Marketing material dynamic optimization method based on AI

Abstract

The invention discloses an AI-based marketing material dynamic optimization method, and relates to the technical field of marketing material optimization, and the method comprises the following steps: 1, determining all put marketing materials contained in a marketing system, and regularly generating a marketing dynamic log corresponding to each put marketing material; 2, determining whether to mark the put marketing materials as to-be-optimized materials or not based on the marketing dynamic logs of the put marketing materials; according to the method, whether all the put marketing materials in the marketing system need to be optimized or not can be dynamically analyzed, and after it is analyzed that the put marketing materials need to be optimized, various features of the materials, feedback conditions in actual marketing and other factors are fully considered through a multi-round comparison screening analysis mechanism, so that the optimization of the put marketing materials is optimized. It can be ensured that the finally selected contrast material and the to-be-optimized material have high relevance and comparability, and a high-quality reference basis meeting actual requirements is provided for optimization work of the to-be-optimized material.

Description

Technical Field

[0001] This invention relates to the field of marketing material optimization technology, and more specifically, to an AI-based method for dynamic optimization of marketing materials. Background Technology

[0002] In today's era of digital marketing, marketing materials are increasingly choosing video formats to meet user needs. With intensifying market competition and constantly evolving consumer demands and preferences, businesses need to continuously invest in high-quality, engaging marketing materials to stand out from competitors, attract the attention of their target audience, and motivate them to make purchases or take action. To achieve this, optimizing existing marketing materials has become a crucial aspect of marketing efforts. This involves analyzing and evaluating various characteristics of the materials (including images, text, and audio) and making targeted improvements based on the analysis results to enhance the overall effectiveness of the marketing materials.

[0003] Currently, while some companies recognize the need to analyze and compare marketing materials to aid in optimization decisions, their methods are often fragmented and simplistic. For example, when selecting materials for comparison, they often simply choose some historically high-performing materials randomly, or compare them based on a single dimension (such as considering only the aesthetics of the image or the word count of the copy), without establishing a scientific and systematic comparative analysis mechanism based on the overall characteristics of the materials. As a result, the selected comparison materials may lack sufficient relevance and comparability with the materials to be optimized, making it difficult to accurately identify the real problems of the materials to be optimized. This fails to provide a truly effective reference for optimization efforts, making it difficult for the optimized materials to achieve the desired marketing results and adapt well to the diverse needs of the market and users.

[0004] Market conditions and user needs are constantly changing, and the performance of marketing materials also changes over time. However, most existing technologies can only perform phased, one-off material analysis and optimization, lacking a dynamic mechanism that can continuously track the performance of launched marketing materials at different stages and make timely optimizations based on changes. This leads to marketing materials potentially losing their effectiveness after a period of time because they fail to adapt to new market conditions, making it impossible to maintain a strong competitive edge and impacting the long-term effectiveness of the company's marketing activities.

[0005] Therefore, this invention proposes an AI-based method for dynamic optimization of marketing materials. Summary of the Invention

[0006] In view of the shortcomings of existing technologies, the purpose of this invention is to provide an AI-based method for dynamic optimization of marketing materials.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] An AI-based method for dynamically optimizing marketing creatives includes the following steps:

[0009] Step 1: Identify all marketing materials already deployed in the marketing system and regularly generate marketing activity logs for each deployed marketing material;

[0010] Step 2: Based on the marketing activity logs of the marketing materials already deployed, determine whether to mark the deployed marketing materials as materials to be optimized;

[0011] Step 3: Once the materials to be optimized are identified, select the relevant materials for comparison.

[0012] Step 4: Obtain the material feature reference index of the related comparison materials. Based on the comparison result between the material feature reference index and the material feature reference threshold index, determine whether to mark the corresponding related comparison materials as pre-selected comparison materials.

[0013] Step 5: Obtain the material feedback index of each pre-selected comparison material. Based on the comparison results between the material feedback index and the material feedback threshold index, determine whether to mark the corresponding pre-selected comparison material as a secondary comparison material.

[0014] Step Six: Obtain the optimization reference index of the secondary comparison materials. Based on the comparison results between the optimization reference index and the optimization reference threshold index, determine whether to mark the corresponding secondary comparison materials as the final comparison materials. Optimize the materials to be optimized by referring to the material image features, material copy features, and material audio features of the final comparison materials.

[0015] Furthermore, the marketing dynamic log includes material number, product type, material image characteristics, material copywriting characteristics, material audio characteristics, material feedback index, and log generation time.

