementation manner, the determining module includes:

[0060] A first obtaining sub-module, configured to obtain the unique identifier of each historical resource in a plurality of historical resources stored in a preset first database;

[0061] A second obtaining sub-module, configured to obtain the historical multimodal information of each historical resource from a preset second database according to the unique identifier;

[0062] A similarity determining sub-module, configured to determine the similarity between the multimodal information and each historical multimodal information by using a preset similarity algorithm.

[0063] As an optional implementation manner, the delivery effect includes a delivery click-through rate, and the first database prestores a historical resource sequence, and the historical resource sequence is arranged in descending order of the delivery click-through rate;

[0064] The first obtaining sub-module is specifically configured to:

[0065] Obtain the unique identifiers of the first N historical resources in the historical resource sequence stored in the first database; N is a positive integer.

[0066] As an optional implementation manner, the data processing module includes:

[0067] A similarity processing sub-module, configured to process a plurality of the similarities according to a preset processing algorithm to obtain a similarity feature;

[0068] A feature obtaining sub-module, configured to obtain the delivery feature of the resource material;

[0069] A data processing sub-module, configured to input the similarity feature, the multimodal information, and the delivery feature into a pre-trained resource effect prediction model to obtain the resource delivery prediction effect of the to-be-predicted resource output by the resource effect prediction model.

[0070] As an optional implementation manner, the similarity processing sub-module is specifically configured to:

[0071] Perform bucketing processing on a plurality of the similarities to obtain bucketing data;

[0072] Perform vector mapping on the bucketing data to obtain the similarity feature.

[0073] As an alternative implementation, the similarity processing sub-module is specifically configured to:

[0074] Perform binary encoding on the multiple similarities to obtain a binary sequence;

[0075] Perform vector mapping on the binary sequence to obtain the similarity feature.

[0076] As an alternative implementation, the feature acquisition sub-module is specifically configured to:

[0077] Obtain the historical delivery effect feature, the historical delivery user feature, and the resource delivery strategy feature corresponding to the resource material;

[0078] The data processing sub-module is specifically configured to:

[0079] Fuse the similarity feature, the multi-modal information, and the historical delivery effect feature to obtain a fused feature;

[0080] Input the fused feature, the historical delivery user feature, and the resource delivery strategy feature into the resource effect prediction model to obtain the resource delivery prediction effect of the to-be-predicted resource output by the resource effect prediction model.

[0081] In a third aspect, the present application provides an electronic device, including: a processor and a memory, where the processor is configured to execute a resource delivery effect prediction program stored in the memory to implement the resource delivery effect prediction method according to any one of the first aspects.

[0082] In a fourth aspect, the present application provides a storage medium, where the storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the resource delivery effect prediction method according to any one of the first aspects.

[0083] The technical solution provided by the embodiments of the present application, when predicting a to-be-predicted resource, obtains the multi-modal information of the to-be-predicted resource, and determines the similarity between the to-be-predicted resource and the historical resources that meet the preset delivery effect conditions according to the multi-modal information. Therefore, the resource effect prediction model trained in advance can be used to process multiple similarities to determine the resource delivery prediction effect of the to-be-predicted resource. It not only introduces the multi-modal information of the to-be-predicted resource, but also introduces the similarity feature, enabling the resource effect prediction model to utilize both the memory ability of the historical delivery effect and the multi-modal information of the resource content, improving the overall generalization ability and accuracy, and achieving the improvement of the accuracy of resource delivery effect prediction and the enhancement of user experience. Description of the Drawings

[0084] The accompanying drawings herein are incorporated into and constitute a part of this specification, showing embodiments consistent with the present invention, and are used in conjunction with the specification to explain the principles of the present invention.

[0085] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required for use in the embodiments or the prior art text. Obviously, for those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0086] One or more embodiments are exemplarily illustrated by the pictures in the corresponding accompanying drawings. These exemplary illustrations do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings are represented as similar elements. Unless otherwise stated, the drawings in the figures do not constitute a scale limitation.

[0087] Figure 1 It is a flowchart of an embodiment of a method for predicting the effect of resource investment provided by an embodiment of the present application;

[0088] Figure 2 It is a flowchart of an embodiment of another method for predicting the effect of resource investment provided by an embodiment of the present application;

[0089] Figure 3 It is a flowchart of an embodiment of yet another method for predicting the effect of resource investment provided by an embodiment of the present application;

[0090] Figure 4 It is a flowchart of an embodiment of still another method for predicting the effect of resource investment provided by an embodiment of the present application;

[0091] Figure 5 It is a block diagram of an embodiment of a device for predicting the effect of resource investment provided by an embodiment of the present application;

[0092] Figure 6 It is a schematic structural diagram of an electronic device provided by an embodiment of the present application. Detailed implementation manners

[0093] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present application in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are a part rather than all of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts fall within the scope of protection of the present application.

[0094] The following disclosure provides many different embodiments or examples for implementing different structures of the present invention. To simplify the disclosure of the present invention, the components and settings of specific examples are described in the following text. Of course, they are only examples and are not intended to limit the present invention. In addition, the present invention may repeat reference numerals and / or letters in different examples. This repetition is for the purpose of simplicity and clarity, and does not itself indicate the relationship between the various embodiments and / or settings discussed.

