Search

CN-121998709-A - Advertisement putting effect optimization method based on big data

CN121998709ACN 121998709 ACN121998709 ACN 121998709ACN-121998709-A

Abstract

The invention relates to the technical field of big data and digital advertisements, and particularly discloses an advertisement putting effect optimization method based on big data. The method constructs an advertising creative gene library, codes text, color matching, layout and background music into gene segments, and realizes iterative evolution and self-optimization of advertising materials through initializing creative population, dynamic feedback evaluation, natural selection, gene crossover recombination and mutation mechanisms driven by environmental pressure. The system updates the fitness score in real time based on the multisource user behavior data, dynamically adjusts mutation probability in combination with external environment factors, introduces a high-quality gene pool to accelerate convergence and ensures compliance. Through the technical scheme, creative fatigue is delayed, advertisement novelty and conversion efficiency are improved, and double optimization of marketing effect and user experience is realized.

Inventors

  • FU XUANHAO

Assignees

  • 西安诚优网络科技有限责任公司

Dates

Publication Date
20260508
Application Date
20260410

Claims (10)

  1. 1. The advertisement putting effect optimizing method based on big data is characterized by comprising the following steps: Constructing an advertising creative gene library, respectively encoding advertising text, visual color matching, page layout and background music into independent gene segments, and assigning a unique identifier and an initial fitness score to each gene segment; Initializing a creative population, randomly selecting a plurality of gene segments from the advertising creative gene library to be combined to form a plurality of initial advertising creative individuals, wherein each individual corresponds to a complete advertising material scheme; Step 3, deploying a dynamic feedback evaluation mechanism, monitoring click rate, conversion rate and stay time of each advertisement creative individual in the creative population in real time based on real user behavior data, and updating fitness scores of each individual according to the click rate, conversion rate and stay time; Step 4, performing natural selection operation, sorting the creative population according to the fitness scores, keeping individuals with scores higher than a preset threshold value as parent populations, and eliminating the rest individuals; step 5, implementing gene cross recombination, randomly selecting two individuals from the parent population, and carrying out segment level exchange on gene segments to generate new child advertising creative individuals; step 6, introducing an environmental pressure driven mutation mechanism, dynamically adjusting mutation probability according to the change of external environmental factors in the generation process of offspring individuals, and randomly replacing part of gene segments or carrying out parameter disturbance; and 7, carrying out iterative evolution and population updating, forming a new generation creative population by the child individuals and the parent individuals which are not eliminated, and repeatedly executing the steps 3 to 6 until a preset termination condition is met.
  2. 2. The method for optimizing advertisement delivery effect based on big data as set forth in claim 1, wherein in the step 1, the construction process of the advertisement creative gene library includes: for the advertisement document gene segment, calling a semantic analysis module to perform word segmentation and part-of-speech tagging on the document, extracting key words, emotion tendencies and sentence pattern structural features, mapping the key words, emotion tendencies and sentence pattern structural features into a numerical sequence with fixed dimensionality, and adding language identification, word number statistics and sensitive word compliance primary screening status bits into the segment; for the visual color matching gene segment, sampling an image material through a color space extraction algorithm, defining a color with the highest pixel occupation ratio in an image as a dominant hue, acquiring a color space coordinate of the dominant hue, and calculating the distribution density of auxiliary colors in a picture, the brightness difference between adjacent color blocks and the hue circle span to form a multidimensional floating point number vector; For page layout gene segments, dividing an advertisement display area into a plurality of equally-divided logic grids by adopting a topological structure description method based on a grid system, defining a starting row index, a starting column index, a occupied row number, a occupied column number and a hierarchical depth of a title, a picture, a button and a price tag in a grid coordinate system, and forming a configuration list of page element arrangement and combination logic; For a background music gene segment, an audio signal is converted from a time domain to a frequency domain through fast Fourier transform, average rhythm strength is calculated, the basic frequency trend of a melody is identified, emotion labeling processing is carried out on music through an audio classification module, and a structural representation comprising a rhythm pulse sequence and a frequency characteristic value is generated.
  3. 3. The advertisement delivery effect optimization method based on big data as set forth in claim 1, wherein in the step 2, the specific process of initializing the creative population includes: Acquiring current computing power resource allocation conditions in real time, wherein the computing power resource allocation conditions comprise memory availability, graphic processor core load and network bandwidth allowance, and dynamically setting the total number of individuals of an initial population according to the computing power resource allocation conditions; Each advertisement creative individual executes service compliance constraint inspection, and a verification module carries out logic consistency audit on the generated individual, wherein the audit content comprises the matching degree of product description in a document and the size of an image area in a page layout, the consistency of a brand tone scheme and a preset brand visual manual specification, and the coverage relation between background music duration and an advertisement display estimated period; and when the generated individuals do not pass any of the service compliance constraint checks, marking and discarding the individuals, and triggering re-extraction logic until the number of the generated compliance individuals meets the preset population scale.
  4. 4. The advertisement delivery effect optimizing method based on big data according to claim 1, wherein in the step 3, the specific process of deploying the dynamic feedback evaluation mechanism comprises: Capturing behavior feedback from a user side through a log stream technology of multi-channel integration, wherein the behavior feedback covers exposure event display, click interaction action, page jump depth and purchase conversion behavior; Establishing a sliding time window mechanism, dividing a preset observation period into a plurality of equal time slices, and endowing corresponding attenuation weight coefficients according to the time distance between the time slices and the current moment, wherein the time slices which are closer to the current moment have higher attenuation weight coefficients; And counting the click frequency and conversion frequency of each advertising creative individual in each time slice, multiplying the click frequency and the conversion frequency with corresponding attenuation weight coefficients respectively, and then executing accumulation operation, so as to calculate the instant performance index of the advertising creative individual, and updating the fitness score of each individual according to the instant performance index.
