Search

CN-122022917-A - Forward automatic optimization method for advertisement materials

CN122022917ACN 122022917 ACN122022917 ACN 122022917ACN-122022917-A

Abstract

The invention relates to the technical field of advertisement material optimization, in particular to a forward automatic advertisement material optimization method. The method comprises the steps of analyzing advertisement request data, identifying a delivery opportunity point with a request but lacking corresponding material coverage, responding to the delivery opportunity point to generate a corresponding material picture, creating an advertisement creative by using the material picture and automatically mounting the advertisement creative to a preset advertisement group, associating a mounting result with the delivery effect data of the advertisement across sources to calculate a delivery effect index, identifying high-quality materials and low-efficiency materials based on the delivery effect index, executing an incremental delivery strategy on the high-quality materials, executing a pause strategy on the low-efficiency materials, and extracting characteristic parameters of the high-quality materials to trigger a new generation flow to perform iterative optimization. The scheme of the invention can improve the efficiency and effect of advertisement delivery.

Inventors

  • Sun Gaozhe

Assignees

  • 钛动科技股份有限公司

Dates

Publication Date
20260512
Application Date
20260206

Claims (10)

  1. 1. A method for forward automatic optimization of advertising material, comprising: Analyzing advertisement request data, and identifying delivery opportunity points with requests but lacking corresponding material coverage; responding to the delivery opportunity point, analyzing the structural configuration file to inquire commodity data, wherein the commodity data at least comprises commodity types and commodity pictures; calling a natural language processing model to generate an advertisement document for commodity data, matching a design template according to commodity types, and injecting the advertisement document and commodity pictures into the matched design template to generate material pictures; cloud storage is carried out on the material pictures, an advertisement platform interface is called, advertisement creatives are created in batches by using uniform resource locators, and the advertisement creatives are automatically mounted to a preset advertisement group; cross-source correlation mounting results and advertisement putting effect data to calculate putting effect indexes; And identifying high-quality materials and low-efficiency materials based on the throwing effect indexes, executing an incremental throwing strategy on the high-quality materials, executing a suspension strategy on the low-efficiency materials, and extracting characteristic parameters of the high-quality materials to trigger a new generation flow to perform iterative optimization.
  2. 2. The method of claim 1, wherein analyzing the advertisement request data to identify delivery opportunity points for which a request exists but coverage of the corresponding material is lacking comprises: And taking the ratio of the number of the covered creatives to the corresponding total number of the requests as a coverage index of the corresponding advertisement spot size, and marking the advertisement spot size with the coverage index lower than a preset threshold and m of the ranking total number of the requests as a delivery opportunity point.
  3. 3. The method of forward automatic optimization of advertising material of claim 2, further comprising, prior to calculating the coverage index: after receiving a user query request, checking whether the local memory cache has the cache data of the same query, if so, directly returning the cache data in the local memory within a preset time period, and if not, executing the query, processing the cache data to calculate a coverage index, storing the coverage index into the local memory cache and setting the cache time.
  4. 4. The method of claim 2, wherein parsing the structured configuration file to query merchandise data comprises: According to the advertisement space size corresponding to the delivery opportunity point, dynamically analyzing a preset YAML configuration file to obtain access parameters of commodity dimensions, and inquiring corresponding commodity data according to the access parameters, wherein the YAML configuration file is used for inquiring database connection information, inquiry table names, partition fields and inquiry fields of the commodity data during inquiry.
  5. 5. The method of claim 1, wherein the matching the design template according to the commodity type comprises: mapping the commodity type to a corresponding design template by using a preset rule; the preset rules comprise mapping table-based matching or semantic vector-based fuzzy matching, and the design templates comprise science and technology sense templates, fashion templates, warm tone templates and templates of solid color backgrounds.
  6. 6. The method for forward automatic optimization of advertisement materials according to claim 1, wherein the cross-source correlation of the mounting result and the advertisement delivery effect data to calculate the delivery effect index comprises: Inquiring a delivery record containing the creative ID of the advertising creative and the delivery time in a database table, and inquiring the delivery effect data of the advertising within a set time range in a relational database by using the creative ID; And combining the delivery record and the delivery effect data of the advertisement to form an effect report, and calculating a delivery effect index, wherein the delivery effect index comprises one or more of click rate, conversion rate and return on investment.
  7. 7. The method for forward automatic optimization of advertisement materials according to claim 6, wherein the process of identifying high-quality materials and low-efficiency materials based on the impression indexes comprises the following steps: And if the click rate of the material picture is smaller than a second preset threshold and the exposure exceeds a preset lower limit, the material picture is an inefficient material, wherein the first preset threshold is larger than the second preset threshold.
  8. 8. The method for forward automatic optimization of advertisement materials according to claim 1, wherein the extracting the characteristic parameters of the high-quality materials to trigger a new generation flow for iterative optimization comprises: extracting characteristic parameters corresponding to the high-quality materials, wherein the characteristic parameters comprise an ID (identity) of an optimal design template and a text style label; regenerating a new advertisement document by taking the characteristic parameters as constraint conditions; And injecting the new advertisement file and the commodity picture requested in real time into an optimal design template to obtain new materials, and mounting the new materials to a preset advertisement group again to realize optimization iteration.
  9. 9. The method for forward automatic optimization of advertisement materials according to claim 1, further comprising a de-duplication process, specifically: And calculating the MD5 value of the material picture, inquiring a database to judge whether the uploaded material picture with the same MD5 value exists, and if so, not repeating uploading.
  10. 10. The method of claim 9, further comprising compressing and optimizing the material picture before cloud storage of the material picture, and storing the material picture in a JPEG format.