[0095] To solve the technical problem in the prior art that when predicting the delivery effect of resources, it depends on the historical delivery data of resources. When the historical delivery data of resource ideas is small, the accuracy of resource prediction will be reduced, affecting the user experience. The present application provides a method, device, electronic device and storage medium for predicting the delivery effect of resources. When predicting the resource to be predicted, it can obtain the multimodal information of the resource to be predicted, and determine the similarity between the resource to be predicted and the historical resources that meet the preset delivery effect conditions according to the multimodal information. Thus, the resource effect prediction model trained in advance can be used to process multiple similarities to determine the resource delivery prediction effect of the resource to be predicted. It not only introduces the multimodal information of the resource to be predicted, but also introduces the similarity feature, enabling the resource effect prediction model to not only utilize the memory ability of historical delivery effects, but also combine the multimodal information of resource content, improving the overall generalization ability and accuracy, and achieving the improvement of the accuracy of resource delivery effect prediction and the enhancement of user experience.

[0096] The following further explains the method for predicting the delivery effect of resources provided by the present application with specific embodiments in conjunction with the accompanying drawings. The embodiments do not constitute a limitation on the embodiments of the present invention.

[0097] See Figure 1 , which is the flowchart of the embodiment of a method for predicting the delivery effect of resources provided by the embodiment of the present application. As Figure 1 shown, the process may include the following steps:

[0098] Step 101, obtain the resource material of the resource to be predicted, and determine the corresponding multimodal information of the resource material.

[0099] The above-mentioned resource to be predicted refers to the resource whose delivery effect is to be predicted, which can be a resource to be delivered or a resource stored in a preset database. The embodiments of the present application do not limit this. Among them, the resource to be predicted can be an advertisement, a movie or TV drama, a book or a novel, etc. The embodiments of the present application do not limit this.

[0100] The above-mentioned resource material refers to the relevant materials used to generate the resource, such as pictures, texts, etc.

[0101] The above multi-modal information refers to information containing different aspects of the resource materials. For example, it may include the text features and image features of the resource materials.

[0102] In the embodiments of the present application, multiple resources can be stored in a preset resource library, and each resource can correspond to multiple resource materials.

[0103] Based on this, as an alternative implementation, the execution entity of the embodiments of the present application can determine each resource in the resource library as a resource to be predicted.

[0104] As another alternative implementation, the user can select the resources that need to be predicted through a visual interface. Based on this, the execution entity of the embodiments of the present application can receive the user's selection through the visual interface and determine the selected resources as resources to be predicted.

[0105] As yet another alternative implementation, the execution entity of the embodiments of the present application can periodically determine multiple recent resources and determine the determined resources as resources to be predicted.

[0106] As still another alternative implementation, every time the execution entity of the embodiments of the present application detects a new resource, it can determine the resource as a resource to be predicted.

[0107] In the embodiments of the present application, after determining the resource to be predicted, the resource materials of the resource to be predicted can be obtained from the preset resource library, and the multi-modal information corresponding to the resource materials can be determined.

[0108] As an alternative implementation, the corresponding relationship between the unique identifier of each resource and the resource materials included in the resource can be stored in the above resource library. Based on this, the execution entity of the embodiments of the present application can determine the unique identifier of the resource to be predicted, and search for the target corresponding relationship including the unique identifier from the above corresponding relationship. After that, the resource materials included in the target corresponding relationship can be determined as the resource materials of the above resource to be predicted.

[0109] As for how to determine the multi-modal information corresponding to the above resource materials specifically, it will be described in the following through Figure 2 the shown process, which will not be elaborated here.

[0110] Step 102: Determine the similarity between the resource to be predicted and each historical resource among multiple historical resources according to the above multi-modal information; wherein, the placement effect of the above historical resources meets the preset placement effect conditions.

[0111] The above historical resources refer to the resources that have been placed in the historical time period and have placement effects.

[0112] The above similarity refers to the similarity between the resource to be predicted and each historical resource.

[0113] The above-mentioned delivery effect condition refers to a condition preset to represent that the delivery effect of the historical resource is good. The delivery effect condition can be the feedback information of the user. For example, when the delivery effect is the click-through rate of the user on the historical resource, then the delivery effect condition is greater than the preset click-through rate threshold.

[0114] In the embodiments of the present application, in order to combine the historical delivery effect of the resource and the resource content of the resource to be predicted, after determining the multi-modal information of the resource material corresponding to the resource to be predicted, the execution subject of the embodiments of the present application can, according to the multi-modal information, determine the similarity between the resource to be predicted and each historical resource among multiple historical resources whose delivery effect in the historical time period meets the preset delivery effect condition.

[0115] As for how to specifically determine the similarity between the resource to be predicted and each historical resource, it can be described in the following through Figure 3 the process shown, which will not be elaborated here first.

[0116] Step 103: Input the multiple similarities into a pre-trained resource effect prediction model to obtain the resource delivery prediction effect of the resource to be predicted output by the resource effect prediction model.

[0117] The above-mentioned resource effect prediction model is a model pre-trained for predicting the delivery effect of a resource. The resource effect prediction model can process the multiple similarities corresponding to the resource to be predicted to predict the delivery effect of the resource to be predicted.

[0118] In the embodiments of the present application, after determining the similarity between the resource to be predicted and each historical resource, the above-mentioned multiple similarities can be input into the above-mentioned pre-trained resource effect prediction model to obtain the resource delivery prediction effect of the above-mentioned resource to be predicted output by the resource effect prediction model.

[0119] Among them, in addition to including the similarity features of the resource to be predicted, the input of the above-mentioned resource effect prediction model can also include other features of the resource to be predicted, which can be specifically described in the following through Figure 4 the process shown, which will not be elaborated here first.

[0120] As an optional implementation manner, when training the above-mentioned resource effect prediction model, a resource feature sample set can be first determined. The resource feature sample set can include multiple historical resource samples, and each historical resource sample can correspond to similarity feature samples, other feature samples with other historical resource samples, and resource delivery effect samples of the historical resource.