[0016] Furthermore, the marketing dynamic log's material feedback index is determined based on the following method: Select a marketing material that has been launched, periodically obtain the user behavior feature set of the launched marketing material, obtain the material feedback model corresponding to the launched marketing material, use the user behavior feature set as input data for the material feedback model, and the material feedback model outputs the material feedback index of the launched marketing material.

[0017] Furthermore, the user behavior feature set of the marketing materials that have been deployed is obtained periodically in the following way: various feedback data generated by the deployed marketing materials in one cycle are collected, the various feedback data are preprocessed and feature extracted, and integrated into a user behavior feature set in the form of a feature set.

[0018] Furthermore, when a creative material to be optimized is identified, related comparison creative materials are selected: all other marketing creative materials already deployed in the marketing system are marked as marketing comparison creative materials, and the product type targeted by each marketing comparison creative material is obtained. When the product type targeted by the marketing comparison creative material is consistent with or related to the product type targeted by the creative material to be optimized, the corresponding marketing comparison creative material is marked as related comparison creative material.

[0019] Furthermore, the reference index for the features of the related comparison materials is obtained as follows: All previously generated marketing dynamic logs of the related comparison materials are obtained, along with the latest generated marketing dynamic logs of the material to be optimized. The latest generated marketing dynamic logs of the material to be optimized are matched with each previously generated marketing dynamic log of the related comparison materials to form a log comparison group. The marketing feature dynamic similarity index side(dv) for each log comparison group is obtained, where d = 1, 2, ..., D-1, D, where d represents the sequence number of the corresponding log comparison group, and D is the total number of log comparison groups. The marketing feature dynamic similarity coefficient is set to feature(pv), and a marketing feature dynamic similarity threshold index is set. When the marketing feature dynamic similarity index of a log comparison group is greater than the marketing feature dynamic similarity threshold index, the marketing feature dynamic similarity count is incremented by one, and the marketing feature dynamic similarity count is marked as mafea. The reference index refexp for the material characteristics of the related comparison materials is obtained.

[0020] Furthermore, the marketing feature dynamic similarity index of the log comparison group is obtained based on the following method: The material image features of the two marketing dynamic logs in the log comparison group are combined into a material image feature group. A material image comparison model is obtained using this material image feature group. The material image feature group is used as input data for the material image comparison model, and the material image comparison model outputs a material image feature similarity index. Similarly, the material copywriting features of the two marketing dynamic logs in the log comparison group are combined into a material copywriting feature group. A material copywriting comparison model is obtained using this material copywriting feature group. The material copywriting comparison model outputs a material copywriting feature similarity index. Finally, the audio features of the two marketing dynamic logs in the log comparison group are combined into an audio feature group. An audio comparison model is obtained using this material audio feature group. The audio comparison model outputs an audio feature similarity index. The summation and mean of the material image feature similarity index, material copywriting feature similarity index, and material audio feature similarity index yield the marketing feature dynamic similarity index of the log comparison group.

[0021] Furthermore, the optimization reference index for the selected comparison materials is obtained in the following way: Select a selected comparison material, obtain the dynamic feedback index and feedback swing index of the selected comparison material, and label them as [C1, C2], where C1 is the dynamic feedback index of the selected comparison material and C2 is the feedback swing index of the selected comparison material. Obtain the dynamic feedback index and feedback swing index of the material to be optimized, and label them as [D1, D2], where D1 is the dynamic feedback index of the material to be optimized and D2 is the feedback swing index of the material to be optimized. Calculate the optimization reference index of the selected comparison materials using the cosine similarity algorithm.

[0022] Furthermore, the dynamic feedback index is obtained in the following way: Select a marketing creative that has been launched, obtain all marketing dynamic logs generated by the marketing creative that has been launched before, obtain the material feedback index and log generation time of each marketing dynamic log, construct a rectangular coordinate system with the log generation time as the X-axis and the material feedback index as the Y-axis, mark the material feedback index corresponding to the generation time of each log in the marketing dynamic log in the rectangular coordinate system as coordinate points, connect adjacent coordinate points to generate a material feedback curve, draw perpendicular lines from the two ends of the material feedback curve to the X-axis, and mark the total area of ​​the graphic formed by the material feedback curve, the two perpendicular lines and the X-axis as the dynamic feedback index;