Description

Forward automatic optimization method for advertisement materials Technical Field The invention relates to the technical field of advertisement material optimization. More particularly, the invention relates to a forward automatic optimization method for advertisement materials. Background With the rapid development of the digital advertising industry, material management and optimization of an advertising platform have become key factors affecting advertising effects. In advertising scenarios in electronics, gaming, branding, etc. industries, advertisement optimizers need to frequently manage a large amount of advertising material and adjust and optimize according to the impression. In the existing advertisement delivery technology, a search algorithm based on the conventional database query is generally adopted to respond to advertisement requests, namely, a material library is directly queried for matching when the request is received, meanwhile, for the production and the supplement of materials, the production and the supplement of the materials often rely on manual regular checking of an offline report to identify gaps, and the uploading of the materials is manually made. However, the above treatment means have obvious drawbacks in practical application, specifically: Firstly, on the core retrieval and data processing algorithm, a local multi-level cache and concurrency control mechanism adapting to a high concurrency scene is lacking, and when a second-level high-frequency request is faced, huge system load and response delay are caused by direct database interaction; Secondly, the system cannot capture the instantaneous release gap of high request quantity but lack of material coverage in a sharp manner due to the lack of a coverage rate analysis algorithm based on real-time data, so that high-value flow is wasted; Finally, the material generation and the delivery effect data are fractured, an automatic iteration closed-loop algorithm based on real-time feedback is lacked, the material winner and the material updating cannot be realized, and the long-term benefit of advertisement delivery is difficult to ensure. Disclosure of Invention The invention aims to provide a forward automatic optimization method of advertisement materials, which is used for solving the problems of low matching efficiency between massive advertisement requests and material supply and poor advertisement putting effect in the prior art. The invention provides a forward automatic optimization method of advertisement materials, which comprises the following steps: Analyzing advertisement request data, and identifying delivery opportunity points with requests but lacking corresponding material coverage; responding to the delivery opportunity point, analyzing the structural configuration file to inquire commodity data, wherein the commodity data at least comprises commodity types and commodity pictures; calling a natural language processing model to generate an advertisement document for commodity data, matching a design template according to commodity types, and injecting the advertisement document and commodity pictures into the matched design template to generate material pictures; cloud storage is carried out on the material pictures, an advertisement platform interface is called, advertisement creatives are created in batches by using uniform resource locators, and the advertisement creatives are automatically mounted to a preset advertisement group; cross-source correlation mounting results and advertisement putting effect data to calculate putting effect indexes; And identifying high-quality materials and low-efficiency materials based on the throwing effect indexes, executing an incremental throwing strategy on the high-quality materials, executing a suspension strategy on the low-efficiency materials, and extracting characteristic parameters of the high-quality materials to trigger a new generation flow to perform iterative optimization. Optionally, the analyzing the advertisement request data, identifying a delivery opportunity point where a request exists but the corresponding material coverage is absent, includes: And taking the ratio of the number of the covered creatives to the corresponding total number of the requests as a coverage index of the corresponding advertisement spot size, and marking the advertisement spot size with the coverage index lower than a preset threshold and m of the ranking total number of the requests as a delivery opportunity point. Optionally, before calculating the coverage index, further comprising: after receiving a user query request, checking whether the local memory cache has the cache data of the same query, if so, directly returning the cache data in the local memory within a preset time period, and if not, executing the query, processing the cache data to calculate a coverage index, storing the coverage index into the local memory cache and setting the cache time. Optionally, the p