[0121] After that, the preset initial resource effect prediction model can be trained by using the above-mentioned resource feature sample set, and the initial resource delivery prediction effect output by the initial resource effect prediction model obtained each time of training can be obtained.

[0122] After that, the current loss value can be calculated according to the preset loss value function for the above initial resource placement prediction effect and the corresponding resource placement effect samples.

[0123] Optionally, when the loss value is less than the preset threshold or the number of training times reaches the preset number threshold, it can be determined that the current training is completed, and a resource effect prediction model is obtained.

[0124] Optionally, when the loss value is greater than or equal to the preset threshold, it indicates that the training of the initial resource effect prediction model is not completed at this time. Therefore, the parameters in the initial resource effect prediction model can be adjusted according to the preset step size, and after the adjustment, the initial resource effect prediction model is continuously trained using the resource feature sample set until the obtained loss value is less than the preset threshold or the number of training times reaches the preset number threshold.

[0125] The technical solution provided in the embodiments of the present application obtains the resource materials of the resource to be predicted, determines the multi-modal information corresponding to the resource materials, and according to the above multi-modal information, determines the similarity between the resource to be predicted and each historical resource among multiple historical resources; wherein, the placement effects of the above historical resources meet the preset placement effect conditions, and inputs the multiple similarities into the pre-trained resource effect prediction model to obtain the resource placement prediction effect of the resource to be predicted output by the resource effect prediction model. This technical solution, when predicting the resource to be predicted, obtains the multi-modal information of the resource to be predicted, and determines the similarity between the resource to be predicted and the historical resources whose placement effects meet the preset placement effect conditions according to the multi-modal information, so that the pre-trained resource effect prediction model can be used to process the multiple similarities to determine the resource placement prediction effect of the resource to be predicted. It not only introduces the multi-modal information of the resource to be predicted, but also introduces the similarity feature, enabling the resource effect prediction model to not only utilize the memory ability of historical placement effects, but also combine the multi-modal information of resource content, improving the overall generalization ability and accuracy, and achieving the improvement of the accuracy of resource placement effect prediction and the enhancement of user experience.

[0126] See Figure 2 , which is a flowchart of an embodiment of another resource placement effect prediction method provided by the embodiments of the present application. Figure 2 The process shown Figure 1 On the basis of the process shown, it describes how to determine the multi-modal information corresponding to the resource materials in the case where the resource materials include creative pictures and creative texts. As Figure 2 shown, the process may include the following steps:

[0127] Step 201, preprocess the creative picture to obtain a processed picture.

[0128] The above-mentioned creative picture refers to the picture involved in the resource creativity for generating the resource to be predicted.

[0129] The above-mentioned preprocessing refers to preprocessing the creative picture before extracting the features of the creative picture. The preprocessing may include, but is not limited to: scaling the creative picture, updating the pixels of the creative picture so that the pixels of the creative picture are within a preset pixel range, etc.

[0130] The above-mentioned processed picture is the picture obtained after preprocessing the creative picture.

[0131] In the embodiments of the present application, in order to more accurately extract the image features of the creative picture, the execution subject of the embodiments of the present application may first preprocess the creative picture so that the creative picture meets the preset feature extraction conditions.

[0132] As an optional implementation manner, the above-mentioned feature extraction conditions may include that the size of the creative picture is smaller than a preset size threshold. Based on this, the execution subject of the embodiments of the present application may determine whether the size of each creative picture is smaller than the above-mentioned size threshold.

[0133] Optionally, when it is determined that the size of the creative picture is smaller than the above-mentioned size threshold, it may be determined that the creative picture meets the feature extraction conditions, so the creative picture may not be preprocessed.

[0134] Optionally, when it is determined that the size of the creative picture is greater than or equal to the above-mentioned size threshold, it may be determined that the creative picture does not meet the feature extraction conditions, so the creative picture may be scaled so that the size of the creative picture is smaller than the above-mentioned size threshold.

[0135] As another optional implementation manner, the above-mentioned feature extraction conditions may include that the pixel value of the creative picture is within a preset pixel range. Based on this, the execution subject of the embodiments of the present application may determine whether the pixel value of each creative picture is within the above-mentioned pixel range.

[0136] Optionally, when it is determined that the pixel value of the creative picture is within the above-mentioned pixel range, it may be determined that the creative picture meets the feature extraction conditions, so the creative picture may not be preprocessed.

[0137] Optionally, when it is determined that the pixel value of the creative picture is not within the above-mentioned pixel range, it may be determined that the creative picture does not meet the feature extraction conditions, so the resolution of the creative picture may be updated so that the pixel value of the creative picture is within the above-mentioned pixel range.

[0138] As another alternative implementation, the above feature extraction conditions may include that the size of the creative picture is smaller than a preset size threshold, and the pixel value of the creative picture is within a preset pixel range. Based on this, the execution entity of the embodiment of the present application can determine, for each creative picture, whether the size of the creative picture is smaller than the above size threshold, and whether the pixel value of the creative picture is within the above pixel range.

[0139] Optionally, when it is determined that the size of the creative picture is smaller than the above size threshold and the pixel value of the creative picture is within the above pixel range, it can be determined that the creative picture meets the feature extraction conditions, so the creative picture does not need to be preprocessed.

[0140] Optionally, when it is determined that the size of the creative picture is greater than or equal to the above size threshold and the pixel value of the creative picture is within the above pixel range, it can be determined that the creative picture does not meet the feature extraction conditions, so the creative picture can be scaled to make the size of the creative picture smaller than the above size threshold.