  5. 5. The advertisement delivery effect optimizing method based on big data as set forth in claim 1, wherein in the step 4, the specific process of performing the natural selection operation includes: Performing fusion calculation on short-term conversion efficiency and long-term user value to generate comprehensive fitness scores, wherein the short-term conversion efficiency is formed by a weighted sum of an instant click rate and an instant conversion rate, and the long-term user value is formed by a weighted sum of user retention probability predicted based on historical data and a repurchase intention score; Dynamically configuring a weighting coefficient between the short-term conversion efficiency and the long-term user value according to the current marketing stage, increasing the weights of the click rate and the stay time in the brand promotion stage, and increasing the weights of the conversion rate and the repurchase intention in the conversion harvesting stage; And calling a sequencing engine to arrange comprehensive fitness scores of all individuals in the population from high to low, setting a selection threshold value to filter the population, wherein the selection threshold value is a preset retention percentage or a dynamic boundary based on a population fitness mean value, retaining individuals with scores higher than the selection threshold value, and transferring a data structure of the eliminated individuals to an offline analysis module to perform failure mode mining.
  6. 6. The advertisement delivery effect optimizing method based on big data according to claim 1, wherein in the step 5, the specific process of implementing the gene crossover recombination comprises: Selecting parent individuals pairwise in the parent population through a random sampling algorithm, and generating a mask vector containing a plurality of binary bits, wherein each bit in the mask vector corresponds to one gene dimension, and the gene dimensions comprise a document, a color matching, a layout and background music; Traversing each bit of the mask vector, and executing fragment level interchange among parent individuals on the corresponding gene dimension when the mask bit is a preset first numerical value; And determining cutting positions in the gene sequence by random numbers in the process of executing fragment level exchange, wherein the cutting positions are set among the phrases for the document genes comprising a plurality of phrases, and generating offspring individuals with new characteristic coupling points by breaking the original gene combination logic.
  7. 7. The method for optimizing advertisement delivery effect based on big data according to claim 1, wherein in the step 6, the specific process of introducing the sudden change mechanism driven by the environmental pressure comprises: External environment factor data is obtained through a real-time crawler interface of the environment perception subsystem, wherein the external environment factor data comprises advertisement putting density of a bid in a media channel, hot topic heat values of a social platform, seasonal consumption trend indexes and macroscopic economic operation fluctuation indexes; carrying out semantic clustering processing on the acquired text environmental data, carrying out normalization processing on the numerical environmental data, and inputting the processed data into a mutation probability mapping function; When the fluctuation degree of the external environment factors is increased, the mutation probability reference value is automatically increased through the mutation probability mapping function, and the random replacement of the gene segment identifiers is performed on the binary stream level of the child individuals, or the random disturbance is performed on the parameters inside the gene segments, wherein the random disturbance comprises the steps of changing the saturation value of the background color of the page and fine tuning the rhythm rate of the background music.
  8. 8. The method according to claim 1, wherein in the step 7, the iterative evolution and population update process involves determination of termination conditions, and the termination conditions include one of the following conditions: The algebra of the current evolution reaches a preset maximum evolution algebra threshold; The performance of the population tends to be in a stable state, and the performance is expressed as that in algebra of a continuous preset number, the improvement amplitude of the average fitness score of the population is lower than the preset minimum deviation; the structural change of the market environment is detected, the environment sensing subsystem identifies industry policy adjustment or sudden public events, and at the moment, the current iteration loop is forcibly terminated and the re-initialization program of the population is triggered.
  9. 9. The advertisement delivery effect optimization method based on big data as set forth in claim 1, further comprising the steps of archiving and applying a high-quality gene pool: Monitoring the fitness score of each individual in real time, and when the fitness score of each individual exceeds the preset proportion of the highest historical record, permanently storing the gene fragments contained in the individual and the corresponding combination parameters in a high-performance read-only database to form a high-quality gene pool; In the subsequent new population initialization process or mutation operation process, an evolution engine is configured to extract gene segments from the high-quality gene pool according to preset priority probability, and an empirical memory mechanism is introduced to shorten the performance optimization period after new advertisement materials are online and accelerate the evolution convergence speed.
  10. 10. The advertisement putting effect optimizing method based on big data as set forth in claim 1, further comprising a process of secondary verification and anomaly monitoring before putting: Before the final advertisement file is distributed to a distribution node, a compliance secondary verification module is called to carry out real-time linkage with a legal and legal database, whether the advertisement file contains an updated forbidden vocabulary is checked, and whether the image content accords with the copyright policy of a target delivery platform is checked; Running an independent abnormal monitoring thread, continuously scanning the characteristic distribution situation of individuals in the population, immediately triggering a fusing mechanism when the situation that the gene segments cause content logical faults, illegal characters or display messy codes due to mutation is found, withdrawing the throwing authority of the corresponding individuals and recording an abnormal log; For dynamic page advertisements containing interaction logic, expanding a layout gene segment into interaction genes containing event triggering logic, defining animation feedback when a user clicks different screen areas, and enabling the interaction genes to synchronously participate in the evolution process of crossing, mutation and selection.