[0023] The feedback swing index is obtained by marking the line connecting adjacent coordinate points as the feedback dynamic line segment, summing and averaging the slopes of all feedback dynamic line segments, and taking the absolute value to obtain the feedback swing index.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] The method of this invention can dynamically analyze whether all marketing materials deployed in the marketing system need to be optimized. After analyzing whether the deployed materials need to be optimized, through a multi-round comparative screening and analysis mechanism, it fully considers various characteristics of the materials themselves and feedback in actual marketing, etc., to ensure that the finally selected comparison materials are highly relevant and comparable to the materials to be optimized. This provides a high-quality reference for the optimization of the materials to be optimized, which helps to ensure that the optimized marketing materials reach a better level in all aspects, better meet the needs of the market and users, and enhance marketing effectiveness. Attached Figure Description

[0026] Figure 1 This is a flowchart of the method of the present invention;

[0027] Figure 2 Flowchart for determining the feedback index of marketing dynamic log materials;

[0028] Figure 3A flowchart for obtaining image features for marketing dynamic logs. Detailed Implementation

[0029] Reference Figures 1-3 A method for dynamically optimizing marketing creatives based on AI includes the following steps:

[0030] Step 1: Identify all marketing materials already deployed in the marketing system and regularly generate marketing dynamic logs for each deployed marketing material. The marketing dynamic logs include the material ID (each deployed marketing material in the marketing system has a different material ID), product type (i.e., the product type targeted by the deployed marketing material), material image characteristics, material copy characteristics, material audio characteristics, material feedback index, and log generation time.

[0031] The marketing activity log's creative feedback index is determined based on the following method: Select a marketing creative that has been launched, periodically obtain the user behavior feature set of the launched marketing creative, obtain the creative feedback model corresponding to the launched marketing creative, use the user behavior feature set as input data for the creative feedback model, and the creative feedback model outputs the creative feedback index of the launched marketing creative.

[0032] The user behavior feature set of the marketing materials that have been deployed is obtained periodically in the following way: (from various marketing material deployment platforms) collect various feedback data generated by the deployed marketing materials in a cycle (the types of feedback data include clicks, browsing time, conversion rate, number of favorites, number of shares, etc.), preprocess the various feedback data, extract features, and integrate them into a user behavior feature set in the form of a feature set.

[0033] Each deployed marketing creative corresponds to a creative feedback model, and each creative feedback model is built based on a neural network model. The only difference in the construction process of all creative feedback models is the training data. This embodiment takes deployed marketing creative A as an example to illustrate the construction process of the creative feedback model for deployed marketing creative A: Collect multiple user behavior feature sets of deployed marketing creative A, construct a neural network model, use the user behavior feature sets as training data for the deep learning model, assign a creative feedback index to each training data, the value range of the creative feedback index is (1.0~5.0), the larger the creative feedback index, the better the marketing effect of deployed marketing creative A, divide the training data into training set, validation set and test set in a ratio of 70%:15%:15%, train on the training set, validation set and test set, and finally, construct the creative feedback model for deployed marketing creative A.

[0034] The image features of the marketing dynamic log are obtained based on the following method: the video of the marketing materials has been released is processed by extracting frames at a preset frame extraction interval (the preset frame extraction interval is adjusted according to actual needs), and multiple material image frames are obtained by extracting image features from multiple material image frames through image processing technology. The extracted material image features of the released marketing materials are obtained (material image features include color features, sharpness features, main object features, scene switching features, key element texture features, image element shape features, etc.).

[0035] The characteristics of the marketing dynamic log's material copywriting are obtained based on the following methods: extracting the copywriting content from the marketing materials that have been deployed, extracting all keywords from the copywriting content using natural language processing (NLP) technology, performing feature extraction on the keywords, and obtaining the material copywriting characteristics of the copywriting elements (the basic characteristics of copywriting elements include keyword frequency characteristics, keyword distribution characteristics, keyword intent characteristics, etc.).

[0036] The audio features of marketing dynamic log materials are obtained based on the following methods: extracting the audio content from the marketing materials that have been deployed, extracting the audio features from the image content through audio processing technology, and obtaining the audio features of the marketing materials that have been deployed (the audio features of the materials include time domain features, frequency domain features, timbre features, rhythm features, prosody features, etc.).