[0141] Optionally, when it is determined that the size of the creative picture is smaller than the above size threshold and the pixel value of the creative picture is not within the above pixel range, it can be determined that the creative picture does not meet the feature extraction conditions, so the resolution of the creative picture can be updated to make the pixel value of the creative picture within the above pixel range.

[0142] Optionally, when it is determined that the size of the creative picture is greater than or equal to the above size threshold and the pixel value of the creative picture is not within the above pixel range, it can be determined that the creative picture does not meet the feature extraction conditions, so the creative picture can be scaled to make the size of the creative picture smaller than the above size threshold, and the resolution of the creative picture can be updated to make the pixel value of the creative picture within the above pixel range.

[0143] Step 202: Process the above processed picture through a pre-trained first processing model to obtain an image vector corresponding to the creative picture.

[0144] The above first processing model refers to a pre-trained model for extracting picture features, and this model can be trained based on the CN-CLIP (Chinese Contrastive Language Image Pretraining) large model.

[0145] The above image vector refers to the picture features extracted from the processed picture.

[0146] In the embodiments of the present application, the execution entity of the present application can pre-train a picture feature extraction model for extracting picture features (hereinafter referred to as the "first processing model" for convenience of description).

[0147] Based on this, when the execution entity of the embodiments of the present application extracts the picture features in the processing picture corresponding to the resource creativity, it can process the above-mentioned processing picture through the first processing model to obtain the image vector corresponding to the creative picture.

[0148] As an implementation manner, the above-mentioned processing picture can be input into the above-mentioned first processing model, and the image vector corresponding to the processing picture output by the first processing model can be obtained.

[0149] Step 203: Perform word segmentation processing on the above-mentioned creative text to obtain a segmented text.

[0150] Step 204: Process the above-mentioned segmented text through a pre-trained second processing model to obtain a text vector corresponding to the creative text.

[0151] The following is a unified description of steps 203 and 204:

[0152] The above-mentioned creative text refers to the text content in the material of the resource to be predicted that is used to construct the creativity of the resource to be predicted.

[0153] The above-mentioned second processing model refers to a model pre-trained for extracting text features of text information, and this model can be trained based on the CN-CLIP (Chinese Contrastive Language Image Pretraining) large model.

[0154] The above-mentioned text vector refers to the text features corresponding to the text information contained in the creative picture.

[0155] In the embodiments of the present application, in order to more accurately extract the own content of the resource creativity, the execution entity of the embodiments of the present application can extract not only the picture features of the resource creativity but also the text features of the resource creativity.

[0156] As an optional implementation manner, the execution entity of the embodiments of the present application can pre-train a text feature extraction model for extracting text features (hereinafter referred to as the "second processing model" for convenience of description). Based on this, when extracting the text features of the resource creativity, word segmentation processing can be performed on the creative text to obtain a corresponding segmented text.

[0157] After that, the above-mentioned segmented text can be input into the above-mentioned pre-trained second processing model, and the text vector of the creative text output by the second processing model can be obtained.

[0158] Step 205: Perform fusion processing on the above-mentioned image vector and text vector to obtain the multi-modal information of the resource creativity.

[0159] The above-mentioned multi-modal information refers to the comprehensive features that simultaneously include the image features of the creative picture and the text features of the creative text, and can be represented by a multi-modal vector.

[0160] In the embodiment of the present application, after obtaining the image vector and text vector of the resource to be predicted, in order to simplify the processing flow, the execution subject of the embodiment of the present application can fuse the above-mentioned image vector and text vector to obtain the multi-modal information of the resource to be predicted.

[0161] As an optional implementation manner, the above-mentioned image vector and text vector can be directly spliced and fused to obtain a multi-modal vector, and this multi-modal vector is determined as the multi-modal information of the above-mentioned resource material.

[0162] As another optional implementation manner, the image weight corresponding to the image vector can be determined, and according to the above-mentioned image weight, the above-mentioned image vector and text vector are weighted and fused to obtain the multi-modal information of the above-mentioned resource material.

[0163] As an exemplary implementation manner, each resource to be predicted may correspond to a preset weight value. Therefore, the execution subject of the embodiment of the present application can directly obtain this image weight.

[0164] As another exemplary implementation manner, the execution subject of the embodiment of the present application can obtain the placement position corresponding to the resource to be predicted, and determine the image weight corresponding to the image vector according to this placement position.

[0165] As an implementation manner, a weight analysis model can be pre-trained, and this placement position is input into the above-mentioned weight analysis model to obtain the image weight output by the weight analysis model.

[0166] As yet another exemplary implementation manner, the execution subject of the embodiment of the present application can analyze the creative text to determine the text content type corresponding to this creative text.

[0167] After that, the target text weight corresponding to this text content type can be determined from the pre-set correspondence between the text content type and the text weight.

[0168] Finally, the preset value (such as 1) can be subtracted from the above-mentioned target text weight to obtain the image weight corresponding to the image vector.

[0169] In one embodiment, when weighted fusion of the image vector and the text vector is performed according to the image weight, the text weight can be obtained by subtracting the image weight from a preset value. Then, the above image vector and the image weight can be multiplied to obtain a first vector, and the above text vector and the text weight can be multiplied to obtain a second vector. Finally, the above first vector and the second vector can be added to obtain the multimodal information of the resource material.