Description

Advertisement putting effect optimization method based on big data Technical Field The invention belongs to the technical field of big data and digital advertisements, and particularly relates to an advertisement putting effect optimization method based on big data. Background Along with the deep fusion of big data and artificial intelligence technology, the digital marketing field has stepped into a new stage of automation and intellectualization, and accurate advertisement delivery becomes a core driving force for improving commercial conversion efficiency. In a large-scale online advertising system, by deep mining and analysis of massive user behavior data, the system can realize accurate image of target audience, and match marketing content with more relevance for different user groups. As a key factor for determining the advertisement reaching effect, the continuous optimization of the delivery strategy directly relates to the utilization rate of the flow resource and the improvement of the overall marketing input-output ratio. The automatic generation and dynamic tuning of the advertising creative materials are key research directions of the current advertising effect optimization. The process generally involves parameterized modeling of multiple creative dimensions, such as document information, visual images, layout configuration, and interaction logic, and performance prediction and real-time screening of massive material combinations using machine learning algorithms. By establishing a feedback closed loop, the system aims to identify a high quality combination with high click rate and high conversion potential from a huge candidate material space to adapt to the changing market environment and user preference. The traditional advertisement optimization algorithm generally faces the serious creative fatigue attenuation problem, materials often fall into optimization stagnation after reaching a local optimal solution, and a static screening mechanism is difficult to cope with aesthetic fatigue of audiences caused by long-term exposure. When facing to a high-dimensional feature space formed by multi-source elements, the existing method has obvious combined explosion problem, and extremely high computational complexity not only consumes a large amount of computational resources, but also limits the capability of the system to find deep cross-boundary creative combinations. The existing optimization model lacks real-time evolution response to dynamic environment pressure, natural selection process with vitality cannot be simulated, the creative material lacks continuous self-iteration capability, long-acting competitiveness is difficult to maintain in strong market competition, and an advertisement putting effect optimization method based on big data is expected. Disclosure of Invention The invention aims to provide an advertisement putting effect optimizing method based on big data, which can solve the problems in the background technology. In order to achieve the aim, the technical scheme adopted by the invention is that the advertisement putting effect optimizing method based on big data comprises the following specific steps: Constructing an advertising creative gene library, respectively encoding advertising text, visual color matching, page layout and background music into independent gene segments, and assigning a unique identifier and an initial fitness score to each gene segment; Initializing a creative population, randomly selecting a plurality of gene segments from the advertising creative gene library to be combined to form a plurality of initial advertising creative individuals, wherein each individual corresponds to a complete advertising material scheme; Step 3, deploying a dynamic feedback evaluation mechanism, monitoring click rate, conversion rate and stay time of each advertisement creative individual in the creative population in real time based on real user behavior data, and updating fitness scores of each individual according to the click rate, conversion rate and stay time; Step 4, performing natural selection operation, sorting the creative population according to the fitness scores, keeping individuals with scores higher than a preset threshold value as parent populations, and eliminating the rest individuals; step 5, implementing gene cross recombination, randomly selecting two individuals from the parent population, and carrying out segment level exchange on gene segments to generate new child advertising creative individuals; step 6, introducing an environmental pressure driven mutation mechanism, dynamically adjusting mutation probability according to the change of external environmental factors in the generation process of offspring individuals, and randomly replacing part of gene segments or carrying out parameter disturbance; and 7, carrying out iterative evolution and population updating, forming a new generation creative population by the child individuals and the pa