[0037] The image comparison model, text comparison model, and audio comparison model are all based on neural network models. The construction processes for these models are similar, differing only in the training data. In this specific implementation, the image comparison model is used as an example to disclose its construction process: A neural network model is constructed, and multiple sets of image features are collected (if constructing a text comparison model, multiple sets of text features are collected; the same applies to the audio comparison model). The neural network model is trained using these image feature sets. For each image... Each feature group is assigned a material image feature similarity index (if constructing a material text comparison model, each material text feature group is assigned a material text feature similarity index, and the same applies to the material audio comparison model). The value range of the material image feature similarity index is (1.0~15.0). The larger the material image feature similarity index, the more similar the two material image features are. Multiple material image feature groups are divided into training set, validation set, and test set according to a set ratio of 5:1:1. The neural network is iteratively trained on the training set, validation set, and test set. After training is completed, the material image comparison model is obtained.

[0038] Step 2: Set the material feedback standard index (the material feedback standard index is a preset index used for comparison with the material feedback index). When the material feedback index of the marketing dynamic log is less than the material feedback standard index, the corresponding marketing material that has been launched will be marked as material to be optimized (when the material feedback index of the launched marketing material is greater than or equal to the material feedback standard index, no further processing will be done).

[0039] Step 3: Once the creative to be optimized is identified, mark all other marketing creatives already deployed in the marketing system as marketing comparison creatives. Obtain the product type targeted by each marketing comparison creative. When the marketing comparison creative targets the same or related product type as the creative to be optimized, mark the corresponding marketing comparison creative as related comparison creative.

[0040] Step 4: Obtain the reference index of material features for related comparison materials, and set the reference threshold index of material features (the reference threshold index of material features is a preset index used for comparison with the reference index of material features). When the reference index of material features for related comparison materials is greater than the reference threshold index of material features, the corresponding related comparison materials are marked as pre-selected comparison materials.

[0041] The reference index for the features of the related comparison materials is obtained as follows: All previously generated marketing dynamic logs of the related comparison materials are obtained, along with the latest generated marketing dynamic logs of the material to be optimized. The latest generated marketing dynamic logs of the material to be optimized are matched with each previously generated marketing dynamic log of the related comparison materials to form a log comparison group. The marketing feature dynamic similarity index side(dv) for each log comparison group is obtained, where d = 1, 2, ..., D-1, D, where d represents the sequence number of the corresponding log comparison group, and D is the total number of log comparison groups. The marketing feature dynamic similarity coefficient is set to feature(pv), where v = 1, 2, ..., V-1, V, where feature(p1) < feature(p2) < ... < feature(pV-1) < feature(pV). Each marketing feature dynamic similarity coefficient corresponds to a range. The marketing feature dynamic similarity index is calculated, with values ​​ranging from (0, side(d1)], (side(d1), side(d2)], ..., (side(dV-1), side(dV)]. When side(dV) ∈ (0, side(d1)], the marketing feature dynamic similarity coefficient is feature(p1). A marketing feature dynamic similarity threshold index is set (this is a preset index used for comparison with the marketing feature dynamic similarity index). When the marketing feature dynamic similarity index of the log comparison group is greater than the marketing feature dynamic similarity threshold index, the marketing feature dynamic similarity count is increased by one (when the marketing feature dynamic similarity index of the log comparison group is less than or equal to the marketing feature dynamic similarity threshold index, no processing is performed). The marketing feature dynamic similarity count is marked as mafea. The reference index refexp for the material characteristics of the related comparison materials is obtained.

[0042] The marketing feature dynamic similarity index of the log comparison group is obtained as follows: The image features of the two marketing dynamic logs in the log comparison group are combined into one image feature group. An image comparison model is obtained from this image feature group. The image feature group is used as input data for the image comparison model, and the image comparison model outputs an image feature similarity index. Similarly, the text features of the two marketing dynamic logs in the log comparison group are combined into one text feature group. A text comparison model is obtained. The text comparison model is used as input data for the text comparison model, and the text comparison model outputs a text feature similarity index. Finally, the audio features of the two marketing dynamic logs in the log comparison group are combined into one audio feature group. An audio comparison model is obtained. The audio comparison model is used as input data for the audio comparison model, and the audio comparison model outputs an audio feature similarity index. The summation and mean of the image feature similarity index, text feature similarity index, and audio feature similarity index yields the marketing feature dynamic similarity index for the log comparison group.