[0170] The technical solution provided by the embodiments of the present application, in the case where the resource material includes a creative picture and creative text, preprocesses the creative picture to obtain a processed picture, processes the above processed picture through a pre-trained first processing model to obtain an image vector corresponding to the creative picture, performs word segmentation processing on the above creative text to obtain a segmented text, processes the above segmented text through a pre-trained second processing model to obtain a text vector corresponding to the creative text, and performs fusion processing on the above image vector and the text vector to obtain the multimodal information corresponding to the resource material. This technical solution, by pre-training a first processing model for extracting image vectors and a second processing model for extracting text vectors, and using the first processing model and the second processing model to extract the image vector of the creative picture and the text vector of the creative text respectively, thereby fusing the image vector and the text vector to obtain the multimodal information corresponding to the resource material. And by using the model to extract the features of the creative picture and the creative text, the accuracy and efficiency of feature extraction can be improved, and the multimodal information corresponding to the resource material can be determined accurately and efficiently.

[0171] See Figure 3 , which is a flowchart of an embodiment of another resource placement effect prediction method provided by the embodiments of the present application. Figure 3 The process shown Figure 1 On the basis of the process shown, it describes how to determine the similarity between the resource to be predicted and each historical resource according to the multimodal information. As Figure 3 shown, the process may include the following steps:

[0172] Step 301, obtain the unique identifier of each historical resource in a plurality of historical resources stored in a preset first database.

[0173] The above first database refers to a database for storing the unique identifiers of historical resources, which may be a couchbase database or other databases, and the embodiments of the present application do not limit this.

[0174] The above-mentioned unique identifier refers to a unique identifier used to identify each resource. Among them, each resource can correspond to a unique identifier, and different resources correspond to different unique identifiers. The unique identifier can be a resource ID (Identity document, identity identification number), or it can be a string, number, letter, or other type. The embodiments of the present application do not limit this.

[0175] In the embodiments of the present application, in order to facilitate the management of the unique identifiers of historical resources, technicians can store historical resources, the unique identifiers of historical resources, and the corresponding relationship between the two in a preset database (for the convenience of description, hereinafter referred to as the "first database"). Based on this, the execution entity of the embodiments of the present application can directly obtain the unique identifiers of multiple historical resources from the above-mentioned first database.

[0176] As an optional implementation manner, in order to better conform to the development status of resources, the execution entity of the embodiments of the present application can obtain the unique identifiers of historical resources within a preset time period closest to the current time from the above-mentioned first database.

[0177] As another optional implementation manner, in order to more accurately predict the placement effect of the resource to be predicted, the execution entity of the embodiments of the present application can select the unique identifiers of historical resources with higher placement effects.

[0178] As an exemplary implementation manner, the placement effect of each historical resource can include the placement click-through rate at which the historical resource is clicked by the user after being placed in the historical time period. Based on this, the above-mentioned first database can pre-store a historical resource sequence, and the historical resource sequence can be arranged in descending order of the above-mentioned click-through rate.

[0179] Based on this, when the execution entity of the embodiments of the present application obtains the unique identifier of each historical resource among multiple historical resources from the first database, it can obtain the unique identifiers of the first N historical resources stored in the historical resource sequence stored in the above-mentioned first database. Among them, the above-mentioned N is a positive integer.

[0180] Step 302: According to the above-mentioned unique identifier, obtain the historical multi-modal information of each historical resource from a preset second database.

[0181] The above-mentioned second database refers to a preset database for storing the historical multi-modal information of each historical resource. It can be a Redis database or other databases. The embodiments of the present application do not limit this.

[0182] In the embodiments of the present application, since the vector dimension of the multimodal information is relatively high, if the multimodal information and historical resources are stored in the same database, problems related to the protobuf 2GB limit may occur when loading and using them in the native way of TensorFlow (a symbolic mathematics system). Here, the protobuf 2GB limit means that when Protobuf serializes and deserializes data, the size of a single serialized message cannot exceed 2GB. This limit is determined by the design of Protobuf and is mainly used to control the data size and prevent problems such as memory overflow.

[0183] Based on this, in the embodiments of the present application, the multimodal information of each historical resource can be stored in a second database different from the first database. The above-mentioned second database may include multimodal information, unique identifiers, and the corresponding relationships between them.

[0184] Based on this, after obtaining the unique identifiers of multiple historical resources, the execution entity of the embodiments of the present application can obtain the historical multimodal information corresponding to each historical resource from the above-mentioned second database according to the above-mentioned unique identifiers.

[0185] As an exemplary implementation manner, the corresponding relationship can be searched according to the above-mentioned unique identifier to obtain a target corresponding relationship including the above-mentioned unique identifier. Then, each historical multimodal information included in the above-mentioned target corresponding relationship can be obtained.

[0186] As an exemplary implementation manner, in order to avoid problems related to the protobuf 2GB limit that may occur when loading and using in the native way of TensorFlow, when obtaining historical multimodal information, the execution entity of the embodiments of the present application can obtain it through a preset data acquisition tool. The above-mentioned data acquisition tool can be TFRA (TensorFlow Recommenders Addons, a toolset for TensorFlow recommendation systems).

[0187] Step 303: Use a preset similarity algorithm to determine the similarity between the multimodal information and each piece of historical multimodal information.

[0188] The above-mentioned similarity algorithm refers to an algorithm preset for calculating the similarity between two pieces of information. It can be a cosine similarity algorithm or an Euclidean distance similarity algorithm. The embodiments of the present application do not limit this.

[0189] In the embodiments of the present application, after obtaining the historical multimodal information corresponding to each historical resource, for each piece of historical multimodal information, a preset similarity algorithm can be used to calculate the similarity between the multimodal information of the resource to be predicted and the historical multimodal information.

[0190] As an alternative implementation, the execution entity of the embodiments of the present application may calculate the above-mentioned multimodal information and historical multimodal information using a preset first similarity algorithm to obtain the cosine similarity between the two. Subsequently, the cosine similarity may be determined as the similarity between the multimodal information and the historical multimodal information.

[0191] As another alternative implementation, the execution entity of the embodiments of the present application may calculate the above-mentioned multimodal information and historical multimodal information using a preset second similarity algorithm to obtain the Euclidean distance between the two. Subsequently, the Euclidean distance may be determined as the similarity between the multimodal information and the historical multimodal information.

[0192] The technical solution provided by the embodiments of the present application obtains the unique identifier of each historical resource stored in a preset first database, and according to the above-mentioned unique identifier, obtains the historical multimodal information of each historical resource from a preset second database, and uses a preset similarity algorithm to determine the similarity between the multimodal information and each historical multimodal information. This technical solution realizes the efficient and accurate acquisition of the multimodal information of each historical resource by using two different databases to store the unique identifiers and multimodal information of historical resources respectively, and directly obtaining the multimodal information of each historical resource using the unique identifier, thereby improving the efficiency and accuracy of determining the similarity between the multimodal information and each historical multimodal information.

[0193] See Figure 4 , which is a flowchart of an embodiment of another resource placement effect prediction method provided by the embodiments of the present application. Figure 4 The process shown Figure 1 On the basis of the process shown, it describes how to specifically use the resource effect prediction model to determine the resource placement prediction effect of the resource to be predicted. As Figure 4 shown, the process may include the following steps:

[0194] Step 401, process multiple similarities according to a preset processing algorithm to obtain similarity features.

[0195] The above-mentioned processing algorithm refers to an algorithm preset for processing the obtained multiple similarities.

[0196] In the embodiments of the present application, in order to facilitate the resource effect prediction model to process multiple similarities, before inputting the multiple similarities into a pre-trained resource effect prediction model, the execution entity of the embodiments of the present application may process the multiple similarities according to a preset processing algorithm to obtain corresponding similarity features.

[0197] As an alternative implementation, multiple similarities can be bucketed to obtain corresponding bucketed data. Subsequently, the above-mentioned bucketed data can be vector-mapped to obtain corresponding similarity features.

[0198] As an exemplary implementation, multiple similarities can be bucketed according to a preset bucketing rule to obtain corresponding bucketed data.

[0199] As another alternative implementation, multiple similarities can be binary-encoded to obtain a binary sequence. Subsequently, the above-mentioned binary sequence can be vector-mapped to obtain corresponding similarity features.

[0200] Step 402: Obtain the placement features of the resource material.

[0201] Step 403: Input the above-mentioned similarity features, multi-modal information, and the above-mentioned placement features into a pre-trained resource effect prediction model to obtain the resource placement prediction effect of the to-be-predicted resource output by the resource effect prediction model.

[0202] The following provides a unified description of Step 402 and Step 403:

[0203] The above-mentioned placement features refer to the features related to the placement of the resource material corresponding to the to-be-predicted resource, which may include but are not limited to: the historical placement effect features of the resource material, the historical placement user features, and the resource placement strategy features.

[0204] The above-mentioned historical placement effect features refer to the placement effect features feedback by users during the placement of the resource material within a historical time period, such as the click-through rate.

[0205] The above-mentioned historical placement user features refer to the user features of the users targeted by the resource material during historical placement, or the user features of the users who provide feedback on the placed resource material, which may include but are not limited to: age features, gender features, and geographical location features, etc.

[0206] The above-mentioned resource placement strategy features refer to the strategies related to the placement of the resource material corresponding to the to-be-predicted resource, which may include but are not limited to: placement location, placement time, and placement duration, etc.

[0207] In the embodiments of the present application, in order to improve the accuracy of predicting the placement effect of the to-be-predicted resource, in addition to obtaining the above-mentioned similarity features and multi-modal information, the execution entity of the embodiments of the present application may also obtain the placement features of the resource material.

[0208] After that, the above similarity features, multimodal information, and placement features can be input into a pre-trained resource effect prediction model, so that the resource effect prediction model can process the above similarity features, multimodal information, and placement features to determine the resource placement prediction effect of the resource to be predicted.

[0209] As an optional implementation, when obtaining the placement features corresponding to the resource material, the historical placement effect features, historical placement user features, and resource placement strategy features corresponding to the resource material can be obtained.

[0210] The sum. Since the above similarity features, multimodal information, and historical placement effect features are all relevant features of the resource to be predicted itself, the above similarity features, multimodal information, and historical placement effect features can be fused to obtain a fused feature.

[0211] As an optional implementation, the above similarity features, multimodal information, and historical placement effect features are all vector features. Therefore, the above three features can be directly concatenated and fused to obtain a fused feature.

[0212] As another optional implementation, the above similarity features, multimodal information, and historical placement effect features are all vector features. The weight value corresponding to each feature can be determined, and the above similarity features, multimodal information, and historical placement effect features can be weighted and fused according to the weight value corresponding to each feature to obtain the corresponding fused feature.

[0213] Finally, the above fused feature, historical placement user features, and resource placement strategy features can be input into the above resource effect prediction model to obtain the resource placement prediction effect of the resource to be predicted output by the resource effect prediction model.