[0043] Step 5: Obtain the material feedback index of each pre-selected comparison material, and set the material feedback threshold index (the material feedback threshold index is a preset index used for comparison with the material feedback index). When the material feedback index of the pre-selected comparison material is greater than the material feedback threshold index, the corresponding pre-selected comparison material is marked as a secondary comparison material (if the material feedback index of the pre-selected comparison material is less than or equal to the material feedback threshold index, no processing is performed).

[0044] Step Six: Obtain the optimization reference index of the selected comparison materials, and set the optimization reference threshold index (the optimization reference threshold index is a preset index used for comparison with the optimization reference index). When the optimization reference index of the selected comparison materials is greater than the optimization reference threshold index, the corresponding selected comparison materials are marked as the final comparison materials. The materials to be optimized are optimized based on the material image features, material copy features, and material audio features of the final comparison materials.

[0045] The optimization reference index for the selected comparison materials is obtained as follows: Select one comparison material, obtain its dynamic feedback index and feedback swing index, and label them [C1, C2], where C1 is the dynamic feedback index and C2 is the feedback swing index. Obtain the dynamic feedback index and feedback swing index of the material to be optimized, and label them [D1, D2], where D1 is the dynamic feedback index and D2 is the feedback swing index. Calculate the optimization reference index for the comparison materials using the cosine similarity algorithm.

[0046] Multiple selection of comparison materials

[0047] The dynamic feedback index is obtained as follows: Select a marketing creative that has been launched, obtain all marketing dynamic logs generated before that marketing creative, obtain the material feedback index and log generation time for each marketing dynamic log, construct a rectangular coordinate system with the log generation time as the X-axis and the material feedback index as the Y-axis, mark the material feedback index corresponding to the generation time of each log in the marketing dynamic log as coordinate points in the rectangular coordinate system, connect adjacent coordinate points to generate a material feedback curve, draw perpendicular lines from both ends of the material feedback curve to the X-axis, and mark the total area of ​​the graph formed by the material feedback curve, the two perpendicular lines, and the X-axis as the dynamic feedback index.

[0048] The feedback swing index is obtained by marking the line connecting adjacent coordinate points as the feedback dynamic line segment, summing and averaging the slopes of all feedback dynamic line segments, and taking the absolute value to obtain the feedback swing index.

[0049] The above method can dynamically analyze whether all marketing materials deployed in the marketing system need optimization. After analyzing whether the deployed materials need optimization, through a multi-round comparison and screening analysis mechanism, it fully considers various characteristics of the materials themselves and feedback in actual marketing, etc., to ensure that the final selected comparison materials are highly relevant and comparable to the materials to be optimized. This provides a high-quality reference for the optimization of the materials to be optimized, which helps to ensure that the optimized marketing materials reach a better level in all aspects, better meet the needs of the market and users, and enhance marketing effectiveness.

[0050] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0051] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0052] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0053] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0054] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0055] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0056] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0057] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for dynamically optimizing marketing materials based on AI, characterized in that, Includes the following steps: Step 1: Identify all marketing materials already deployed in the marketing system and regularly generate marketing activity logs for each deployed marketing material; Step 2: Based on the marketing activity logs of the marketing materials already deployed, determine whether to mark the deployed marketing materials as materials to be optimized; Step 3: Once the materials to be optimized are identified, select the relevant materials for comparison. Step 4: Obtain the material feature reference index of the related comparison materials. Based on the comparison result between the material feature reference index and the material feature reference threshold index, determine whether to mark the corresponding related comparison materials as pre-selected comparison materials. Step 5: Obtain the material feedback index of each pre-selected comparison material. Based on the comparison results between the material feedback index and the material feedback threshold index, determine whether to mark the corresponding pre-selected comparison material as a secondary comparison material. Step Six: Obtain the optimization reference index of the secondary comparison materials. Based on the comparison results between the optimization reference index and the optimization reference threshold index, determine whether to mark the corresponding secondary comparison materials as the final comparison materials. Optimize the materials to be optimized by referring to the material image features, material copy features, and material audio features of the final comparison materials.

2. The AI-based dynamic optimization method for marketing materials according to claim 1, characterized in that, The marketing activity log includes the material number, product type, material image characteristics, material copywriting characteristics, material audio characteristics, material feedback index, and log generation time.

3. The AI-based dynamic optimization method for marketing materials according to claim 2, characterized in that, The marketing activity log's creative feedback index is determined based on the following method: Select a marketing creative that has been launched, periodically obtain the user behavior feature set of the launched marketing creative, obtain the creative feedback model corresponding to the launched marketing creative, use the user behavior feature set as input data for the creative feedback model, and the creative feedback model outputs the creative feedback index of the launched marketing creative.