[0214] The technical solution provided by the embodiments of the present application processes multiple similarities according to a preset processing algorithm to obtain similarity features, obtains the placement features of the resource material, inputs the above similarity features, multimodal information, and the above placement features into a pre-trained resource effect prediction model, and obtains the resource placement prediction effect of the resource to be predicted output by the resource effect prediction model. This technical solution obtains the placement features of the resource material and uses the resource effect prediction model to process the similarity features, multimodal information, and placement features of the resource to be predicted at the same time to obtain the resource placement prediction result of the resource to be predicted. Through features from multiple different aspects, it can make the resource effect prediction model more accurately predict the resource placement effect of the resource to be predicted, achieving an improvement in the accuracy of the resource placement prediction effect.

[0215] See Figure 5, which is a block diagram of an embodiment of a resource placement effect prediction device provided by an embodiment of the present application. As Figure 5 shown, the device may include:

[0216] An acquisition module 51, configured to acquire resource materials of a resource to be predicted and determine multimodal information corresponding to the resource materials;

[0217] A determination module 52, configured to determine the similarity between the resource to be predicted and each of a plurality of historical resources according to the multimodal information; wherein the placement effects of the historical resources meet a preset placement effect condition;

[0218] A data processing module 53, configured to input the plurality of similarities into a pre-trained resource effect prediction model to obtain a resource placement prediction effect of the resource to be predicted output by the resource effect prediction model.

[0219] As an optional implementation manner, the resource materials include creative pictures and creative texts;

[0220] The acquisition module 51 includes:

[0221] A preprocessing sub-module, configured to preprocess the creative picture to obtain a processed picture;

[0222] A picture processing sub-module, configured to process the processed picture through a pre-trained first processing model to obtain an image vector corresponding to the creative picture;

[0223] A word segmentation processing sub-module, configured to perform word segmentation processing on the creative text to obtain a segmented text;

[0224] A text processing sub-module, configured to process the segmented text through a pre-trained second processing model to obtain a text vector corresponding to the creative text;

[0225] A fusion processing sub-module, configured to perform fusion processing on the image vector and the text vector to obtain multimodal information corresponding to the resource materials.

[0226] As an optional implementation manner, the fusion processing sub-module is specifically configured to:

[0227] Concatenate the image vector and the text vector to obtain multimodal information of the resource materials;

[0228] Or,

[0229] Determine an image weight corresponding to the image vector;

[0230] Perform weighted fusion on the image vector and the text vector according to the image weight to obtain multimodal information of the resource materials.

[0231] As an alternative implementation, the determining module 52 includes:

[0232] A first acquisition sub-module, configured to acquire the unique identifier of each of the multiple historical resources stored in a preset first database;

[0233] A second acquisition sub-module, configured to acquire the historical multimodal information of each of the historical resources from a preset second database according to the unique identifier;

[0234] A similarity determination sub-module, configured to determine the similarity between the multimodal information and each piece of historical multimodal information by using a preset similarity algorithm.

[0235] As an alternative implementation, the placement effect includes the placement click-through rate. The first database prestores a historical resource sequence, and the historical resource sequence is arranged in descending order of the placement click-through rate;

[0236] The first acquisition sub-module is specifically configured to:

[0237] Acquire the unique identifiers of the first N historical resources in the historical resource sequence stored in the first database; N is a positive integer.

[0238] As an alternative implementation, the data processing module 53 includes:

[0239] A similarity processing sub-module, configured to process the multiple similarities according to a preset processing algorithm to obtain similarity features;

[0240] A feature acquisition sub-module, configured to acquire the placement features of the resource material;

[0241] A data processing sub-module, configured to input the similarity features, the multimodal information, and the placement features into a pre-trained resource effect prediction model to obtain the resource placement prediction effect of the to-be-predicted resource output by the resource effect prediction model.

[0242] As an alternative implementation, the similarity processing sub-module is specifically configured to:

[0243] Perform bucketing processing on the multiple similarities to obtain bucketing data;

[0244] Perform vector mapping on the bucketing data to obtain the similarity features.

[0245] As an alternative implementation, the similarity processing sub-module is specifically configured to:

[0246] Perform binary encoding on the multiple similarities to obtain a binary sequence;

[0247] Perform vector mapping on the binary sequence to obtain the similarity feature.

[0248] As an optional implementation manner, the feature acquisition sub-module is specifically configured to:

[0249] Obtain the historical delivery effect feature, the historical delivery user feature, and the resource delivery strategy feature corresponding to the resource material;

[0250] The data processing sub-module is specifically configured to:

[0251] Fuse the similarity feature, the multimodal information, and the historical delivery effect feature to obtain a fused feature;

[0252] Input the fused feature, the historical delivery user feature, and the resource delivery strategy feature into the resource effect prediction model to obtain the resource delivery prediction effect of the to-be-predicted resource output by the resource effect prediction model.

[0253] As Figure 6 shown, a schematic structural diagram of an electronic device provided by an embodiment of the present application includes a processor 61, a communication interface 62, a memory 63, and a communication bus 64. Among them, the processor 61, the communication interface 62, and the memory 63 complete communication with each other through the communication bus 64,

[0254] The memory 63 is used to store a computer program;

[0255] In an embodiment of the present application, when the processor 61 is used to execute the program stored on the memory 63, it implements the resource delivery effect prediction method provided by any one of the foregoing method embodiments, including:

[0256] Obtain the resource material of the to-be-predicted resource and determine the multimodal information corresponding to the resource material;

[0257] According to the multimodal information, determine the similarity between the to-be-predicted resource and each of the multiple historical resources, where the delivery effect of the historical resource meets a preset delivery effect condition;

[0258] Input the multiple similarities into a pre-trained resource effect prediction model to obtain the resource delivery prediction effect of the to-be-predicted resource output by the resource effect prediction model.

[0259] An embodiment of the present application further provides a storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of the resource delivery effect prediction method provided by any one of the foregoing method embodiments.

[0260] The device embodiments of the above text are merely illustrative. The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple grid units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0261] Through the above text of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the above technical solutions, in essence, or the part that contributes to the relevant technologies, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a grid device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0262] It should be understood that the terms used in the text are only for the purpose of specific example embodiments of the text and are not intended to be restrictive. Unless the context clearly indicates otherwise, the singular forms "a", "an", and "the" as used in the text may also represent the plural form. The terms "include", "comprise", "contain", and "have" are inclusive and thus specify the presence of the stated features, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, elements, components, and / or their combinations. The method steps, processes, and operations in the text are not to be construed as necessarily requiring them to be executed in the specific order stated or illustrated, unless the execution order is explicitly indicated. It should also be understood that additional or alternative steps can be used.

[0263] The above description is only the specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown in the text, but will conform to the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A resource delivery effect prediction method, characterized in that: The method comprises: Acquire resource material of the resource to be predicted, and determine multimodal information corresponding to the resource material; Determine, based on the multimodal information, the similarity between the resource to be predicted and each of the multiple historical resources; wherein the delivery effect of the historical resources satisfies a preset delivery effect condition; The multiple similarities are input into a pre-trained resource effect prediction model to obtain the resource delivery prediction effect of the resource to be predicted output by the resource effect prediction model.

2. The method according to claim 1, characterized in that The resource materials include creative pictures and creative texts; The determining the multimodal information corresponding to the resource material includes: Preprocessing the creative image to obtain a processed image; Processing the processing picture by using a pre-trained first processing model to obtain an image vector corresponding to the creative picture; Performing word segmentation processing on the creative text to obtain a word segmentation text; Processing the segmented text by a pre-trained second processing model to obtain a text vector corresponding to the creative text; The image vector and the text vector are fused to obtain multimodal information corresponding to the resource material.

3. The method according to claim 2, characterized in that The fusing the image vector and the text vector to obtain multimodal information corresponding to the resource material includes: splicing the image vector and the text vector to obtain multimodal information corresponding to the resource material; or, Determining an image weight corresponding to the image vector; According to the image weight, the image vector and the text vector are weightedly fused to obtain multimodal information corresponding to the resource material.

4. The method according to claim 1, characterized in that: The determining, based on the multimodal information, a similarity between the resource to be predicted and each of the plurality of historical resources includes: Obtaining a unique identifier of each of the plurality of historical resources stored in the preset first database; According to the unique identifier, acquiring the historical multimodal information of each of the historical resources from a preset second database; The similarity between the multimodal information and each piece of historical multimodal information is determined using a preset similarity algorithm.

5. The method according to claim 4, characterized in that The delivery effect includes delivery click-through rate, the first database pre-stores a historical resource sequence, and the historical resource sequence is arranged in descending order according to the delivery click-through rate; The obtaining of the unique identifier of each of the plurality of historical resources stored in the preset first database includes: Obtaining unique identifiers of the first N historical resources in the historical resource sequence stored in the first database; Said N is a positive integer.

6. The method according to claim 1, characterized in that The step of inputting the plurality of similarities into a pre-trained resource effect prediction model to obtain a resource delivery prediction effect of the resource to be predicted outputted by the resource effect prediction model comprises: Processing the plurality of similarities according to a preset processing algorithm to obtain similarity features; Obtaining delivery characteristics of the resource material; The similarity feature, the multimodal information, and the delivery feature are input into a pre-trained resource effect prediction model to obtain a resource delivery prediction effect of the resource to be predicted output by the resource effect prediction model.

7. The method according to claim 6, characterized in that The processing of the plurality of similarities according to a preset processing algorithm to obtain similarity features includes: Performing bucket processing on the plurality of similarities to obtain bucket data; The bucketed data is vector-mapped to obtain the similarity feature.

8. The method according to claim 6, characterized in that The processing of the plurality of similarities according to a preset processing algorithm to obtain similarity features includes: Binary encoding the plurality of similarities to obtain a binary sequence; Perform vector mapping on the binary sequence to obtain the similarity feature.

9. The method according to claim 6, characterized in that The obtaining of the delivery characteristics of the resource material includes: Obtaining historical delivery effect characteristics, historical delivery user characteristics, and resource delivery strategy characteristics corresponding to the resource material; The inputting the similarity feature, the multimodal information, and the delivery feature into a pre-trained resource effect prediction model to obtain the resource delivery prediction effect of the resource to be predicted output by the resource effect prediction model includes: The similarity feature, the multimodal information, and the historical delivery effect feature are fused to obtain a fusion feature; The fusion feature, the historical delivery user feature, and the resource delivery strategy feature are input into the resource effect prediction model to obtain the resource delivery prediction effect of the resource to be predicted output by the resource effect prediction model.

10. A resource delivery effect prediction device, characterized in that: The device comprises: An acquisition module, used to acquire resource materials of the resource to be predicted and determine multimodal information corresponding to the resource materials; A determination module, configured to determine, based on the multimodal information, a similarity between the resource to be predicted and each of the plurality of historical resources; wherein the delivery effect of the historical resources satisfies a preset delivery effect condition; The data processing module is used to input the multiple similarities into a pre-trained resource effect prediction model to obtain the resource delivery prediction effect of the resource to be predicted output by the resource effect prediction model.

11. An electronic device, characterized in that: include: A processor and a memory, wherein the processor is used to execute a resource placement effect prediction program stored in the memory to implement the resource placement effect prediction method according to any one of claims 1 to 9.

12. A storage medium, characterized in that: The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the resource placement effect prediction method according to any one of claims 1 to 9.

Images