4. The AI-based dynamic optimization method for marketing materials according to claim 3, characterized in that, The user behavior feature set of the marketing materials that have been deployed is obtained periodically in the following way: collect various feedback data generated by the deployed marketing materials in one cycle, preprocess and extract features from the various feedback data, and integrate them into a user behavior feature set.

5. The AI-based dynamic optimization method for marketing materials according to claim 1, characterized in that, Once a creative to be optimized is identified, select related comparison creatives: mark all other marketing creatives already deployed in the marketing system as marketing comparison creatives, obtain the product type targeted by each marketing comparison creative, and mark the corresponding marketing comparison creative as related comparison creative when the product type targeted by the marketing comparison creative is the same as or related to that of the creative to be optimized.

6. The AI-based dynamic optimization method for marketing materials according to claim 1, characterized in that, The reference index for the features of the related comparison materials is obtained as follows: All previously generated marketing dynamic logs of the related comparison materials are retrieved, along with the latest generated marketing dynamic logs of the material to be optimized. The latest generated marketing dynamic logs of the material to be optimized are matched with each previously generated marketing dynamic log of the related comparison materials to form a log comparison group. The marketing feature dynamic similarity index side(dv) for each log comparison group is obtained, where d = 1, 2, ..., D-1, D, where d represents the sequence number of the corresponding log comparison group, and D is the total number of log comparison groups. The marketing feature dynamic similarity coefficient is set to feature(pv). A marketing feature dynamic similarity threshold index is set. When the marketing feature dynamic similarity index of a log comparison group is greater than the marketing feature dynamic similarity threshold index, the marketing feature dynamic similarity count is incremented by one, and the marketing feature dynamic similarity count is marked as mafea. The reference index refexp for the material characteristics of the related comparison materials is obtained.

7. The AI-based dynamic optimization method for marketing materials according to claim 6, characterized in that, The marketing feature dynamic similarity index of the log comparison group is obtained as follows: The image features of the two marketing dynamic logs in the log comparison group are combined into one image feature group. An image comparison model is obtained from this image feature group. The image feature group is used as input data for the image comparison model, and the image comparison model outputs an image feature similarity index. Similarly, the text features of the two marketing dynamic logs in the log comparison group are combined into one text feature group. A text comparison model is obtained. The text comparison model is used as input data for the text comparison model, and the text comparison model outputs a text feature similarity index. Finally, the audio features of the two marketing dynamic logs in the log comparison group are combined into one audio feature group. An audio comparison model is obtained. The audio comparison model is used as input data for the audio comparison model, and the audio comparison model outputs an audio feature similarity index. The summation and mean of the image feature similarity index, text feature similarity index, and audio feature similarity index yields the marketing feature dynamic similarity index for the log comparison group.

8. The AI-based dynamic optimization method for marketing materials according to claim 1, characterized in that, The optimization reference index for the selected comparison materials is obtained as follows: Select one comparison material, obtain its dynamic feedback index and feedback swing index, and label them [C1, C2], where C1 is the dynamic feedback index and C2 is the feedback swing index. Obtain the dynamic feedback index and feedback swing index of the material to be optimized, and label them [D1, D2], where D1 is the dynamic feedback index and D2 is the feedback swing index. Calculate the optimization reference index for the comparison materials using the cosine similarity algorithm.

9. The AI-based dynamic optimization method for marketing materials according to claim 8, characterized in that, The dynamic feedback index is obtained as follows: Select a marketing creative that has been launched, obtain all marketing dynamic logs generated by the marketing creative that has been launched before, obtain the material feedback index and log generation time of each marketing dynamic log, construct a rectangular coordinate system with the log generation time as the X-axis and the material feedback index as the Y-axis, mark the material feedback index corresponding to the generation time of each log in the marketing dynamic log as coordinate points in the rectangular coordinate system, connect adjacent coordinate points to generate a material feedback curve, draw perpendicular lines from the two ends of the material feedback curve to the X-axis, and mark the total area of ​​the graphic formed by the material feedback curve, the two perpendicular lines and the X-axis as the dynamic feedback index; The feedback swing index is obtained by marking the line connecting adjacent coordinate points as the feedback dynamic line segment, summing and averaging the slopes of all feedback dynamic line segments, and taking the absolute value to obtain the feedback swing index.

Images